Review log: 29 October 2026

A full review of this organization on 28 and 29 October 2026, with every verified finding fixed in place. This log lists each change: the document and block, what changed, why, and the source. The pass entry in the Research log points here.

Summary

Scope: all 16 documents: AGENTS.md, the five canon documents, the Research log, the Loop and the eight source notes.
Method: 39 review lanes checked factual accuracy against primary sources, internal contradictions, gaps against the stated scope, duplicated concepts, links and docrefs, outdated statements, unclear writing, structure, missing citations, dashes and public safety. Two independent skeptics checked every finding (a source check and an editorial check), with a third on a split. One fixer per document applied the surviving findings and one checker per document reread every changed block. A completeness pass then snapshotted the live documents and ran four critics (coverage, drift between documents, a sweep of all 517 cited URLs and every docref, and an adversarial recheck of the riskiest fixes), followed by a skeptic, a fixer and a checker per document.
Result: 440 findings raised, 392 survived verification, 382 fixed. The completeness pass raised 32 tasks; 22 survived its skeptics and 19 were fixed. Checkers corrected 21 problems in fresh edits. 11 items are open for the owner (below). 409 logged changes in all.
Mechanical state after the pass: no em or en dashes in any document; every docref inside the organization resolves; the one docref leaving it (AGENTS.md to the UX and UI organization) resolves and is public.
Legal notes stay orientation, not legal advice. Every legal correction was made against the official text (gesetze-im-internet.de, the Commission's AI Act pages) and cites it with its read date.

Open for the owner

These were not changed, or were changed and need a final look:

Loop scope. The legal recheck Loop covers two blocks, but other statute-based blocks also go stale on a calendar: the activation note's German price display checklist and statute register, and the content design note's BFSG block (now dated "Last checked: 29 October 2026"). Decide whether to widen the Loop; its sentence that its two blocks are the parts that go stale on a calendar then needs the same change.
Loop owner field. The Owner cell now names Oddlynew; the Loop's metadata owner still names a person. Decide whether it should name Oddlynew too.
Research pass rule. AGENTS.md says "one agent, one model, one session", while Pass 1 and this review ran many parallel lanes under one coordinating session. Decide whether the rule should read "one coordinating session owns each pass, and its log entry names every agent and model".
Duplicated method. The positioning note repeats the step-by-step method and templates that From product strategy to page copy owns (about 85 blocks). Proposal: replace them with one pointer, after retargeting the one docref that cites them.
Summary label in source notes. Entries now end with different labels ("Summary:", "What it gives:", "What it offers:", "Why it matters:"). Pick one for all notes.
Deep fake reading. The AI note now lists four disclosure duties, including Art. 50(4) deep fakes, and says a misleading AI product or food photo can count as one, following the Commission's guidelines. Principle 32 follows it with "can count". Confirm the reading.
Default tone. The cited NN/g study found casual, conversational and moderately enthusiastic copy did best, so the default changed from "matter-of-fact" to moderate enthusiasm in the content design note and Principles. Confirm the canon default.
Legal corrections, final look. Applied against the statute text: § 11 PAngV covers price reductions on goods only, and a service plan shows a former price only if it was really charged (Principle 43, persuasion note, activation note, Myths); § 312k BGB needs a confirmation page after the cancel button (persuasion note); the § 356a BGB withdrawal function (persuasion note, activation checklist); the Art. 50(2) grace to 2 December 2026 for generators already on the market; the role line that names AI under Art. 50(5); AI text drafted before 2 August 2026 and published later; "no card" only where no card is asked for (activation checklist); published volume steps instead of "ab" prices marked as house practice, not law.
Persuasion section 5. Its list of failed or misapplied effects duplicated Myths and had started to drift, so it now points to Myths. One skeptic preferred keeping the list; the change is reversible from the document's history.
A Myths row on mentioning AI. "Naming AI always costs trust, so never mention it" was not added because the skeptics split. Add it if Myths should hold it too.
Readability target. The playbook and Principles say "grade 7 to 9"; two source notes say "about grade 8". Compatible; the canon owner may align the wording.

Decided 30 October 2026 (owner)

Loop scope: widened. The legal recheck Loop now also covers the activation note's German price display checklist and the statute rows (E1 to E9) of its evidence register, and the content design note's BFSG block. Its "go stale on a calendar" sentence, its Outcome, Measure, Writebacks and Authority cells and steps 2 and 5 now speak of all the legal blocks. Same trigger, no outside action. The AGENTS.md route line names the four notes.
Loop owner field: Oddlynew. The Loop's metadata owner now says Oddlynew, as the Owner cell does; the measure's name says "oldest" instead of "older".
Research pass rule: changed. AGENTS.md now says "one coordinating session owns each pass, and its log entry names every agent and model".
Duplicated method: replaced by a pointer. The positioning note's sections (a) and (b), 81 blocks, gave way to one short section, Method and templates, that points to the eight steps and Templates in From product strategy to page copy. The one positioning rule that step 8 there cites stays in the note. The Templates attribution line now points to the note itself. Rows in the positioning table below that cite removed blocks now point to blocks that no longer exist; the document's history keeps them.
Summary label: "Summary:". It replaced "What it gives:" in 20 entries of the copywriting canon note and "Why it matters:" in 14 entries of the positioning note and 13 of the content design note. No "What it offers:" was left.
Deep fake reading: confirmed. Principle 32 keeps "can count". The documents carried no "until confirmed" hedge, so nothing else changed.
Default tone: confirmed. Moderate enthusiasm stays the default. The documents carried no pending-confirmation hedge, so nothing changed.
Legal corrections: they stay applied. One item stays open, below.
Persuasion section 5: stays removed in favour of Myths.
Myths row on mentioning AI: stays out.
Readability target: "grade 7 to 9" everywhere. Changed in Principle 14, the web reading note and the content design note. Figures quoted from sources, such as NN/g's grade 8, are unchanged.

Still open: The owner or counsel may give the legal corrections one more look when convenient.

Outside this organization

No fix was found that belongs in another organization.

Not checked

87 cited URLs answer scripted fetches with a bot block (HTTP 403): journal publishers (SAGE, Oxford, PNAS, ScienceDirect, Wiley, ACM, SSRN), CXL, Medium and the OpenAI help center. They are reachable but their wording was not reread in this pass.
EUR-Lex answered with a bot challenge, so the EU wording of DSA Art. 25 and Directive (EU) 2023/2673 Art. 11a was confirmed only through search results. The German § 312k and § 356a BGB texts were read directly.
The ChatGPT plan eligibility for custom MCP connectors and the Plus message limit rest on reviewers' reads and third-party reports; the OpenAI help center blocked every fetch, and the cited page's address names a different model version than the text.
The in-force date of § 356a BGB (19 June 2026) rests on the Federal Law Gazette as the reviewers read it, not on gesetze-im-internet.de, which shows the text without the date.
Two hosts timed out rather than failing: ai.stanford.edu (an author copy of the cited Kohavi 2007 paper exists) and jtbd.info (archived copies exist). Neither is confirmed dead.
Claims that rest on books with no accessible primary text: Ries and Trout, The Mom Test, Competing Against Luck, Schwartz's sophistication stages, Ogilvy on Advertising.
Studies held from earlier passes and not refetched: Hagger et al. 2016, Klein et al. 2014, Open Science Collaboration 2015, Johnson and Goldstein 2003, Dietvorst et al. 2015 and 2018, Crolic et al. 2022, Jakesch et al. 2019; vendor figures from Ahrefs, Profound, Semrush, SE Ranking and Qubit.
Pages that return HTTP 200 were checked for "not found" titles only, not against each claim.

Improvement pass, 30 October 2026

A single-agent pass for the organization's two readers, a person writing product copy and an agent asked to write or review it: routing by task, a glossary, examples where a rule had none, short entry blocks on the long source notes, and one owner pointer. No fact, figure, legal statement or citation changed. 19 changes; the last five, added later the same day, replace the web reading note's duplicate testing playbook and stats list with pointers to their owners.

Document	Block	What changed	Why
AGENTS.md	Find the right note by task	New section after Start here: one intro line and a table of 18 tasks (landing page, use-case page, pricing, limit message, interface text, first run, email, AI disclosure, privacy, persuasion, German, agent-facing text, voice, AI drafting, review, testing, citing a number), each routed to its Principles section and one deeper block.	Start here and Routes routed by document, not by task; a newcomer could not reach the right note in two hops.
AGENTS.md	How to work here, step 1	Step 1 now links the new task table.	The step named "the section your task needs" without saying where to find it.
Principles	22. Interface text is copy	Added an illustrative example: a button, an error message and an empty state.	Rule without an example; the most common interface task.
Principles	36. Make "no" reachable	Added an illustrative disqualifier and prerequisite line.	Rule without an example.
Principles	44. Limit messages have four parts	Added an illustrative limit message showing all four parts and one number per sentence.	Rule without an example.
Principles	Words used here	New closing section: a table of 18 terms (awareness stage, hero, fold, false floor, marketese, social proof, anchoring, dark pattern, message house, PR/FAQ, switch interview, language bank, claims register, agent reader, answer engine, fit check, five-second test, A/B test), each with a one-line meaning and a link to its template or principle where one exists.	The rules use practitioner jargon a writer may not know, with no definitions anywhere in the organization.
Principles	Pointer after Rules at a glance	One line naming the terms and linking Words used here.	The glossary sits at the end; readers meet the terms in Rules at a glance.
Source: Copywriting canon	In short	New block after Status: the principles the note backs, and the section for headlines, disagreements and first reading.	The note runs to 35,000 to 50,000 characters and opened with provenance only; a writer had no quick way to the rules and the section for the task.
Source: Psychology of persuasion	In short	New block after Status: principles 19 to 21, then the scorecard, pricing-page psychology and legal context.	The note runs to 35,000 to 50,000 characters and opened with provenance only; a writer had no quick way to the rules and the section for the task.
Source: Positioning	In short	New block after Status: principles 1, 2, 4 and 24, then the method pointer, customer research and message hierarchy.	The note runs to 35,000 to 50,000 characters and opened with provenance only; a writer had no quick way to the rules and the section for the task.
Source: Content design	In short	New block after Status: principles 14, 22, 23 and 25, then microcopy patterns, the voice chart, German specifics, AI drafting and the 25 working rules.	The note runs to 35,000 to 50,000 characters and opened with provenance only; a writer had no quick way to the rules and the section for the task.
Source: How people read the web	In short	New block after Status: principles 6, 7, 9, 10, 15 and 50, then reading, homepage and pricing evidence, testing, search and accessibility.	The note runs to 35,000 to 50,000 characters and opened with provenance only; a writer had no quick way to the rules and the section for the task.
Source: Activation, plan and limit copy	In short	New block after Scope: principles 40 to 49, then the use-case checklist, developer copy, plan and limit sections and the German price checklist.	The note runs to 35,000 to 50,000 characters and opened with provenance only; a writer had no quick way to the rules and the section for the task.
Source: How people read the web	Myths owns the verdicts	One line at the top of "Commonly cited stats" saying Myths owns these verdicts and holds where the two differ.	The same claims carry different labels in the two places (for example "Half true" against "overgeneralized", and the Ogilvy headline claim as "Dated and unsourced" against "misapplied").
From product strategy to page copy	Preference tests are tie-breakers only	New step in the low-traffic testing playbook, citing the web note's block on other qualitative sources.	The web note's duplicate playbook had this step and the canonical one did not. Its other extra step, mining customer language, was not added: step 1 of the eight steps already covers it.
From product strategy to page copy	Instrument big effects only	The web reading note citation now points to 6. Testing copy properly.	It cited a block of the duplicate playbook, which was removed.
Myths	Web reading and testing table	Six evidence cells that cited the web note's stats list now carry that list's sources themselves: the Microsoft Canada report, "no primary source", Mehrabian's quote and URL, Porter 2003 and its URL, Fogg et al. 2002, Lindgaard et al. 2006. The Ogilvy headline row gained the CXL 2018 eye tracking figures, citing the web note's evidence section.	The list they cited was removed. All 17 of its claims were already Myths rows, so no row was added. Not carried: an Optimizely 12 percent win rate that gave no source, and Loranger 2017 as a second citation on plain language.
Source: How people read the web	Testing playbook pointer	The 11-block source draft of the testing playbook under its heading gave way to one pointer to the canonical playbook and to 6. Testing copy properly. The Status line no longer says the note keeps the draft.	One owner per method, as for the positioning note's method. Older rows of this log that cite the removed blocks resolve through the document's history.
Source: How people read the web	Stats pointer	The 17-claim stats list gave way to one pointer to Myths, Web reading and testing; the earlier "Myths owns the verdicts" line is now that pointer.	Myths owns the verdicts; two lists with different labels drift.

Left for the owner

Email and sales copy. The organization names emails and sales material in its audience, but its email guidance is two bullets in the content design note. The queued pass on emails and sales material would close the gap.
Not covered by this pass. The AI disclosure note and the agent reader note got no In short block, since both open with a summary section (Bottom line; Source synthesis). No rule stated as a story was found in AGENTS.md, Principles, From product strategy to page copy, the playbook, Myths or Conflicts; the source note bodies were not reread in full.

Changes by document

AGENTS.md

10 changes.

Block	What changed	Why	Source
This organization gives method and evidence; a ...	'(template in the playbook)' now docrefs Playbook; 'its claims register' gains '(template)' docref to Strategy to page copy.	Template pointers were prose with no link.	Playbook and Strategy to page copy, live blocks read 29 October 2026
After real copy work, follow Learn from ...	'"Learn from corrections"' is now a docref to Playbook; numbered list item kept.	Section named in prose with no link.	Playbook, live block read 29 October 2026
From product strategy to page copy	'every reusable template' to 'the reusable strategy and messaging templates'.	Contradicted the next route, which places the voice and tone template in the Playbook.	AGENTS.md; MentionAI, AgentReader, Activation checklists
Legal notes cover EU and German law ...	'Legal notes are orientation, not legal advice;' to 'Legal notes cover EU and German law only and are orientation, not legal advice; for any other market, check local law.' Purpose AGENTS.md unchanged.	No jurisdiction stated for a public organization whose legal floors are all EU or German.	Skeptic revisedFix (two votes); skeptic search found no non-EU/German legal floor
Copywriting canon	Appended '; frameworks and headline formulas, where the schools disagree, and working rules for website copy.'	Route listed authors only and hid Parts 3 to 5.	Canon note headings Part 3, Part 4, Part 5 (local copy lines 319, 345, 376)
Maintained by Oddlynew through research passes recorded ...	'Maintained by Oddlynew, with research by parallel research agents and a separate coverage review.' to 'Maintained by Oddlynew through research passes recorded in the Research log.' Rule AGENTS.md unchanged.	Ownership line narrated the founding multi-agent pass and contradicted the one-agent pass rule.	Editorial skeptic revisedFix; Research log
Positioning, customer research and product narrative	Inserted 'voice-of-customer research' after Torres and 'message house and messaging hierarchy' after Sierra.	Route omitted Positioning section 4 and the voice-of-customer section.	Positioning note headings 'Voice-of-customer research as copy input', '4. Message hierarchy and messaging documents'
Activation, plan and limit copy for agent-operated ...	Inserted 'a use-case page checklist, a first-run study design' after 'first-run and empty-state copy'.	Route omitted Activation A2 and A5.	Activation note headings A2 and A5
Content design, UX writing, voice and tone	Inserted 'plain language and readability formulas' after 'microcopy patterns'; 'voice and tone' not repeated.	Route omitted section 5; the link label already names voice and tone.	content design note '5. Plain language and readability'
Sources: one note per body of work, ...	Completeness pass: Applied the skeptic's revisedFix. The Sources lead-in now ends its description with a period, adds 'Where a note's rules differ from Principles, Principles holds; where its method or template differs from From product strategy to page copy or the Copy review playbook, those hold.' and introduces the nested list with 'The notes:'. The bullet level 1 attributes and the key stay the same, and the nested list blocks were not touched.	Persuasion, Activation and MentionAI carry rule lists with no precedence line. One line in AGENTS covers every source note, including future ones.	skeptic revisedFix, task 0

Principles

30 changes.

Block	What changed	Why	Source
Status: canonical research synthesis for this organization; ...	Status: ', first pass on 29 September 2026' to '; the Research log (docref Research log) holds its passes'.	A pass date in canon reads as the age of the current rules; history belongs in the Research log.	AGENTS.md; Research log
Each principle names its evidence class. Experiment: ...	'test anything rated mixed or weak' to 'test anything a source note rates mixed or weak (evidence ratings persuasion note, replication scorecard persuasion note)'.	The mixed or weak scale is not this document's evidence classes; the reader could not find it.	persuasion note, persuasion note
6. Clarity first. Answer what it is, ...	'visitors leave within 10 to 20 seconds unless...' to 'visitors often leave within 10 to 20 seconds, and pages with a clear value proposition hold them far longer'; Limit names the 10-second advice as the inference.	Dropped 'often' and the Limit referred to guidance the body never stated.	web reading note, Nielsen 2011 row
7. Front-load and write for scanning. People ...	Evidence now ties each sample to its claim: 2005 logs of 25 users (percentages), 80-participant link test (first two words), 500+ eye tracking (scanning).	'500+ participants' lent weight to a percentage that comes from a small modelled log sample.	web reading note (Nielsen 2008, 2009, Moran 2020 rows)
10. Show the price, or at least ...	'hidden costs are the top reason' to 'extra costs (shipping, tax, fees) are the top reason shoppers give for abandoning an e-commerce checkout (40 percent), and 12 percent leave because they cannot see the total up front'.	Baymard's top reason is extra costs too high; the unclear total is a separate 12 percent reason.	https://baymard.com/lists/cart-abandonment-rate (finding, read 29 October 2026); web reading note
41. Say what happens at the limit ...	'Surprise costs are the top reason for abandonment.' to 'Extra costs are the most named reason for abandoning a checkout.'	Same swap of extra costs for surprise costs as Principle 10.	Activation (a table cell) E12 (live, 29 October 2026); Baymard as above
14. Plain words for every reader. Needless ...	Kept the grade 7 to 9 and 12 targets and added evidence: published reading-level targets from NN/g, Shopify and Hemingway (content design note, content design note).	The target had no source beside it; a related finding's rewrite to 'about grade 8' was not applied because skeptics on both findings preferred the citation and Canon's fix already points to the 7 to 9 target.	content design note, content design note
15. Cut marketese. Objective copy measured 27 ...	'they discount everything after a superlative' to 'a superlative leads readers to discount everything else you say (Hopkins)', with a docref to canon note.	The 51-user study does not measure discounting; it is Hopkins's practitioner claim.	canon note
16. Length follows awareness and stakes. A ...	'disagreements' docref retargeted from the Canon's H1 canon note to Part 4 canon note.	The docref landed at the top of a long note instead of the section that resolves long against short copy.	canon note
23. Voice is constant, tone flexes. Default ...	'Default to casual, serious, respectful and matter-of-fact' to 'Default to casual, serious and respectful, with moderate enthusiasm'.	NN/g found matter-of-fact samples rated less friendly; casual, moderately enthusiastic tones performed best.	Spill a related finding (content design note owner fix)
27. Write for translation from the start, ...	Added a docref 'length expansion' to content design note beside the localization docref.	The 40 percent figure sits under Length expansion, not the cited localization block.	content design note
21. Never use persuasion you would not ...	Legal list 'section 11 PAngV' to 'section 5 UWG, section 11 PAngV for goods'.	Section 11 PAngV covers reductions on goods only; services fall under section 5 UWG.	https://www.gesetze-im-internet.de/pangv_2022/__11.html and https://www.gesetze-im-internet.de/uwg_2004/__5.html, read 29 October 2026
29. When you disclose AI, state roles, ...	'naming the human step removes most of the penalty' to 'shrinks the penalty'; example 'Drafted from your notes, checked by you' to 'AI draft from your notes, checked by you'; Limit kept.	No study supports 'removes most'; the old example disclosed nothing about AI.	AI note, (a table cell); Schilke and Reimann 2025 and Altay and Gilardi 2024 as cited in the findings (read 29 October 2026)
32. Meet the legal floor where it ...	'From 2 August 2026' to 'Since 2 August 2026'; added 'a provider whose generator was on the market before 2 August 2026 has until 2 December 2026' (docref AI note); evidence 'read 29 September 2026' replaced by 'as amended by the Digital Omnibus on AI ... kept current in the legal floor' plus the Service Desk URL.	The Art. 50(2) marking duty ignored the Digital Omnibus transition to 2 December 2026.	https://ai-act-service-desk.ec.europa.eu/en/ai-act/timeline/timeline-implementation-eu-ai-act, read 29 October 2026; AI note (Regulation (EU) 2026/1744). EUR-Lex returned an empty page to the fetch.
37. Ask a visiting agent only what ...	'Imperatives are also ineffective: models ... discount them' to 'Imperatives gain nothing: models trained to treat tool output as untrusted give it no authority (inference)'.	It stated an untested effect as fact.	agent reader note (live, owner fix applied 29 October 2026)
39. Discovery by answer engines comes mostly ...	'cite earned media, community threads and top-ranked pages' to 'cite earned media and community threads more than brand pages, and engines differ: Google's AI Overviews draw partly on pages that already rank, ChatGPT search mostly on pages ranked 21 or lower'.	It merged engines that behave in opposite ways, from a Google-only 2025 snapshot.	https://www.semrush.com/blog/ai-search-seo-traffic-study/ and https://ahrefs.com/blog/ai-overview-citations-top-10/ (finding, read 29 October 2026); agent reader note, agent reader note
40. Lead each plan with the number ...	'about 60 edits a day' to 'about N edits a day'; unit list to 'A system's own names for its limits, and raw counts such as tokens'; 'AI vendors have stopped publishing...' to 'the AI chat plans in the register state limits as multiples of a usage window or as message counts per window, not as token counts'; docref split into plan principles and evidence register.	One product's internal vocabulary and number in public canon, and an unsourced 'have stopped' market claim.	activation note, activation note, (a table cell) E20/E21 and activation note (live, 29 October 2026)
43. One total price, with its unit ...	'price reductions show the 30-day lowest price' to 'a price reduction starts from a price really charged, not one charged only for an unreasonably short time, and a reduction on goods also shows the lowest price of the previous 30 days'; evidence adds the § 11 goods scope and UWG section 5 with URLs.	§ 11 PAngV covers reductions for goods only; § 5(5) UWG governs reductions generally.	https://www.gesetze-im-internet.de/pangv_2022/__11.html and https://www.gesetze-im-internet.de/uwg_2004/__5.html, read 29 October 2026
44. Limit messages have four parts: what ...	'One number per line' to 'One number per sentence'; docref split into activation principles activation note and tensions activation note.	The docref missed the block that holds the evidence; 'line' contradicted the source's example.	activation note, activation note, activation note (live)
46. When the customer's own agent does ...	Example 'Bring the agents you use; your organization keeps the work' to 'Use the agent you already work with; [Product] keeps what it makes'.	Public canon quoted one product's shipped tagline.	AGENTS.md; activation note (live)
50. Test comprehension with real readers before ...	'testing, playbook' docref from web reading note to web reading note (testing) and Strategy to page copy (playbook).	It cited the duplicate playbook instead of the canonical one.	Strategy to page copy; web reading note
51. Hold every copy opinion loosely. Even ...	'across 6,700 experiments' to 'across about 2,600 categorized retail and travel experiments'; 'a bias-free meta-analysis' to 'a vendor meta-analysis that covers all its tests, not only winners'; docref split into experiments web reading note and landing web reading note.	The averages come from the categorized subset, and 'bias-free' overstated the source.	WebReading Qubit row ('roughly 2,600 categorized tests (from 6,700 in total), retail and travel') and web reading note (live, 29 October 2026)
52. Learn from every correction. A founder's, ...	'A founder's or reader's correction ... product organization' to 'A founder's, reviewer's or test's correction ... product's organization'; docref Playbook to Playbook.	The docref landed on the Playbook H1, and the trigger list did not match the Playbook.	Playbook
39. Discovery by answer engines comes mostly ...	Checker correction: Split into two sentences ('... ranked 21 or lower. On owned pages, ...') and added 'in the top 10' per agent reader note.	After the fix, one sentence ran the engine comparison and the owned-page advice together with a semicolon, so 'on owned pages' read as a third engine. 'Pages that already rank' was vaguer than the source (top 10).	Checker reread of the live block
7. Front-load and write for scanning. People ...	Completeness pass: Principle 7: 'People read about 20 to 28 percent' -> 'People have time to read about 20 percent of a page's words, at most 28 percent'	28% is the ceiling, 20% the likely figure; matches web reading note and Persuasion	https://www.nngroup.com/articles/how-little-do-users-read/, read 29 October 2026 (skeptic revisedFix, task 1)
14. Plain words for every reader. Needless ...	Completeness pass: Principle 14: 'grade 7 to 9' -> 'about grade 8' for general readers	Both owners (WebReading rule, ContentDesign rule 25 content design note) now say about grade 8	https://www.nngroup.com/articles/legibility-readability-comprehension/, read 29 October 2026
29. When you disclose AI, state roles, ...	Completeness pass: Principle 29 Limit: 'a small disclosure penalty remains' -> 'the trust cost to the person who discloses remains under every framing tested (pooled d = 0.81, Schilke and Reimann)' with new docref to AI note 'bottom line'	d = 0.81 is not small; the Limit now names the discloser penalty, separate from the content penalty the body discusses	AI note; https://par.nsf.gov/servlets/purl/10597787, read 29 October 2026 (skeptic revisedFix, task 0)
32. Meet the legal floor where it ...	Completeness pass: Principle 32: added the Art. 50(4) first-subparagraph deep fake duty (AI image, audio or video that resembles real people, objects, places or events and would falsely appear authentic says it is AI-generated or manipulated; a misleading AI product or food photo 'can count'), docref AI note 'deep fakes'	Canon rule was narrower than its own legal floor; product-photo reading kept as 'can count' because MentionAI marks it with a flag for the owner's final look	https://ai-act-service-desk.ec.europa.eu/en/ai-act/article-50, read 29 October 2026 (official text confirmed: 'Deployers of an AI system that generates or manipulates image, audio or video content constituting a deep fake, shall disclose ...')
34. Write so each claim survives one ...	Completeness pass: Principle 34 docref label 'principles, specs' -> 'source synthesis' (agent reader note) plus new 'specs' docref to agent reader note	AgentReader section 1 was renamed 'Source synthesis'; specs live in section 2	AgentReader live headings agent reader note, agent reader note
46. When the customer's own agent does ...	Completeness pass: Principle 46 example -> '("Use the agent you already work with. You keep everything it makes.")'	Matches Activation P5 activation note; generic placeholder example, no product promise changed	activation note

From product strategy to page copy

23 changes.

Block	What changed	Why	Source
Status: canonical method for this organization, first ...	Plain-text pointer to the nonexistent AGENTS.md section "Where product application lives" now a docref to the AGENTS.md H2 "How a product organization uses this" (AGENTS.md).	The named heading does not exist in AGENTS.md.	AGENTS.md 01a0ed9d AGENTS.md, live read 29 October 2026
Interview recent switchers and churners about specific ...	"8 to 12 switch interviews per segment ... Stop when pains and workarounds repeat" to "switch interviews in rounds of about ten, stepping back after each round; stop when the stories repeat" with a docref to the new Positioning interview-count principle positioning note; Mom Test and Torres keep the positioning note docref.	Count was uncited, cited a bare heading, and "per segment" multiplied the source's guidance.	positioning note (Moesta and Spiek, https://jobstobedone.org/radio/the-mattress-interview-qa/, read 29 October 2026), as written by the Positioning fixer today
Output: the four forces per segment (push, ...	Added Moesta docref (positioning note) after "the buying timeline".	Four forces and buying timeline carried no evidence.	positioning note
Stress-test through prose before pages: a PR/FAQ ...	Anchor "Working Backwards, six-pager" split into Working Backwards (positioning note) and six-pager (positioning note).	The memo practice lives in a separate block from the one cited.	positioning note, positioning note
Capture the result in the positioning statement ...	Appended docref "after Neumeier's onlyness test" to positioning note, the Into copy block that holds the three-qualifier line.	The test carried no evidence; the skeptic noted the heuristic is the note's application, not Neumeier's words.	positioning note and positioning note
Frame the promised land as the user's ...	Appended Graham docref (positioning note) to "Pitch it live to real prospects before it becomes a page."	Sentence was uncited and read as Sierra's.	positioning note, positioning note
One message house per product: a roof ...	"two or three proofs under each pillar, and a base of glossary and voice rules" to "a foundation of two or three proofs under each pillar; the glossary and voice rules sit beside the house" (three pillars kept per skeptic revisedFix).	Cited source makes proof points the foundation; glossary and tone are optional extras.	positioning note; https://umbrex.com/resources/frameworks/marketing-frameworks/message-house-framework/ (finding evidence, read 29 October 2026)
Message house. Roof: one sentence of category ...	"Under each: two or three proofs ... Base: glossary and voice rules." to "Foundation: two or three proofs under each pillar ... Beside the house: glossary and voice rules."	Template repeated the base drift.	positioning note
Keep one glossary for code, UI, docs ...	"usually a sign" to "often a sign".	Overstated the cited synthesis, which says often.	positioning note / positioning note
Write the hero last. Reuse language-bank phrases ...	"working rules" docref retargeted from Canon H1 canon note to Canon Part 5 canon note (the Playbook half belongs to the Playbook fixer).	Docref landed on the Canon title, not the rules.	canon note
Ugly first draft, then edit in passes ...	Anchor "Handley, Graham" to "Handley" plus docrefs Graham (canon note) and Bezos memos (positioning note).	One docref covered only Handley; Graham and "let it rest" were unreachable.	canon note, canon note; positioning note
See the testing playbook below. When copy ...	"fails in tests twice" to "fails in tests repeatedly" with a Dunford docref to positioning note (skeptic revisedFix).	"Twice" was an uncited number that contradicted the thesis and the source.	positioning note
Two cases the source notes do not ...	Heading "Two cases the general literature does not cover" to "Two cases the source notes do not cover"; new line Strategy to page copy under it: "Evidence class for both: synthesis (docref Principles), a hypothesis until tested on the real audience."	Overclaim (buying groups are widely covered) and unmarked unsourced claims.	https://www.gartner.com/en/sales/insights/b2b-buying-journey (finding evidence, read 29 October 2026); Principles
Message testing panel with verified target buyers ...	"10 to 20 respondents usually reach saturation" (docref web reading note) to "about 15 respondents per narrow audience reach saturation, Wynter's default test size" with docref (a table cell) and the Wynter URL; matches web reading note as fixed today.	Number contradicted the source and cited a block without it.	https://wynter.com/research-playbook, read 29 October 2026 ("15 is the default and fine for most cases", "Qualitative saturation point"); WebReading (a table cell)
Moderated usability: three rounds of five people ...	Appended Nielsen docref (web reading note).	Number with no evidence.	web reading note
Instrument big effects only: scroll depth, CTA ...	Appended docref "web reading note" to web reading note, without the "practitioner judgment" label (two skeptics against that label).	Step carried no evidence route.	web reading note
If you A/B test: pre-register metric, minimum ...	"distrust any lift above 10 to 20 percent (Twyman's law)" to "distrust any lift far above the 0 to 4 percent that real tests average (Twyman's law, Kohavi et al.)"; anchor to "Qubit, Berman et al., Kohavi"; added docref "Evan Miller, Kohavi" (web reading note) after "do not peek".	Unsourced threshold, Twyman's law misattributed, peeking and full weeks uncited; the owner removed the threshold from web reading note today, so canon now matches.	WebReading (a table cell) (0 to 4%, Kohavi), web reading note, web reading note; web reading note live read 29 October 2026
Simulated persona panels are an exploration tool, ...	Conflicts docref retargeted from H1 Conflicts to entry Conflicts, anchor "Conflicts, simulated persona panels".	Docref landed on the Conflicts title, not the named entry.	Conflicts
Paraphrased from their authors; attribution in the ...	"attribution in the source notes" to a docref "the positioning source note" (positioning note).	Attribution pointer had no link.	positioning note (still present live)
Audience brief (this organization). Reader and role ...	Appended "Agent readers: writing for an agent reader" (agent reader note).	Agent role listed with no route to the agent-reader rules.	agent reader note
Claims register (this organization). Claim as published; ...	Appended "Legal floors: UWG and PAngV (persuasion note), AI Act (AI note), price display (activation note)."	Claims register named laws with no route to the legal-floor blocks.	persuasion note, AI note, activation note
Status: canonical method for this organization, first ...	Completeness pass: Status line: replaced 'first pass on 29 September 2026' with a docref to the Research log (Research log) that holds its passes; Principles and AGENTS docrefs carried through.	Canon states the rule that holds now; history lives in the Research log. Matches the Principles and Myths Status lines.	Live Principles; AGENTS.md
Job story (Adams and Klement). When situation, ...	Completeness pass: Job story template credit '(Adams and Klement)' changed to '(Paul Adams's team at Intercom; named by Alan Klement)'.	Aligns with owner positioning note and the primary source.	https://www.intercom.com/blog/accidentally-invented-job-stories/, read 29 October 2026

Copy review playbook

13 changes.

Block	What changed	Why	Source
Consequence pass. Legal floor (fake urgency, unverified ...	Consequence pass: 'cancellation, AI disclosure where it applies) (legal context)' became 'cancellation, withdrawal button) (legal context); AI disclosure where it applies (AI legal floor AI note, disclosure checklist AI note)'.	AI disclosure was cited to a legal block with no AI Act content, and the list omitted the withdrawal button that Principles rule 21 and the cited legal context name.	persuasion note item persuasion note (live v5); AI note, AI note; https://www.gesetze-im-internet.de/bgb/__356a.html (read 29 October 2026)
table or removed block	Agent-facing bullet adds 'tool descriptions and agent errors pass the tool-description checklist'; prices bullet adds 'German prices pass the price display checklist'; new Purpose bullet 'Use-case pages pass the use-case page checklist'; AI disclosure checklist routed from the consequence pass.	The playbook linked none of the four surface checklists the organization already holds.	agent reader note; activation note and activation note; AI note
table or removed block	Added 'Sources: microcopy patterns, plain language, German specifics and the working rules of the content design note' under the checklist heading.	The checklist's rules and numbers carried no evidence next to them.	content design note, content design note, content design note, content design note
Layout survives text 40 percent longer than ...	'40 percent expansion (short labels 200 percent)' became 'text 40 percent longer than the English, labels of up to 10 characters two to three times as long or more' with a length expansion docref.	200 percent was ambiguous, sat at the floor of the sources and had no evidence.	content design note, content design note (W3C citing IBM), content design note (Microsoft)
Fix the vagueness under the words, not ...	'Detectors and people are near chance at spotting AI text' became 'Most readers are near chance at spotting AI text, and detectors make real errors (Signs of AI writing)'.	Detectors perform better than chance; the detector half contradicted the organization's own source.	content design note; https://en.wikipedia.org/wiki/Wikipedia:Signs_of_AI_writing (read 29 October 2026 by the finder)
Name the reader, their awareness stage, the ...	Docref 'templates' to the Strategy H1 became 'template' to the audience brief Strategy to page copy.	The link landed at the top of the method document.	Strategy to page copy
Status: canonical execution document for this organization, ...	Status docref 'working rules' to Canon H1 became 'copywriting working rules' to Canon Part 5 canon note.	The link landed on the note's H1 and the anchor read like the content design working rules.	canon note
Reader pass. Who arrives, from where, knowing ...	'in about ten seconds of reading (Nielsen 2011)' became 'before most visitors would leave (Nielsen 2011)'.	The ten-second figure is an inference from the dwell-time curve, not a measurement.	WebReading Nielsen 2011 row under web reading note; spill finding a related finding
Product, languages, German address form (du or ...	Reading level target gains a 'readability targets' docref to content design note next to the grade numbers.	Numeric targets had no evidence next to them.	content design note, content design note
No jokes in errors, billing or data ...	'no "simply" or "easy"' became 'no "simply", "just" or "easy"'.	The checklist passed 'just', which the template's banned words in the same document ban.	Playbook; https://developers.google.com/style/word-list (read 29 October 2026 by the finder)
Default position on the four tone dimensions ...	Docref 'NN/g tone study' retargeted from the Four Dimensions heading content design note to the companion study bullet content design note; anchor kept.	The trust and casual-versus-formal findings come from the companion study, a separate block.	content design note; https://www.nngroup.com/articles/tone-voice-users/
Reader pass. Who arrives, from where, knowing ...	Checker correction: replaced it with 'in the first seconds, when visitors are most likely to leave (Nielsen 2011)'. The Nielsen 2011 docref is unchanged. Checked against https://www.nngroup.com/articles/how-long-do-users-stay-on-web-pages/ (read 29 October 2026) and the table row in the web reading note.	'before most visitors would leave (Nielsen 2011)' says more than the source. Nielsen 2011 says users 'often leave Web pages in 10 to 20 seconds' and that 'the probability of leaving is very high during these first few seconds'. It never says most visitors leave.	Checker reread of the live block
Status: canonical execution document for this organization, ...	Completeness pass: Status line: replaced 'first pass on 29 September 2026' with '; the Research log holds its passes.' All four existing docrefs carried through.	Canon states the rule that holds now; pass history belongs in the Research log, matching the Principles and Myths Status lines.	Task 1 (mechanics critic, skeptic confirmed); live block read 29 October 2026

Myths

14 changes.

Block	What changed	Why	Source
Status: canonical list for this organization; the ...	Status: 'first pass on 29 September 2026' and 'Keep this route so later agents do not restore attractive shortcuts.' replaced by '; the Research log holds its passes', matching the live Principles status.	Pass history and a justification clause inside canon.	Principles (live, 29 October 2026); AGENTS.md
Status: canonical list for this organization; the ...	Status adds: 'Claims about mentioning AI and its legal floor are in the do-not-rely-on list of the note on mentioning AI' (docref AI note).	Myths told readers to check every memorable claim here but neither held nor linked the AI note's list.	AI note
Status: canonical list for this organization; the ...	Legend: failed now covers a direct measurement; misattributed glossed; unlawful added (prohibited toward consumers under the cited law, whatever its effect); 'A phrase after a verdict limits its scope.' E-E-A-T row verdict misattributed to misapplied.	Legend did not match the verdicts used; E-E-A-T is a real concept misused as a ranking claim.	web reading note; editorial skeptic revisedFix
table or removed block	New Psychology row: 'People trust algorithms more than people, so lead with AI (algorithm appreciation)', overgeneralized, Qin et al. 2025 and the do-not-rely-on bullet. The 'never name AI' row was not added.	A named effect the AI note rejects had no Myths entry. The skeptics split on the second row; the editorial skeptic showed it was a strawman whose proposed wording misstated Schilke and Reimann, so only the agreed row went in.	AI note (Qin et al. 2025 row), AI note
table or removed block	'Many Labs failures; Kahneman's own retraction of the priming chapter' to 'Flag and currency priming did not replicate in Many Labs (Klein et al. 2014); Kahneman wrote that he "placed too much faith in underpowered studies" (Kahneman 2017, docref persuasion note)'.	Kahneman admitted reliance on weak studies; he did not retract the chapter.	persuasion note, persuasion note; https://retractionwatch.com/2017/02/20/placed-much-faith-underpowered-studies-nobel-prize-winner-admits-mistakes (per finding, read 29 October 2026)
table or removed block	'in 91 attempts with realistic stimuli only 11 reliable effects (Frederick et al. 2014)' to '91 attempts produced only 11 reliable effects, and with verbal or pictorial descriptions the effect fell to chance (Yang and Lynn 2014; Frederick et al. 2014)'.	Wrong author and misdescribed stimuli.	persuasion note (live, Yang & Lynn 2014 now in section 1.9); https://journals.sagepub.com/doi/10.1509/jmr.14.0020
table or removed block	Claim 'think harder and remember more' to 'think harder and solve problems better'.	Meyer et al. 2015 measured problem solving, not memory.	https://pubmed.ncbi.nlm.nih.gov/25844628/ (per finding, read 29 October 2026)
table or removed block	Claim 'Making users feel scarce makes them act' to 'Making people feel short of money or time changes how they decide (scarcity mindset priming)'; evidence '4 of 20 replications significant' to '4 of the 18 originally significant studies replicated, and the original authors dispute the audit', plus 'These studies concern lacking resources, not product scarcity cues'.	Myths carried the same wrong gloss and count the findings name; wording follows the owner's live fix.	persuasion note (live, 29 October 2026)
table or removed block	Verdict 'failed and unlawful' to 'unlawful'; evidence now: UWG annex numbers 7, 23b, 23c; section 5 UWG on availability, price advantages and price with the presumption for a briefly charged price; section 11 PAngV for price reductions on goods; annex number 5 dropped; docref anchor 'legal context, not legal advice'; three official URLs read 29 October 2026. No effect claim added.	'failed' broke the legend; section 11 PAngV covers goods only; fake stock is section 5 UWG, not annex number 5.	https://www.gesetze-im-internet.de/pangv_2022/__11.html, https://www.gesetze-im-internet.de/uwg_2004/__5.html, https://www.gesetze-im-internet.de/uwg_2004/anhang.html (read 29 October 2026)
table or removed block	'Ogilvy limited the rule to expensive purchases' to 'Ogilvy stressed it most for expensive purchases'; 'disagreements' docref retargeted from canon note to canon note.	Ogilvy did not limit the rule; the docref landed on the Canon H1.	canon note (live: 'particularly when you are asking the reader to spend a lot of money'), canon note
table or removed block	Quote restored to 'For every curiosity headline that succeeds in getting results, a dozen will fail'; docref retargeted from canon note to canon note.	Words dropped inside quotation marks without an ellipsis.	canon note; https://archive.org/details/tested-advertising-methods (per finding)
table or removed block	'In 6,700 experiments' to 'In roughly 2,600 categorized tests out of 6,700'.	The averages come from the categorized subset; figure matches the owner's live fix (2,600, not the finding's 3,000).	web reading note and Qubit row in web reading note section (live, 29 October 2026)
table or removed block	Checker correction: replaceBlock on cell (a table cell): 'Prohibited toward consumers' became 'Prohibited or regulated toward consumers', and the number 7 gloss became 'false claim that an offer is available only for a very limited time'. The docref persuasion note and the three URLs are kept.	UWG annex number 7 was glossed as 'false claim of very limited availability'. The annex text (read 29 October 2026) covers a false claim that goods or services are available only for a very limited time. That is a fake countdown, not fake stock, which falls under section 5 UWG. The cell also opened with 'Prohibited', but section 11 PAngV sets a duty (state the lowest price of the last 30 days with a price reduction on goods) and prohibits nothing.	Checker reread of the live block
table or removed block	Completeness pass: Replaced the evidence cell 'Counts participants' comments, not causes; figure not checked against the paper (myths)' with '46.1 percent of comments, not of people, in a study of 2,684 participants; it does not show cause (myths docref; Consumer Reports URL, read 29 October 2026).' Verdict 'overgeneralized' kept.	The figure is now checked: the study summary gives 46.1 percent of comments from 2,684 participants, so the stale caveat goes and the new numbers carry their URL next to them.	https://advocacy.consumerreports.org/research/how-do-people-evaluate-a-web-sites-credibility/, read 29 October 2026 (curl confirmed both '2,684 people' and '46.1 percent of the comments')

Conflicts and open questions

15 changes.

Block	What changed	Why	Source
Mandatory AI disclosure against not leading with ...	Added a Resolved by context entry 'Mandatory AI disclosure against not leading with AI' after the category-creation entry, citing MentionAI section 6 and Principles 28, 29 and 32.	Conflicts calls itself the canonical list of source tensions but omitted the AI disclosure tension that MentionAI section 6 resolves.	MentionAI 01a0ed8c AI note (section 6, residual tension line); Principles, Principles, Principles; live reads 29 October 2026
Plan and limit copy against a real ...	Added a Resolved by context entry 'Plan and limit copy against a real capacity model' routing to Activation A4 (binding limit, 'about' on approximate units, honest variant without a retry time; broad grant is a product decision).	The nine capacity tensions in Activation A4 were unrouted; the skeptics rejected the proposed Open entry wording ('each need a product decision') as inaccurate, so the source skeptic's Resolved by context wording was used.	Activation 01a0ed8f activation note and its items activation note, activation note, activation note, activation note, live read 29 October 2026
Status: canonical list of genuine source tensions, ...	Status line: removed ', first pass on 29 September 2026' (history belongs in the Research log).	Pass history inside canon; the line was also stale since the list has been extended.	Spill finding a related finding (owner Myths), Conflicts part; How to Write voice rule
Hype against understatement. Direct-response agitation (Kennedy, Halbert) ...	Added a Morkes and Nielsen 1997 docref (canon note) beside the 27 percent figure and retargeted 'disagreements' from the Canon H1 canon note to Part 4 canon note.	A number lacked its evidence docref and the disagreements link landed on the H1.	Canon 01a0ed82 canon note ('27% for objective') and canon note (Part 4), live read 29 October 2026
Long copy against short copy. Hopkins, Ogilvy, ...	Retargeted the 'disagreements' docref from canon note to canon note.	Docref landed on the Canon H1, not the section holding the disagreements.	Canon 01a0ed82 canon note, live read 29 October 2026
Clarity against curiosity. Caples ranks curiosity third; ...	Retargeted the 'disagreements' docref from canon note to canon note.	Docref landed on the Canon H1, not the section holding the disagreements.	Canon 01a0ed82 canon note, live read 29 October 2026
Facts against story and identity. Unknown products ...	'(Schwartz's fifth stage)' to '(Schwartz's fifth stage of market sophistication, identification)'.	Bare 'fifth stage' reads as the fifth awareness stage (Unaware), the opposite of a saturated market; wording matches the Canon owner's fix.	Canon 01a0ed82 canon note and Part 4; skeptic evidence quoting Schwartz on identification
Social proof effect size. Descriptive norms work ...	'a German hotel replication did not beat the standard appeal' to 'in two German hotels, norm messages beat no message but not the standard appeal (Bohner and Schlüter 2014 docref persuasion note)'; Cialdini docref kept.	Study unnamed and the sentence omitted that norm messages beat no message.	persuasion note; https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0104086 abstract (read 29 October 2026 by both skeptics)
Simulated persona panels against real readers. Some ...	Docref retargeted from Research log H1 Research log to Research log, anchor 'Research log, queued persona-panel pass'.	Link landed on the Research log H1, not the queued brief it names.	Research log (h3 'Queued: validity of simulated persona panels'), live read 29 October 2026
These gaps are owned by queued research ...	'Research log' docref retargeted from Research log to the 'Queued passes' H2 Research log.	Link landed on the Research log H1.	Research log, live read 29 October 2026
These gaps are owned by queued research ...	Appended 'Already covered: the distinction between Leichte and Einfache Sprache (content design note) and generic trust and privacy copy (AI note).' Briefs in the Research log not touched (not my document).	Partial existing coverage was unlinked; narrow 'already covered' wording used because two skeptics said the core (dignity tension, B2B claims law) is still uncovered.	ContentDesign 01a0ed84 content design note; MentionAI 01a0ed8c AI note; live reads 29 October 2026
These gaps are owned by queued research ...	Added 'emails and sales material, including the consent rule for advertising email (§ 7 UWG, URL, read 29 October 2026)' to the uncovered-evidence list.	Emails and sales material are in the organization's scope but no gap entry recorded them.	https://www.gesetze-im-internet.de/uwg_2004/__7.html read 29 October 2026 (Abs. 2 Nr. 2: advertising by electronic mail without prior express consent; Abs. 3: existing-customer exception)
Conflicts between a product's own canonical documents ...	'see "Where product application lives" in the organization's AGENTS.md' to a docref to AGENTS.md 'How a product organization uses this'.	The named heading does not exist in AGENTS.md.	AGENTS.md 01a0ed9d AGENTS.md (h2), live read 29 October 2026
These gaps are owned by queued research ...	Checker correction: 'Already covered:' became 'Already covered in part: the basic'. Both new docrefs and the Research log docref were kept.	The appended 'Already covered: the distinction between Leichte and Einfache Sprache' read as a contradiction, because the same sentence still lists 'Leichte against Einfache Sprache and the dignity tension' as an uncovered gap. content design note covers only the basic distinction and a SaaS default, not the dignity tension.	Checker reread of the live block
Mandatory AI disclosure against not leading with ...	Completeness pass: Dropped 'small' twice in the AI disclosure conflict entry; named deep fake images, audio or video among the places the law requires disclosure; stated the discloser's trust cost under every framing tested, with a new docref to AI note on 'pooled d = 0.81'; kept all four existing docrefs.	The entry drifted from AI note (pooled d = 0.81, no framing removed it) and AI note (deep fakes, Art. 50(4) first subparagraph, are the third place).	skeptic revisedFix; AI note and AI note read live 29 October 2026

Research log

8 changes.

Block	What changed	Why	Source
Agents and models: one coordinating agent. Five ...	'voice-of-customer desk research for the first audience' to 'for one product's first audience'; 'lanes' kept per both skeptics.	'the first audience' is undefined for a public reader and points at one product.	Findings file a related finding revisedFix; AGENTS.md
Briefs live with each product organization: Mom-Test ...	'Mom-Test interviews with the first audience' to 'with each product's first audience'.	Same undefined reference in the queued brief.	Findings file a related finding revisedFix
Sources covered: see the eight source notes, ...	'see the five source notes' to 'see the eight source notes, five from the first wave and three from the second'; rest of the sentence kept.	Pass 1 produced eight source notes, not five.	Research log; AGENTS.md routes eight source notes
Run before a product's next trust block, ...	'BetrVG section 87; verify' to 'section 87(1) Nr. 6 BetrVG' with URL and read date (systems that monitor employees' behaviour or performance); added 'Already covered' pointer to AI note (Strategy pointer dropped per source skeptic).	Open 'verify' and imprecise citation; brief listed covered topics as open.	https://www.gesetze-im-internet.de/betrvg/__87.html, read 29 October 2026; MentionAI 01a0ed8c AI note
Run when a product with vulnerable or ...	Added 'Already covered: Leichte against Einfache Sprache and DIN SPEC 33429:2025-03 (docref content design content design note); ISO 24495-1 is not.' after the Goal.	Brief presented already covered topics as open gaps.	ContentDesign 01a0ed84 content design note (exists, read 29 October 2026)
Run before a content or recipe site's ...	Added 'Already covered' pointers to AI note (recipe AI disclosure) and agent reader note (structured data) after the Goal.	Brief presented already covered topics as open gaps.	MentionAI 01a0ed8c AI note; AgentReader 01a0ed8b agent reader note (both exist, read 29 October 2026)
Queued: emails and sales material	Inserted new queued pass 'Queued: emails and sales material' (Research log, Research log) after Research log, with pointers to content design note and Strategy to page copy; legal scope stated as German floor only: prior express consent § 7 Abs. 2 Nr. 2 UWG and existing-customer exception § 7 Abs. 3 UWG.	Emails and sales material are in scope but uncovered and unrecorded as a gap.	https://www.gesetze-im-internet.de/uwg_2004/__7.html, read 29 October 2026
Run before a product's next lifecycle, onboarding ...	Checker correction: Anchor text changed to 'from product strategy to page copy'; same documentId and blockKey;	Docref anchor text 'method' for the Strategy document did not tell a reader what the link points to.	Checker reread of the live block

Loop: Recheck the legal blocks

3 changes.

Block	What changed	Why	Source
table or removed block	Owner cell 'the owner, for Oddlynew, which maintains this organization ...' to 'Oddlynew, which maintains this organization. It decides anything that widens the blocks' scope.'	A public Loop named an internal person, against its own Authority row ('no internal names').	Loop 01a17e95 table (a table cell), Authority row (a table cell); AGENTS.md 01a0ed9d AGENTS.md 'Maintained by Oddlynew'
table or removed block	Trigger cell 'No harness schedule or installation exists.' to 'No scheduler starts a session for it.'	Internal agent-tooling wording a public reader cannot parse; the skeptics' wording keeps the row consistent with 'Runs when due in a session'.	Loop 01a17e95 Trigger row (revisedFix, source skeptic)
Add one Research log entry with every ...	Step 6 'Add one Research log entry: sources read, what changed, what held.' to 'Add one Research log entry with every field the log's status line names.' (numbered list item kept)	The step listed three fields while the Research log's status line requires eight, so each cycle would write an incomplete entry.	Research log 01a0ed93 status line Research log

Source: Copywriting canon

30 changes.

Block	What changed	Why	Source
Use plain words. Aim for grade 7 ...	Rule 17: 'Aim for a 5th to 7th grade reading level' to 'Aim for grade 7 to 9 for general readers and no higher than about grade 12 for specialists' with a docref to Principles 14; Unbounce marked observational.	The Canon's working rule set a reading target that contradicts Principles 14 and rests on one correlational benchmark.	Principles Principles; canon note caveat
table or removed block	Added under the Part 5 heading: 'These rules sum up this note's sources. The canonical rules are in Principles, which holds where the two differ.' with a docref.	Part 5 is a second rule list beside Principles with no statement of which holds.	AGENTS 01a0ed9d (Principles is the canonical distilled rules); Principles
table or removed block	Inserted the missing fourth sophistication stage ('When rivals copy the mechanism: enlarge or elaborate it, so it works easier, quicker or more surely.') and marked identification '(the fifth stage)'.	The sophistication list skipped Schwartz's fourth stage, so 'fifth stage' pointed at nothing.	https://nordiccopy.com/market-sophistication/ (read 29 October 2026): stage 4 'The new mechanism is enlarged upon'; finding evidence (Breakthrough Advertising pp. 37 to 50)
table or removed block	'his fifth stage' to 'his fifth stage of market sophistication, identification'; rule 23 'Schwartz's fifth stage of market sophistication'; Apple lesson 'Schwartz's identification stage' to 'Schwartz's fifth stage of market sophistication'.	The phrase was ambiguous between the awareness and sophistication scales. (The Conflicts part belongs to that document's fixer.)	canon note list as now corrected
table or removed block	Replaced the Wikipedia paraphrase in quotation marks with Ogilvy's own line: 'Unless your advertising is built on a BIG IDEA it will pass like a ship in the night' (How to Create Advertising That Sells); 'Show the benefit "in a specific and factual manner."' to 'Show the benefit specifically and factually.'	Two Wikipedia summary sentences were quoted as Ogilvy's words.	https://www.ogilvy.com/sites/g/files/dhpsjz106/files/pdfdocuments/Quotations_of_David_Ogilvy_v2.pdf (read 29 October 2026); URL added to canon note
Resolution: length follows the reader's awareness and ...	'Ogilvy himself limited the long-copy rule to cases where...' to 'Ogilvy said long copy sells more "particularly when you are asking the reader to spend a lot of money" (Ogilvy on Advertising, 1983, confirmed through reviews; see Verification notes).' canon note left unchanged, as both skeptics advised.	Ogilvy strengthened the rule for expensive purchases; he did not limit it. (The Myths FTroOxuCoxl part belongs to that document's fixer.)	Ogilvy on Advertising p. 217 via https://ricardovice.substack.com/p/ogilvy-on-advertising (read 29 October 2026, per finding)
table or removed block	'Long copy sells more than short copy when the reader needs the information (from Ogilvy's summary in the foreword)' to 'Long copy sells more than short copy (Ogilvy's foreword to the fourth edition). Caples adds that long copy is for crowding in sales arguments, not for filling space.'; author line 'has a foreword by David Ogilvy' to 'reprints David Ogilvy's foreword to the fourth edition'.	The added condition is not in Ogilvy's foreword, and the foreword is the fourth edition's.	https://archive.org/details/tested-advertising-methods full text (read 29 October 2026): 'FOREWORD TO THE FOURTH EDITION ... 7. Long copy sells more than short copy'; chapter 11 'crowd in as many sales arguments as possible'
table or removed block	'The customer is not a moron, she's your wife.' to 'The consumer isn't a moron; she is your wife.' (with source) in Part 1 and rule 18.	Misquote.	Ogilvy quotation book PDF (read 29 October 2026), attributed to Confessions of an Advertising Man
table or removed block	Unquoted the Kennedy line: 'Copyhackers writes that Kennedy called PAS the most reliable copywriting formula for sales ever invented' with URL and read date; table cell reworded the same way.	The quoted words were not on the cited page and differed between the two places.	https://copyhackers.com/2015/10/copywriting-formula/ (read 29 October 2026)
table or removed block	Added the Confessions chapter 6 excerpt URL for the headline quote and keyword examples; verification note now reads 'Ogilvy's long-copy line for expensive purchases, via reviews of Ogilvy on Advertising.'	The keyword examples and five-times quote come from Confessions (1963), not Ogilvy on Advertising.	https://www.campaignlive.co.uk/article/excerpt-confessions-advertising-man/996114 (read 29 October 2026 by the skeptics; returns 403 to direct fetch, content confirmed via search snippets the same day)
Pattern catalogues: Schwab's "100 good headlines" and ...	'Caples's chapter of 29 headline formulas' to '35 headline formulas (fifth edition, 1997; the fourth edition has 29)'.	The linked 1997 edition has 35 formulas.	archive.org full text (read 29 October 2026): 'THIRTY-FIVE PROVEN FORMULAS FOR WRITING HEADLINES'
What it gives: A compact course built ...	'It includes "100 good headlines and why they work" and 14 split-run tests.' to 'It includes a list of 100 good headlines with the reasons they worked, and a chapter of 14 split-run tests.'	The quoted chapter title was not the book's. The library table of contents (bot-protected) and archive.org text (restricted) could not be reopened today, so the quotation marks were removed rather than replaced with an unchecked title.	Finding evidence: British Council library record (read 29 October 2026 by skeptics)
table or removed block	Closed the Collier flag: 'The popular one-line version is a paraphrase. In the 1950 edition Collier writes that every reader holds "a mental conversation with himself, the burden of which is his own interests" and that a letter should "chime in on that conversation"' with the full-text URL; dropped the Collier item from the not-checked list.	The flag was open and the one-liner is not Collier's wording.	https://idoc.pub/documents/the-robert-collier-letter-book-6nq8egkr7pnw (read 29 October 2026)
Sources: RC Victorino, "How Stripe Built a ...	'(checked Sept 2026)' to '(read 29 October 2026)'; order corrected: logo strip directly under the hero, then the numbers, then the enterprise section's Fortune 100 figure. Linear canon note left as is.	The described order was wrong and the date was a month.	https://stripe.com (read 29 October 2026): logo strip after hero, stats block 'The backbone of global commerce', Fortune 100 in the enterprise section
Author / year / format: April Dunford, ...	'2019 (an updated edition has followed)' to '2019; updated and expanded second edition 2026 (five components, five steps)'.	Undated edition.	https://aprildunford.substack.com/p/announcement-obviously-awesome-the (read 29 October 2026): 'five components and five steps'; https://www.aprildunford.com/books
What it gives: The model for opinionated, ...	'Fried still writes "all the key pages"...' to 'In a December 2023 interview, Fried said he still writes "all the key pages" (landing, pricing, features) himself.'	Present-tense claim on an undated interview.	https://churnkey.co/blog/jason-fried/ (read 29 October 2026), dated December 14, 2023
table or removed block	Dropped 'now' and 'currently': 'founder of CXL; founder and CEO of Wynter', 'Framework article (November 2022)', 'book (reprinted)', and the Bourgoin line now carries a read date. canon note handled under a related finding.	Time words without a date.	https://wynter.com/post/b2b-message-layers-framework-wynter (read 29 October 2026): 'November 22, 2022, Peep Laja Founder & CEO @Wynter'
"Writing simply keeps you honest. If you ...	Completed the Graham quote: '...it will be obvious to everyone, including you.'	The quote was cut inside quotation marks.	https://paulgraham.com/simply.html (read 29 October 2026)
NN/g: Concise, Scannable, and Objective (Morkes and ...	'124% more usable than promotional copy' to '124% more usable than a promotional version of the same site, in one 1997 test with 51 users' with a docref to the web reading note.	The must-read summary stated a single-site result as a general law, which Myths lists as overgeneralized.	WebReading web reading note; Myths row on the 124 percent claim
table or removed block	'Agitation reads as marketese on B2B pages (NN/g)' to 'Agitation slides easily into promotional language, which NN/g's 1997 test found less usable than objective copy'.	NN/g tested neither agitation nor B2B pages.	canon note; WebReading Morkes and Nielsen row
table or removed block	Added a pointer line to the web reading note under the heading; the 79% line now names its article ('How Users Read on the Web'). Unbounce wording left unchanged.	The section duplicated WebReading's evidence without its sample sizes and misplaced the 79% figure's source.	WebReading web reading note; https://www.nngroup.com/articles/how-users-read-on-the-web/
table or removed block	Added under the Part 4 heading: 'These are the source positions. The organization's rulings on these tensions are in Conflicts; where the two differ, Conflicts holds.' with a docref.	Part 4 resolves the same tensions as Conflicts with no statement of which holds.	Conflicts Conflicts
table or removed block	Removed unevidenced superlatives: 'The only framework' to 'A framework for choosing', 'The best single model for a SaaS homepage' to 'A model for the order of a B2B homepage', 'the best self-edit checklist' to 'a short self-edit checklist', 'the writer founders trust most' to 'a founder-essayist', 'The clearest account' to 'An account'.	The note's own rule is to replace superlatives with facts.	canon note; AGENTS 01a0ed9d
table or removed block	Renamed the field label 'Why it matters:' to 'What it gives:' in all 20 entries, text after it unchanged.	The voice rules exclude 'why it matters' framing.	How to Write in Oddlynew voice rule
table or removed block	Replaced the invented Wynter quote 'with proof rather than adjectives' with Wynter's words 'Do not invent difference by adding adjectives' in all three places.	The quoted phrase is not on either Wynter page.	https://wynter.com/post/b2b-message-layers-framework-wynter (read 29 October 2026)
URL: https://marketingexamples.com/copywriting(https://marketingexamples.com/copywriting) ; https://marketingexamples.com/copywriting/tips(https://marketingexamples.com/copywriting/tips) ; https://marketingexamples.com/copywriting/conversion(https://marketingexamples.com/copywriting/conversion) ; ...	Replaced the dead Substack URL with https://www.youtube.com/watch?v=TUMjnmfsPeM (read 29 October 2026).	Dead link (404).	Substack URL returns 404 and YouTube oEmbed confirms 'Learn Copywriting in 76 Minutes, Harry Dry' by David Perell (both checked 29 October 2026)
table or removed block	'Honesty feels warmer' to 'An honest line always feels warmer' in both places.	Misquote.	https://marketingexamples.com/copywriting/tips (read 29 October 2026)
table or removed block	Restored Handley's wording: Good writers "are excellent editors of their own work."; "...too simple to understand."; "...don't assume the reader is stupid."	Spliced and shortened quotations.	https://annhandley.com/13-writing-rules/ (read 29 October 2026), rules 6, 7 and 10
Apply the header test: "If the visitor ...	Shapiro quote restored: 'If the visitor reads only this text on your page, will they know exactly what you sell?'	Words dropped inside the quote.	https://www.julian.com/guide/growth/landing-pages (read 29 October 2026)
Write for trigger moments. Every purchase starts ...	'from not having the problem' to 'from not knowing they have the problem'.	Bourgoin's trigger is the move from unaware to in the market.	https://learnwhywebuy.com/the-trigger-technique-turn-buyer-stories-into-smarter-marketing-campaigns/ (read 29 October 2026)

Source: Psychology of persuasion

32 changes.

Block	What changed	Why	Source
Meta-rule: test anything rated mixed or weak ...	Meta-rule 'A/B test anything rated mixed or weak' became 'test anything rated mixed or weak before relying on it: an A/B test where traffic allows, otherwise the methods in the low-traffic testing playbook' (docref Strategy to page copy); OSC 2015 sentence kept.	The meta-rule contradicted the organization's low-traffic method, which says not to A/B test conversion below the needed sample.	Strategy to page copy item 1; Principles
5. Do not rely on	Section 5's 14 items replaced by one line pointing to Myths: Psychology (persuasion note kept as that line, 13 items removed); Mandel & Johnson 2002 moved into the scorecard's priming row; 1.3 principle 5 now cites Myths by docref. Item-level fixes of related findings and the section 5 item 14 part of a related finding have no block left here; their corrected content lives in Myths (owner) and in this note's 1.15 and section 3 blocks.	Section 5 duplicated the canonical Myths list and had started to drift.	Myths and Myths; getBacklinks on 29 October 2026 showed no inbound citation to section 5 items
Make it easy to read: high contrast, ...	'users read about 20 to 28% of a page's words (NN/g, URL)' became 'readers scan more than they read: one model puts what users have time to read at about 20% of a page's words, at most 28% (NN/g 2008, URL; a model estimate, not a constant)'.	A model estimate (28% is a ceiling, 20% likely) was stated as fact inside a rule graded Robust.	https://www.nngroup.com/articles/how-little-do-users-read/; Myths row on the 20 percent claim
Expressions of concern: Mazar, Amir & Ariely ...	Unsourced 'a 2008 JMR paper (2024) and a 2004 Psychological Science paper (2021)' now names Mazar, Amir & Ariely 2008 (JMR, September 2024) and Heyman & Ariely 2004 (Psychological Science, July 2021) with both notice URLs, read 29 October 2026.	A public claim about named papers carried no source.	Crossref records for 10.1177/00222437241285882 (20 Sep 2024) and 10.1177/09567976211035782 (23 Jul 2021), read 29 October 2026
Status: source note, first research pass on ...	Status kept (per two skeptics' revisedFix) and made true by sourcing the two unsourced claims in place (Mertens, Ariely expressions of concern); added '; section 3 carries its own "Last checked" date'.	'Exceptions are flagged in place' did not hold while two claims had no source, and the Status did not point to section 3's later check date.	Blocks persuasion note and persuasion note of this note; MentionAI spill fix text
These figures come from secondary sources, not ...	Lead-in 'Two numbers ... primary papers:' became 'These figures come from secondary sources, not the primary works:'; the O'Donnell bullet (persuasion note) now reads 'The Economist decoy figures: taken from Wikipedia's account of Ariely's Predictably Irrational (2008); the book itself was not checked.'	The list undercounted (decoy figures come from Wikipedia) and the O'Donnell caveat was stale once the primary paper was cited.	This note persuasion note; PNAS correction https://www.pnas.org/doi/10.1073/pnas.2217321119 (read via Europe PMC PMC9860137, 29 October 2026)
Weak / failed: failed preregistered replication, the ...	New legend bullet after the three classes: 'A range such as "weak to mixed" sits between two classes; for testing, treat it as the lower one. Mandatory marks a legal requirement, not an evidence class.'	The note uses range and legal labels the legend did not define, so the meta-rule could not be applied to them.	This note's labels in persuasion note, persuasion note, persuasion note, persuasion note
Truth and fluency effects: robust (Dechêne et ...	Split into 'Truth and fluency effects: robust (Dechêne et al. 2010 meta-analysis).' and a new bullet 'The long-words finding: mixed (Oppenheimer 2006, small student samples; not replicated at scale as far as I found).'; rule 1 (persuasion note) grade 'Robust.' became 'Robust (fluency) / mixed (long words).'	One small 2006 paper failed the note's own definition of Robust.	Legend persuasion note; Oppenheimer 2006 https://onlinelibrary.wiley.com/doi/10.1002/acp.1178 (finding evidence)
Nudges in general: mixed. Mertens et al. ...	'Mertens et al. 2022 reported d = 0.43.' now cites PNAS 119(1) https://www.pnas.org/doi/10.1073/pnas.2107346118 and its May 2022 correction https://www.pnas.org/doi/10.1073/pnas.2204059119, read 29 October 2026.	The effect size had no source; the original article says 0.45 and only the correction gives 0.43.	Europe PMC PMC8740589 (d = 0.45) and PMC9171817 correction (d = 0.43), read 29 October 2026
A Bayesian synthesis still supports the norm ...	'Critics found a smaller, still largely supportive effect.' became 'A multilevel Bayesian reanalysis of the same seven studies found the evidence moderate to inconclusive once the studies may differ in effect: Carlsson, Schimmack, Williams & Bürkner 2017, Psychological Science, https://doi.org/10.1177/0956797616684682, read 29 October 2026.'	The critique was unnamed and its summary was wrong.	https://doi.org/10.1177/0956797616684682 (author PDF, read 29 October 2026): 'even a small amount of between-samples variability renders the evidence inconclusive'
Endowment can help: a trial with the ...	'Endowment works: a trial ... creates something to lose.' became 'Endowment can help: a trial with the user's own data can create something to lose (no source yet; test it).'	A flat unsourced claim sat under a Mixed verdict.	Section 1.14 sources and Evidence line persuasion note of this note
Loss aversion is weaker for small stakes ...	Rule now cites '(Harinck et al. 2007; Mrkva et al. 2020)'; new 1.14 source bullet Harinck, Van Dijk, Van Beest & Mersmann 2007 (Psychological Science 18(12)): in three experiments, loss aversion reversed for small amounts.	The small-stakes clause had no source.	https://doi.org/10.1111/j.1467-9280.2007.02031.x (Crossref abstract, read 29 October 2026)
Remove sludge from paths that serve the ...	'In Germany this is also a legal requirement (section 3).' became 'For consumer contracts in Germany, the cancel and withdrawal paths are also legal requirements (section 3).'	Section 3 covers cancellation and withdrawal only, not sign-up or export friction.	https://www.gesetze-im-internet.de/bgb/__312k.html and https://www.gesetze-im-internet.de/bgb/__356a.html, read 29 October 2026
Keep choice symmetric and avoid dark patterns. ...	'DSA Art. 25' became 'DSA Art. 25 (online platforms)' in rule 20.	The bare citation read as binding any website, while Art. 25 binds only online platforms.	EUR-Lex Regulation (EU) 2022/2065 Art. 25(1), via EUR-Lex search result read 29 October 2026; persuasion note
DSA Art. 25: online platforms "shall not ...	Unofficial eu-digital-services-act.com link replaced by '(Regulation (EU) 2022/2065, Art. 25 and Recital 67: https://eur-lex.europa.eu/eli/reg/2022/2065/oj/eng, read 29 October 2026)'.	A quoted EU article was cited from a private mirror.	https://eur-lex.europa.eu/eli/reg/2022/2065/oj/eng (Art. 25(1) wording confirmed through a EUR-Lex search result, 29 October 2026)
Last checked: 27 October 2026, against the ...	'Everything below held; that day added the new environmental-claims entries ... status.' became 'Everything below held on that date.'	Change history inside a block that states what holds now.	Research log holds the same history
Summary: The 2021 edition adds a seventh ...	Label 'Why it matters:' became 'Summary:' in all nine blocks (persuasion note, persuasion note, persuasion note, persuasion note, persuasion note, persuasion note, persuasion note, persuasion note, persuasion note); dropped 'this is the standard vocabulary of persuasion' (1.1) and 'the most rigorous practitioner book of the group' (1.6).	The voice rules exclude 'why it matters' and unsourced praise.	How to Write in Oddlynew; organization AGENTS.md voice rules
Precise figures read as measured: "cut build ...	Rule head kept; 1.12 gained Jerez-Fernandez, Angulo & Oppenheimer 2014 (precision as a cue to confidence) and Loschelder et al. 2016 (precise offers backfired with experts); persuasion note 'reverses with experts in some work' now cites Loschelder et al. 2016.	The only precision source measured anchoring, not credibility, and the expert reversal had no source.	https://doi.org/10.1177/0956797616666074 (Crossref abstract, read 29 October 2026); https://doi.org/10.1177/0956797613504301 (title 'Show me the numbers: precision as a cue to others' confidence')
Show the product. Real screenshots and diagrams ...	'remembered far better than' became 'remembered better than'.	'Far' overstated 98% against 90% recognition.	This note persuasion note (Shepard 1967)
Yechiam & Zeif 2025, "Loss aversion is ...	'Yechiam & Zeif 2025 re-meta-analysis manuscript ... The debate continues.' now cites Journal of Economic Psychology 107, 102801 and states that on Brown et al.'s data (84 papers, 163 estimates) lambda was about 1.07, not significantly above 1, for symmetric, unordered designs; scorecard Loss aversion row adds Yechiam & Zeif 2025.	The paper is published and its counter-evidence was missing.	https://yeldad.net.technion.ac.il/files/2025/02/YZ_2025.pdf abstract, read 29 October 2026; Crossref 10.1016/j.joep.2025.102801
O'Donnell et al. 2021 (PNAS 118(44), https://www.pnas.org/doi/10.1073/pnas.2103313118(https://www.pnas.org/doi/10.1073/pnas.2103313118); ...	O'Donnell bullet now cites the PNAS paper and its November 2022 correction, says the studies concern lacking money or time (not product scarcity cues), and gives '4 of the 18 originally significant studies replicated'; Collins blog link dropped; scorecard row Status 'Weak (4 of 18 replicated)'.	The count used the wrong denominator and the bullet did not say the audit concerns lacking resources.	PNAS correction 10.1073/pnas.2217321119 (Europe PMC PMC9860137, read 29 October 2026): '18 of the effects we attempted to replicate were significant when they were originally published'
Barton, Zlatevska & Oppewal 2022 (Journal of ...	New 1.15 bullet after Barton: 'Luguri & Strahilevitz 2021 (section 1.16): in their second experiment, "must act now" messages did not make consumers more likely to buy a costly service.'	Section 1.15 gave only intention evidence while the note already cited a behavioural null for urgency.	Crossref abstract of 10.1093/jla/laaa006, read 29 October 2026
Frederick, Lee & Baskin 2014, "The limits ...	Frederick, Lee & Baskin bullet now gives their own result (38 studies; 4 of 5 numeric, 2 of 5 mixed, 0 of 27 experienced attributes); new bullet attributes '91 attempts, 11 reliable effects' to Yang & Lynn 2014; 'Huber et al. replied' became 'Huber, Payne & Puto replied to both papers'.	The headline number belonged to a different paper.	Frederick et al. 2014 General Discussion (NUS author PDF, read 29 October 2026); Yang & Lynn 2014 abstract via Crossref 10.1509/jmr.14.0020, read 29 October 2026
Show the highest tier or annual total ...	Pricing principle now reads: price reduction on goods offered to consumers states the 30-day lowest price (§ 11 PAngV); for a service such as a subscription, a reduction from a price never really charged misleads about the price (§ 5 UWG; section 3).	§ 11 PAngV covers goods only; the rule was stated as universal.	https://www.gesetze-im-internet.de/pangv_2022/__11.html and https://www.gesetze-im-internet.de/uwg_2004/__5.html, read 29 October 2026
Ariely & Wertenbroch 2002 procrastination and deadlines ...	'retracted in September 2026 after Data Colada found signs of tampering' became 'retracted on 2 September 2026 after a replication raised questions about its data and Data Colada questioned the data's veracity' with the notice URL plus the Retraction Watch report, read 29 October 2026.	Only a news report was cited and the replication ground was missing.	Crossref 10.1177/09567976261488042; notice wording as quoted by Retraction Watch, read 29 October 2026
Processing fluency explains part of the effect. ...	'The effect ran through processing fluency.' became 'Processing fluency explains part of the effect.'	Mediation held only in Experiment 1; the author says fluency is at least partially responsible.	Oppenheimer 2006, https://onlinelibrary.wiley.com/doi/10.1002/acp.1178 (finding and skeptic evidence)
§ 11 PAngV (since 28 May 2022): ...	'any advertised price reduction ... 30 days' became 'an advertised price reduction for goods offered to consumers ... lowest total price ... 30 days', plus: § 11 names goods only; for services, SaaS plans included, § 5 UWG still applies (§ 5 Abs. 2 Nr. 2 price advantage; § 5 Abs. 5 presumption for a price charged only briefly); law-firm link replaced by official texts; rule 8 (persuasion note) now cites '§ 5 UWG; § 11 PAngV for goods'.	§ 11 PAngV was stated as covering every price reduction, which misleads a SaaS reader.	https://www.gesetze-im-internet.de/pangv_2022/__11.html; https://www.gesetze-im-internet.de/uwg_2004/__5.html, both read 29 October 2026
Withdrawal button (Directive (EU) 2023/2673, new Art. ...	Withdrawal item now names § 356a BGB (since 19 June 2026) with 'Vertrag widerrufen' and the EU label, 'available throughout the withdrawal period' instead of 'always-available', and the confirmation function 'Widerruf bestätigen' ('confirm withdrawal'); law-firm link replaced by EUR-Lex and gesetze-im-internet.de.	The German law, its labels and the mandatory confirmation step were missing, and 'always-available' overstated the duty.	https://www.gesetze-im-internet.de/bgb/__356a.html, read 29 October 2026; https://eur-lex.europa.eu/eli/dir/2023/2673/oj/eng (Art. 11a wording via EUR-Lex search result, 29 October 2026)
§ 312k BGB cancellation button (since 1 ...	§ 312k item now adds the confirmation page with a 'jetzt kündigen' button and that both buttons and the page stay permanently available and easy to reach; source now gesetze-im-internet.de, read 29 October 2026.	§ 312k requires two steps; the note named only the first button.	https://www.gesetze-im-internet.de/bgb/__312k.html, read 29 October 2026 (Abs. 2 and 6)
Since 27 September 2026 (Drittes Gesetz zur ...	'(Nr. 23d and following)' became 'of goods, including goods with digital elements (Nr. 23d, points a to g)'.	There is one Nr. 23d with points a to g, it covers goods only, and Nr. 24 is an aggressive practice.	https://www.gesetze-im-internet.de/uwg_2004/anhang.html, read 29 October 2026
UWG blacklist (Anhang zu § 3 Abs. ...	dejure.org UWG annex link replaced by https://www.gesetze-im-internet.de/uwg_2004/anhang.html (read 29 October 2026); the § 312k dejure link in persuasion note replaced by gesetze-im-internet.de.	Statutes were cited through a private mirror though the Loop requires official texts.	https://www.gesetze-im-internet.de/uwg_2004/anhang.html and https://www.gesetze-im-internet.de/bgb/__312k.html, read 29 October 2026
Precision effects: mixed (fewer replications; with experts, ...	Checker correction: replaceBlock: 'Precision effects: mixed (fewer replications; with experts, too much precision backfires: Loschelder et al. 2016).' List bullet level 1 kept;	Nested parentheses, and the text overstated the source: 'the effect reverses with experts (Loschelder et al. 2016)'. The paper found an inverted U for experts, where too much precision backfires; it did not find a plain reversal.	Checker reread of the live block

Source: Positioning, customer research and product narrative