A phone directory case set copyright's creativity floor
Feist rejected sweat-of-the-brow protection and required minimal creativity, a standard the Office's registration practice still applies.
3 min · source-led record↗
100 records / One research shelf
Find the question
behind the tool.
Search documentation-led notes, historical cases and studies. These retained drafts were prepared on 16 September 2026; current prices, policies and legal developments need rechecking before you act.
Historical event dates are not site publication dates. This local redesign does not freshly verify every inherited claim.
No matching record. Try fewer words or reset the filters.
Copyright required a human choice behind the camera
Burrow-Giles tied photographic copyright to posing and arrangement, the human-authorship line the 2023 AI guidance calls its leading case.
3 min · source-led record↗
A monkey's selfie could not hold its own copyright
The Ninth Circuit held a macaque lacks Copyright Act standing, reasoning the Office's own AI guidance still leans on against non-human authors.
3 min · source-led record↗
An engineer imagined linked ideas, not machine-written prose
Vannevar Bush's 1945 memex proposal describes associative retrieval decades before word processors, not a machine that composes text.
3 min · source-led record↗
A program's illusion of listening was not comprehension
Weizenbaum's 1966 ELIZA paper describes keyword pattern-matching, not language generation, and warns against reading fluency as understanding.
3 min · source-led record↗
Spelling checks made software correction unremarkable
Microsoft's own developer documentation describes spelling and grammar checks as flagged suggestions a user decides on, not silent rewrites.
3 min · source-led record↗
Word made every edit visible, attributable and reversible
Word's tracked-changes documentation records who changed what, a visible-edit mechanism disclosure proposals now borrow for AI-assisted drafts.
3 min · source-led record↗
A keypad patent taught software to guess the word
Tegic's 1995 patent describes a frequency-ranked dictionary lookup for ambiguous keypresses, the direct ancestor of today's predictive text.
3 min · source-led record↗
Dictation software let authors speak a manuscript
Nuance's Dragon dictation line, now under Microsoft, still documents converting spoken words into editable text rather than drafting them.
3 min · source-led record↗
A startup's browser word processor became Google Docs
Google's own 2006 posts show Writely acquired, then merged with Spreadsheets into the always-online, multi-editor document.
3 min · source-led record↗
The Copyright Office says digital replicas need federal law
Part 1 of the Copyright Office's AI report recommends a federal right against unauthorized digital replicas, leaving other issues to later parts.
3 min · source-led record↗
The Copyright Office ties AI training's fairness to its use
The pre-publication Part 3 report says AI training can be fair use for research but not for commercially competing output.
3 min · source-led record↗
A registration form now forces an AI authorship disclosure
The Copyright Office's March 2023 guidance requires applicants to disclose and exclude AI-generated material from what they claim.
3 min · source-led record↗
Amazon draws its own line between AI-made and AI-assisted
KDP's living content guidelines require disclosing AI-generated text, images and translations but not AI-assisted editing or brainstorming.
3 min · source-led record↗
Apple Books separates an AI credit from an author's
Apple's own formatting guide requires an AI Generated by attribution distinct from a human author credit.
3 min · source-led record↗
Barnes and Noble Press asks its own AI disclosure question
B&N Press requires authors to declare AI-Created versus AI-Assisted work, using terms that do not match Amazon's or Apple's.
3 min · source-led record↗
Kobo says it will not train models on authors' books
Kobo's blog and content policy say it won't feed books into an LLM while barring content generated primarily by automated tools.
3 min · source-led record↗
The EU's AI Act now requires labeling synthetic text
From 2 August 2026 the AI Act requires marking some AI-generated content, on top of copyright duties already binding AI model providers.
3 min · source-led record↗
Three publishing bodies agree AI cannot be an author
ICMJE, COPE and WAME each state AI tools cannot meet authorship criteria because they cannot take responsibility for a manuscript.
3 min · source-led record↗
Two wire services keep photo authenticity off-limits to AI
AP and Reuters both bar generative AI from creating news imagery while allowing it as production infrastructure elsewhere in the newsroom.
3 min · source-led record↗
An authors' lawsuit put AI training itself on trial
The Authors Guild's 2023 complaint against OpenAI targeted how books entered a model, not just what a chatbot outputs.
3 min · source-led record↗
Authors asked a judge to rule OpenAI's training unlawful
A 2026 summary-judgment motion asks the court to decide fair use as a matter of law, not send it to a jury.
3 min · source-led record↗
A judge split AI training from the piracy that fed it
Bartz v. Anthropic found training fair use but ruled downloading pirated books was not, sending only that question to trial.
3 min · source-led record↗
A settlement pays authors only if a copyright was registered
Anthropic's piracy settlement pays class members by title, but only for books the Copyright Office shows were registered on time.
3 min · source-led record↗
Meta won on fair use because authors could not show harm
Kadrey v. Meta found for the platform not because copying was harmless, but because these plaintiffs did not prove market dilution.
3 min · source-led record↗
A machine cannot register a copyright it authored alone
A federal court upheld the Copyright Office's refusal to register a work whose applicant claimed no human involvement at all.
3 min · source-led record↗
The Copyright Office split a comic book from its pictures
Zarya of the Dawn kept copyright in its text and layout while the individual Midjourney images lost protection entirely.
3 min · source-led record↗
A newspaper's suit tested reproduction, not just training
The Times sued OpenAI and Microsoft over chatbots it says both trained on and reproduced its journalism nearly verbatim.
3 min · source-led record↗
One judge now oversees a dozen AI copyright lawsuits
A federal panel consolidated twelve author, publisher and news suits against OpenAI into a single New York proceeding.
3 min · source-led record↗
A registration gap can erase an author's settlement award
An Authors Guild survey found members reporting that publishers often failed to register the copyrights AI settlements now require.
3 min · source-led record↗
Grammarly's own plans page draws the free-to-Pro line
Grammarly's current pricing and support pages state what free, Pro and Enterprise tiers include, as retrieved in September 2026.
3 min · source-led record↗
Sudowrite prices fiction help by the credit, not the seat
Sudowrite's own plans page ties its tiers to monthly AI credits rather than word counts, as retrieved in September 2026.
3 min · source-led record↗
Jasper prices itself for marketing teams, not manuscripts
Jasper's own trial and pricing pages describe a platform built for brand content, not book-length prose, as retrieved in September 2026.
3 min · source-led record↗
Notion built AI writing into the workspace, not a new app
Notion's own help center and product pages describe AI writing tools inside its existing workspace, as retrieved in September 2026.
3 min · source-led record↗
Google put a drafting assistant inside the blank Doc
Google's March 2023 announcement and its current Docs help page describe the same drafting feature under different names.
3 min · source-led record↗
Copilot joined Word's toolbar without replacing Editor
Microsoft's March 2023 announcement and current support page describe what Copilot in Word drafts, rewrites and requires you to verify.
3 min · source-led record↗
ProWritingAid prices credits and promises to delete text
ProWritingAid's own pricing tiers and a separate data-safety page describe cost and manuscript handling, as retrieved in September 2026.
3 min · source-led record↗
LanguageTool separates its checker from its AI training
LanguageTool's premium page and privacy policy state what Premium adds and that inputs are not used to train its models.
3 min · source-led record↗
ChatGPT launched free, then Plus arrived ten weeks later
OpenAI's original November 2022 post and its February 2023 Plus announcement mark what changed and what stayed free.
3 min · source-led record↗
Both chat assistants warn, in writing, not to trust them fully
OpenAI's and Anthropic's own help pages both flag confident-sounding errors as a named risk for writing and research tasks.
3 min · source-led record↗
Elicit prices its tiers around paper counts, not general chat
Elicit's own pricing page limits its claims to searching, summarizing and extracting from a defined set of papers and trials.
3 min · source-led record↗
Scite classifies what a citation actually does to a claim
Scite's own site defines Smart Citations as showing support, challenge or discussion, not just recording a reference.
3 min · source-led record↗
ACX ties an audiobook's royalty rate to exclusivity
ACX's own royalty page sets two production-cost models and two royalty tiers due to converge by the end of 2026.
3 min · source-led record↗
Apple's synthetic narration still starts from the author's own rights
Apple's authors documentation sets eligibility and voice rules for digital narration, present on its site by January 2023.
3 min · source-led record↗
KDP's AI disclosure rule reaches translated editions too
Amazon's content guidelines extend AI-generated disclosure to translations, and its own translation tool adds a licensing term.
3 min · source-led record↗
ChatGPT's business tiers default away from training use
OpenAI's own pages show consumer plans default to training use with an opt-out, while business tiers default to none.
3 min · source-led record↗
ElevenLabs requires authorization before it clones a voice
ElevenLabs' own terms and use policy require consent or legal right before a real person's voice can be replicated.
3 min · source-led record↗
DeepL's privacy terms are more checkable than its quality claim
DeepL separates a marketed accuracy claim from a concrete difference in what a free versus a paid account retains.
3 min · source-led record↗
Gemini Notebook cites the exact line an answer drew from
Google's help pages describe quote-level citations restricted to a user's chosen sources, not open web retrieval.
3 min · source-led record↗
Perplexity tells its own users to check every citation
Perplexity's own documentation states each answer links to numbered citations that the user is expected to verify.
3 min · source-led record↗
Amazon's 70 percent option comes with a price floor
KDP's own royalty page shows the 70 percent ebook rate is conditional on territory, price floor and file size, not a flat number.
3 min · source-led record↗
Apple's ebook royalty holds steady across every price
Apple's own documentation states a flat 70 percent ebook royalty with no exclusivity requirement, a different bargain from Amazon's tiered program.
3 min · source-led record↗
Kobo's royalty rate depends on crossing a price line
Kobo's own earnings page ties its ebook royalty rate to a price threshold rather than to the retailer alone.
3 min · source-led record↗
Barnes and Noble Press ties its print rate to printing cost
Barnes & Noble Press documents a flat ebook royalty and a print royalty set against a variable per-book printing cost.
2 min · source-led record↗
IngramSpark separates its fees from its print cost
IngramSpark's 2025 rate card shows title set-up and pre-approval revisions are free, with cost concentrated in a per-sale market-access fee.
3 min · source-led record↗
Draft2Digital pairs a flat fee with a per-sale commission
Draft2Digital's own FAQ and terms show a flat activation fee, an annual maintenance fee and a nonexclusive license alongside its commission.
3 min · source-led record↗
PublishDrive sells shelf space, not a cut of royalties
PublishDrive's own pricing page sells tiered subscriptions with a stated 0 percent commission, a different bet from percentage-based aggregators.
2 min · source-led record↗
BookBaby pays more when a reader buys direct
BookBaby's own pages show a direct-storefront royalty well above its stated range for sales through other retailers.
2 min · source-led record↗
Lulu's storefront pays more than its wide reach does
Lulu's own selling page ties a higher royalty share to its own storefront and a lower-friction path to wide retail distribution.
2 min · source-led record↗
Reedsy prices self-publishing from its own quote data
Reedsy's 2026 guide prices self-publishing services from its marketplace's quote data, labeled indicative rather than a survey finding.
2 min · source-led record↗
Print still led a record year for UK publishing
The Publishers Association's own 2025 figures show UK revenue at a record high while print still supplies most consumer sales.
3 min · source-led record↗
TikTok's BookTok sales number is a platform's own claim
TikTok cites NielsenIQ and Media Control data for 50 million BookTok-linked sales in Europe, a platform claim, not proof of causation.
3 min · source-led record↗
Sanderson's record Kickstarter took years of quiet prep
Kickstarter's own data and retrospective show a $41.75 million campaign built on an existing fanbase and a 30-person publisher.
3 min · source-led record↗
NetGalley charges by listing, not by review received
NetGalley's own pricing page prices advance-copy distribution per listing and promotion, not per review, as retrieved 16 September 2026.
3 min · source-led record↗
BookSirens prices reviews by reader, not by review
BookSirens' own pricing page ties author fees to reader downloads, with a stated average review rate it does not guarantee.
3 min · source-led record↗
A book review's free-copy origin still needs disclosure
FTC guidance and its 2024 reviews rule both reach book reviews obtained through free copies, regardless of who or what wrote them.
3 min · source-led record↗
Goodreads polices review manipulation, not just opinion
Goodreads' own guidelines separate rating manipulation and undisclosed free copies from ordinary negative reviews.
3 min · source-led record↗
Amazon's review rules ban conduct, not how text was made
Amazon's own community guidelines prohibit compensated, fake and seller-linked reviews, rules written without regard to a review's origin.
3 min · source-led record↗
APA's sales survey and its consumer survey differ
The Audio Publishers Association's own release separates a Toluna-run sales survey from an Edison Research consumer survey.
3 min · source-led record↗
An editor's own numbers closed a magazine's submission queue
Neil Clarke's blog post on Clarkesworld's 2023 spam spike shows what an AI-writing flood cost one small magazine.
3 min · source-led record↗
A journal standard defines authorship to exclude any AI tool
ICMJE's recommendations tie authorship to accountability, a threshold its own guidance says a chatbot cannot meet.
3 min · source-led record↗
A publishing-ethics group also rules out an AI byline
COPE's position statement says AI tools cannot meet authorship criteria and must be disclosed wherever they contribute.
2 min · source-led record↗
Medical editors warn a chatbot's citations may not be real
WAME's recommendations bar chatbot authorship and require authors to verify that a citation actually supports what a chatbot claims.
2 min · source-led record↗
A writers' group drafted contract language against silent AI use
The Authors Guild's model clauses let an author reserve, license or require approval over specific AI uses of their book.
2 min · source-led record↗
A writers' union's own survey ties AI use to falling income
The Society of Authors reports member-survey findings on falling income and offers a label for books written without generative AI.
2 min · source-led record↗
A free-expression group's AI paper is not a copyright brief
PEN America's Speech in the Machine sets out free-expression risks from generative AI, distinct from copyright claims in court.
2 min · source-led record↗
A scholarly publisher separates AI citation from AI authorship
APA's journals policy requires citing a generative AI tool used in a manuscript while barring the tool from being named an author.
3 min · source-led record↗
A wire service bars AI from photos, not from drafting text
AP's generative AI standards flatly prohibit altering news photos while treating AI text output as unvetted source material.
3 min · source-led record↗
A certification mark targets AI training data, not book bylines
Fairly Trained's Licensed Model certification flags AI companies that license training data, without ruling on any book's authorship.
3 min · source-led record↗
One writing tool keeps brainstorming apart from full drafting
Sudowrite's staged outline tools and Notion AI's blended prompts show two different vendor answers to where assistance starts.
3 min · source-led record↗
Grammarly and Google Docs define tone editing narrowly
Both vendors document a tone or rewrite feature that changes register, not claims, a narrower job than full drafting.
3 min · source-led record↗
A confidentiality checklist differs by AI product tier
OpenAI, Anthropic and ProWritingAid document different training and retention rules depending on which product tier is used.
3 min · source-led record↗
Track Changes attributes edits by name, not by author type
Word's documentation shows tracked changes record a self-entered name, with no built-in category for AI-generated edits.
3 min · source-led record↗
Two AI vendors tell users to verify output before use
OpenAI and Notion each instruct users to check generated text against sources before treating it as finished.
3 min · source-led record↗
Version history shows who edited, not what kind of edit
Google's and Microsoft's documentation show version history records timing and access, not whether text was AI-generated.
3 min · source-led record↗
APA and WAME set different AI disclosure rules for papers
APA requires a methods-section disclosure, while WAME's recommendations ask for the prompts and dates behind the drafting.
3 min · source-led record↗
A manuscript tool says its beta reader is not an editor
ProWritingAid's own pages frame its Virtual Beta Reader report as feedback before human readers, not a substitute for one.
3 min · source-led record↗
DeepL and KDP treat machine translation as a starting point
DeepL's benchmark page cites reduced post-editing, and KDP's rule keeps a machine translation AI-generated after edits.
3 min · source-led record↗
Research tools keep a source ledger before AI touches prose
Elicit's and Scite's documentation describe sentence-level citation records that a claim ledger practice would draw on.
3 min · source-led record↗
Readers rated AI-generated poems as more human than real ones
A Scientific Reports study found non-expert readers misjudged AI poems as human more often than they misjudged real poems as AI.
3 min · source-led record↗
One writing assistant nudged many writers toward similar ideas
A 33-participant study found ChatGPT users produced less distinct ideas than users of a non-AI tool, and AI-written book reviews cluster narrowly too.
3 min · source-led record↗
Most Americans expect AI to make creative thinking worse
A representative Pew survey found 53 percent of U.S. adults expect AI to worsen creative thinking, versus 16 percent who expect improvement.
3 min · source-led record↗
An opinionated writing assistant shifted what people believed
A CHI 2023 experiment found a one-sided writing assistant shifted both what 1,506 participants wrote and what they later said they believed.
3 min · source-led record↗
ChatGPT access cut writing time and lifted quality in a trial
A Science-published experiment found ChatGPT access cut professional writing time by 40 percent and raised rated quality by 18 percent among 453 workers.
3 min · source-led record↗
An annual benchmark scored machine translation of web fiction
The WMT24 findings paper reports 10 systems judged mainly by humans on Chinese-English web fiction, a narrower scope than the task page had announced.
3 min · source-led record↗
AI-text detectors flagged most non-native writers' essays
A Stanford study found seven GPT detectors misclassified 61 percent of non-native English essays as AI-written, versus near-perfect accuracy on native essays.
3 min · source-led record↗
Word-frequency analysis found AI edits in conference reviews
A Stanford method estimates 6.5-16.9 percent of reviews at four AI conferences were AI-modified; a related study found up to 17.5 percent in CS papers overall.
3 min · source-led record↗
A readership survey found researchers split on chatbot use
Nature's survey of about 1,600 self-selected researchers found many use generative AI for manuscripts; its methodology says the sample is not representative.
3 min · source-led record↗
A classifier caught AI-written book reviews with high accuracy
A transfer-learning classifier reached 96.86 percent accuracy distinguishing real Amazon book reviews from reviews generated by the Vicuna language model.
3 min · source-led record↗
A famous word-of-mouth study measured rank, not sales
Chevalier and Mayzlin's own paper explains it tracked Amazon and Barnes & Noble sales ranks from two days in May 2003, not verified sales.
3 min · source-led record↗