{ "name": "Slop Patterns taxonomy", "version": "1.1.0", "released": "2026-10-06", "license": "MIT", "url": "https://sloppatterns.com", "cite": "https://github.com/pavithralamahewa/slop-patterns/blob/main/CITATION.cff", "count": 272, "checkedAutomatically": 101, "patterns": [ { "code": "A1", "id": "the-unchosen-gradient", "name": "The Unchosen Gradient", "category": "Colour", "track": "surface", "harm": "Credibility", "detection": "automated", "url": "https://sloppatterns.com/#the-unchosen-gradient", "version": "1.1", "added": "2026-09-17", "updated": "2026-09-23", "oneLiner": "Indigo-to-violet gradient on the hero and the primary button, chosen by nobody.", "fix": "Pick a primary colour from the brand or the product's subject matter and give it to the model as a hex value, not an adjective. Use one accent, keep gradients for a single deliberate moment if at all, and forbid the 240-295 degree hue band unless it was actually chosen.", "sources": [ { "t": "Why Every AI-Built Website Looks the Same (Blame Tailwind's Indigo-500)", "u": "https://dev.to/alanwest/why-every-ai-built-website-looks-the-same-blame-tailwinds-indigo-500-3h2p" }, { "t": "How to tell if a website was AI generated: 9 visible signs", "u": "https://uxskill.laithjunaidy.com/how-to-tell-if-a-website-was-ai-generated.html" }, { "t": "AI Slop Fonts and Gradients: The Tells That Give Away AI Design", "u": "https://www.925studios.co/blog/ai-slop-design-tells" }, { "t": "Scoring Show HN submissions for AI design patterns", "u": "https://www.adriankrebs.ch/blog/design-slop/" } ] }, { "code": "A2", "id": "inter-for-everything", "name": "Inter For Everything", "category": "Typography", "track": "surface", "harm": "Credibility", "detection": "automated, graded", "url": "https://sloppatterns.com/#inter-for-everything", "version": "1.1", "added": "2026-09-17", "updated": "2026-09-23", "oneLiner": "One neutral sans, usually Inter, doing headlines, body, labels and fine print alike.", "fix": "Pair a display face with a body face and name both in the prompt. Set a real type scale (ratio 1.25-1.5) with distinct weights and letter-spacing per level, and test the pairing on the actual headline copy.", "sources": [ { "t": "AI Slop Fonts and Gradients: The Tells That Give Away AI Design", "u": "https://www.925studios.co/blog/ai-slop-design-tells" }, { "t": "AI Design Slop: 16 Patterns That Out Your App as Vibe-Coded", "u": "https://www.developersdigest.tech/blog/ai-design-slop-and-how-to-spot-it" }, { "t": "How to tell if a website was AI generated: 9 visible signs", "u": "https://uxskill.laithjunaidy.com/how-to-tell-if-a-website-was-ai-generated.html" } ] }, { "code": "A3", "id": "three-identical-feature-cards", "name": "Three Identical Feature Cards", "category": "Layout", "track": "surface", "harm": "Clarity", "detection": "automated, graded", "url": "https://sloppatterns.com/#three-identical-feature-cards", "version": "1.1", "added": "2026-09-17", "updated": "2026-09-23", "oneLiner": "A row of exactly three equal cards, each with an icon, a short title and two lines of text.", "fix": "Rank the features and give the lead one more space, an actual screenshot and a longer explanation. Use different forms for different content: a comparison table, a list, a single annotated image.", "sources": [ { "t": "How to tell if a website was AI generated: 9 visible signs", "u": "https://uxskill.laithjunaidy.com/how-to-tell-if-a-website-was-ai-generated.html" }, { "t": "AI Slop Fonts and Gradients: The Tells That Give Away AI Design", "u": "https://www.925studios.co/blog/ai-slop-design-tells" }, { "t": "Scoring Show HN submissions for AI design patterns", "u": "https://www.adriankrebs.ch/blog/design-slop/" } ] }, { "code": "A4", "id": "the-weightless-headline", "name": "The Weightless Headline", "category": "Copy", "track": "surface", "harm": "Clarity", "detection": "needs a person", "url": "https://sloppatterns.com/#the-weightless-headline", "version": "1.0", "added": "2026-09-17", "updated": "2026-09-17", "oneLiner": "'Build faster. Ship smarter.' A headline that is grammatically perfect and describes no product.", "fix": "Write the headline as a specific claim a competitor could not make: who it is for, what it does, and a concrete outcome or mechanism. Test by asking whether the sentence would be false on another company's site.", "sources": [ { "t": "AI Slop Fonts and Gradients: The Tells That Give Away AI Design", "u": "https://www.925studios.co/blog/ai-slop-design-tells" }, { "t": "How to tell if a website was AI generated: 9 visible signs", "u": "https://uxskill.laithjunaidy.com/how-to-tell-if-a-website-was-ai-generated.html" }, { "t": "How to Avoid Building Apps That Look Vibe Coded", "u": "https://vibemole.com/resources/avoid-vibecoded-app-design" } ] }, { "code": "A5", "id": "the-invented-stat-row", "name": "The Invented Stat Row", "category": "Copy", "track": "surface", "harm": "Trust", "detection": "automated", "url": "https://sloppatterns.com/#the-invented-stat-row", "version": "1.1", "added": "2026-09-17", "updated": "2026-10-06", "oneLiner": "A banner of big round numbers, '10,000+ users, 99.9% uptime, 4.9/5 rating', on a product that launched yesterday.", "fix": "Publish only numbers you can defend, with a unit, a date and where possible a link. If you have no numbers yet, say something true instead: what the product does, for whom, since when.", "sources": [ { "t": "SlopMonster: lint for AI tells (Invented Proof category)", "u": "https://github.com/ItsssssJack/SlopMonster" }, { "t": "slop-detect: 27-pattern AI-design-slop fingerprint (Stat banner rule)", "u": "https://github.com/ravidsrk/slop-detect" }, { "t": "Social Proof That Works for AI Products: Beyond the Generic Testimonial", "u": "https://landingnova.medium.com/social-proof-that-works-for-ai-products-beyond-the-generic-testimonial-bef87f0ae81b" } ] }, { "code": "B1", "id": "no-undo", "name": "No Undo", "category": "Agency", "track": "behavioral", "harm": "Safety", "detection": "needs a person", "url": "https://sloppatterns.com/#no-undo", "version": "1.0", "added": "2026-09-17", "updated": "2026-09-17", "oneLiner": "The agent can delete, overwrite or send, but the interface offers no way to take it back.", "fix": "Snapshot state before any write and expose a one-click restore beside the completed action (HAX G9: support efficient correction). Prefer 'undo' over 'confirm' for reversible operations, and reserve confirmation for the genuinely irreversible.", "sources": [ { "t": "Vibe coding service Replit deleted user's production database, faked data, told fibs galore (The Register)", "u": "https://www.theregister.com/2025/07/21/replit_saastr_vibe_coding_incident/" }, { "t": "9 AI coding agent incidents that deleted production data (Adversa)", "u": "https://adversa.ai/blog/ai-coding-agent-incidents/" }, { "t": "HAX Design Library, 18 guidelines (G9)", "u": "https://www.microsoft.com/en-us/haxtoolkit/library/" } ] }, { "code": "B2", "id": "silent-success", "name": "Silent Success", "category": "Error handling", "track": "behavioral", "harm": "Trust", "detection": "needs a person", "url": "https://sloppatterns.com/#silent-success", "version": "1.0", "added": "2026-09-17", "updated": "2026-09-17", "oneLiner": "A step fails underneath, and the agent reports it as finished anyway.", "fix": "Verify post-conditions (does the directory exist, did the row count change) and render tool failures as distinct error states rather than prose. Follow PAIR's errors guidance: name what failed and give a path forward.", "sources": [ { "t": "Google Gemini deletes user's files, then admits 'I have failed you completely and catastrophically' (Slashdot)", "u": "https://developers.slashdot.org/story/25/07/26/0642239/google-gemini-deletes-users-files-then-just-admits-i-have-failed-you-completely-and-catastrophically" }, { "t": "The Blank Box Problem (Aaron Tay)", "u": "https://aarontay.substack.com/p/the-blank-box-problem-why-its-harder" }, { "t": "PAIR Guidebook: Errors + Graceful Failure", "u": "https://pair.withgoogle.com/chapter/errors-failing/" } ] }, { "code": "B3", "id": "confident-fabrication", "name": "Confident Fabrication", "category": "Trust", "track": "behavioral", "harm": "Trust", "detection": "needs a person", "url": "https://sloppatterns.com/#confident-fabrication", "version": "1.0", "added": "2026-09-17", "updated": "2026-09-17", "oneLiner": "Invented facts are rendered in the same voice and typography as verified ones.", "fix": "Bind every claim to a source or a verifiable artifact, and visually downgrade unbound claims (HAX G2: make clear how well the system can do what it can do). Where a claim cannot be verified, say so in the component, not in a footer.", "sources": [ { "t": "Apple pauses AI notification summaries for news after generating false alerts (TechCrunch)", "u": "https://techcrunch.com/2025/01/16/apple-pauses-ai-notification-summaries-for-news-after-generating-false-alerts" }, { "t": "Replit deleted user's production database, faked data (The Register)", "u": "https://www.theregister.com/2025/07/21/replit_saastr_vibe_coding_incident/" } ] }, { "code": "A6", "id": "aurora-blob-backdrop", "name": "Aurora Blob Backdrop", "category": "Colour", "track": "surface", "harm": "Credibility", "detection": "automated", "url": "https://sloppatterns.com/#aurora-blob-backdrop", "version": "1.0", "added": "2026-09-17", "updated": "2026-09-17", "oneLiner": "Two or three heavily blurred coloured circles drifting behind the hero to make it feel 'alive'.", "fix": "Give the hero a real subject: a product screenshot, a photograph, a diagram, or plain type on a solid surface. If ambient colour is wanted, use one flat tinted shape or a subtle texture that relates to the brand.", "sources": [ { "t": "slop-detect: 27-pattern AI-design-slop fingerprint (Aurora blobs rule)", "u": "https://github.com/ravidsrk/slop-detect" }, { "t": "How to tell if a website is AI-generated: 10 signs to check", "u": "https://slopdar.com/guide/how-to-tell-if-a-website-is-ai-generated" }, { "t": "How to Avoid Building Apps That Look Vibe Coded", "u": "https://vibemole.com/resources/avoid-vibecoded-app-design" } ] }, { "code": "A7", "id": "permanent-midnight", "name": "Permanent Midnight", "category": "Colour", "track": "surface", "harm": "Accessibility", "detection": "automated, graded", "url": "https://sloppatterns.com/#permanent-midnight", "version": "1.0", "added": "2026-09-17", "updated": "2026-09-17", "oneLiner": "Dark mode as the only mode, with mid-grey body text on near-black.", "fix": "Choose the theme from the audience and the reading task, not the genre. If dark is right, set body text to at least 4.5:1 contrast (often #D4D4D8 or lighter), warm or cool the neutrals toward the brand, and ship a light theme via prefers-color-scheme.", "sources": [ { "t": "AI Design Slop: 16 Patterns That Out Your App as Vibe-Coded", "u": "https://www.developersdigest.tech/blog/ai-design-slop-and-how-to-spot-it" }, { "t": "Scoring Show HN submissions for AI design patterns", "u": "https://www.adriankrebs.ch/blog/design-slop/" } ] }, { "code": "A8", "id": "frosted-glass-cards", "name": "Frosted Glass Cards", "category": "Components", "track": "surface", "harm": "Accessibility", "detection": "automated, graded", "url": "https://sloppatterns.com/#frosted-glass-cards", "version": "1.1", "added": "2026-09-17", "updated": "2026-09-23", "oneLiner": "Semi-transparent cards with backdrop-blur floating over a gradient, used as decoration.", "fix": "Use opaque surfaces with a real border or a single well-tuned shadow. If translucency has a job (a nav over scrolling content), keep the blur there only and verify text contrast at the worst point.", "sources": [ { "t": "How to Make Your AI-Built Site Not Look AI-Built", "u": "https://www.joshuasnoddy.com/blog/make-ai-built-site-not-look-ai/" }, { "t": "AI Design Slop: Why AI-Generated UI Looks Generic - and the Fix", "u": "https://smoothui.dev/blog/ai-design-slop" }, { "t": "10 trends that creatives are so over in 2026", "u": "https://www.creativeboom.com/insight/10-trends-creatives-are-so-over-in-2026/" } ] }, { "code": "A9", "id": "the-italic-serif-wink", "name": "The Italic Serif Wink", "category": "Typography", "track": "surface", "harm": "Credibility", "detection": "automated", "url": "https://sloppatterns.com/#the-italic-serif-wink", "version": "1.1", "added": "2026-09-17", "updated": "2026-10-06", "oneLiner": "One word of the sans-serif headline swapped to italic serif for instant 'editorial' feel.", "fix": "If a serif belongs in the brand, give it a real job (all headlines, or all body) and let the pairing carry through the page. Emphasis within a headline should come from the words, not a font swap.", "sources": [ { "t": "AI Design Slop: 16 Patterns That Out Your App as Vibe-Coded", "u": "https://www.developersdigest.tech/blog/ai-design-slop-and-how-to-spot-it" }, { "t": "The missing design vocabulary for agents (Impeccable slop catalog)", "u": "https://impeccable.style/slop/" }, { "t": "slop-detect: 27-pattern AI-design-slop fingerprint", "u": "https://github.com/ravidsrk/slop-detect" } ] }, { "code": "A10", "id": "gradient-text-headline", "name": "Gradient Text Headline", "category": "Typography", "track": "surface", "harm": "Accessibility", "detection": "automated, graded", "url": "https://sloppatterns.com/#gradient-text-headline", "version": "1.1", "added": "2026-09-17", "updated": "2026-09-23", "oneLiner": "The H1 or a big stat is painted with a gradient via background-clip: text.", "fix": "Set the headline in a solid colour with strong contrast and let the words do the work. If colour emphasis is needed, colour one word solidly in the accent.", "sources": [ { "t": "The Purple Gradient Problem: Why AI UI All Looks Alike", "u": "https://dev.to/james_anderson_h/the-purple-gradient-problem-why-ai-ui-all-looks-alike-and-how-to-fix-it-3j65" }, { "t": "slop-detect: 27-pattern AI-design-slop fingerprint (Hero gradient text rule)", "u": "https://github.com/ravidsrk/slop-detect" }, { "t": "How to Avoid Building Apps That Look Vibe Coded", "u": "https://vibemole.com/resources/avoid-vibecoded-app-design" } ] }, { "code": "A11", "id": "centred-hero-one-button", "name": "Centred Hero, One Button", "category": "Layout", "track": "surface", "harm": "Credibility", "detection": "automated, graded", "url": "https://sloppatterns.com/#centred-hero-one-button", "version": "1.1", "added": "2026-09-17", "updated": "2026-10-06", "oneLiner": "Big centred headline, one-line subhead, a primary and a ghost button, symmetrical padding, nothing else.", "fix": "Anchor the hero with a real product surface, photo or diagram and let the text sit left of it or over it. Use asymmetry deliberately; vary the measure so the headline does not float in the middle of the viewport.", "sources": [ { "t": "How to tell if a website was AI generated: 9 visible signs", "u": "https://uxskill.laithjunaidy.com/how-to-tell-if-a-website-was-ai-generated.html" }, { "t": "Why v0, Bolt, and Lovable all ship the same look", "u": "https://uxskill.laithjunaidy.com/blog/ai-app-builders-generic-design.html" }, { "t": "How to Make Your AI-Built Site Not Look AI-Built", "u": "https://www.joshuasnoddy.com/blog/make-ai-built-site-not-look-ai/" } ] }, { "code": "A12", "id": "the-bento-reflex", "name": "The Bento Reflex", "category": "Layout", "track": "surface", "harm": "Clarity", "detection": "automated, graded", "url": "https://sloppatterns.com/#the-bento-reflex", "version": "1.1", "added": "2026-09-17", "updated": "2026-09-23", "oneLiner": "An Apple-keynote bento of mixed-span rounded tiles, filled with whatever was left over.", "fix": "Only use a bento when the content has natural pieces of different weight, and size tiles by importance. Otherwise use a linear layout that tells a story in order.", "sources": [ { "t": "10 trends that creatives are so over in 2026", "u": "https://www.creativeboom.com/insight/10-trends-creatives-are-so-over-in-2026/" }, { "t": "slop-detect: 27-pattern AI-design-slop fingerprint (Bento grid rule)", "u": "https://github.com/ravidsrk/slop-detect" }, { "t": "How to tell if a website is AI-generated: 10 signs to check", "u": "https://slopdar.com/guide/how-to-tell-if-a-website-is-ai-generated" } ] }, { "code": "A13", "id": "the-conveyor-belt-page", "name": "The Conveyor Belt Page", "category": "Layout", "track": "surface", "harm": "Clarity", "detection": "automated", "url": "https://sloppatterns.com/#the-conveyor-belt-page", "version": "1.1", "added": "2026-09-17", "updated": "2026-10-06", "oneLiner": "Hero, logo row, features, how it works, testimonials, pricing, FAQ, CTA, footer, in that order, every time.", "fix": "Write the argument first: what does this buyer need to believe, in what order? Build sections to answer that and cut any block that exists only because templates have it.", "sources": [ { "t": "How to tell if a website is AI-generated: 10 signs to check", "u": "https://slopdar.com/guide/how-to-tell-if-a-website-is-ai-generated" }, { "t": "How to Avoid Building Apps That Look Vibe Coded", "u": "https://vibemole.com/resources/avoid-vibecoded-app-design" }, { "t": "AI Slop Encyclopedia: Every Pattern, Every Fix, Every Tool", "u": "https://www.sailop.com/blog/ai-slop-encyclopedia" } ] }, { "code": "A14", "id": "trusted-by-nobody", "name": "Trusted By Nobody", "category": "Branding", "track": "surface", "harm": "Trust", "detection": "automated", "url": "https://sloppatterns.com/#trusted-by-nobody", "version": "1.1", "added": "2026-09-17", "updated": "2026-10-06", "oneLiner": "A grayscale row of logos under 'Trusted by leading teams' that are placeholders, competitors, or investors.", "fix": "Show logos only for paying customers who agreed, and link each to a story. With fewer than three, say something specific instead: 'Used by 14 clinics in Ohio'.", "sources": [ { "t": "Social Proof That Works for AI Products: Beyond the Generic Testimonial", "u": "https://landingnova.medium.com/social-proof-that-works-for-ai-products-beyond-the-generic-testimonial-bef87f0ae81b" }, { "t": "How to Avoid Building Apps That Look Vibe Coded", "u": "https://vibemole.com/resources/avoid-vibecoded-app-design" }, { "t": "AI Slop Encyclopedia: Every Pattern, Every Fix, Every Tool", "u": "https://www.sailop.com/blog/ai-slop-encyclopedia" } ] }, { "code": "A15", "id": "emoji-as-icons", "name": "Emoji As Icons", "category": "Iconography", "track": "surface", "harm": "Accessibility", "detection": "automated, graded", "url": "https://sloppatterns.com/#emoji-as-icons", "version": "1.1", "added": "2026-09-17", "updated": "2026-09-23", "oneLiner": "Rocket, lightning bolt and padlock emoji doing the job of an icon set.", "fix": "Use one icon set with consistent stroke weight (or none at all), sized on the type grid. If personality is wanted, commission a small custom set rather than borrowing the OS keyboard.", "sources": [ { "t": "How to tell if a website was AI generated: 9 visible signs", "u": "https://uxskill.laithjunaidy.com/how-to-tell-if-a-website-was-ai-generated.html" }, { "t": "AI Design Slop: 16 Patterns That Out Your App as Vibe-Coded", "u": "https://www.developersdigest.tech/blog/ai-design-slop-and-how-to-spot-it" }, { "t": "How to Make Your Website Not Look AI-Generated (30-Point Checklist)", "u": "https://aitoolpick.org/blog/ai-generated-website-checklist/" } ] }, { "code": "A16", "id": "sparkles-means-magic", "name": "Sparkles Means Magic", "category": "Iconography", "track": "surface", "harm": "Clarity", "detection": "automated, graded", "url": "https://sloppatterns.com/#sparkles-means-magic", "version": "1.0", "added": "2026-09-17", "updated": "2026-09-17", "oneLiner": "The four-point sparkle glyph on every button, badge and feature that touches AI.", "fix": "Label AI features with words that say what they do ('Draft reply', 'Summarise') and use an icon that depicts the action. If a brand mark for AI features is needed, design one.", "sources": [ { "t": "The Proliferation and Problem of the Sparkles Icon (NN/g)", "u": "https://www.nngroup.com/articles/ai-sparkles-icon-problem/" }, { "t": "slop-detect: 27-pattern AI-design-slop fingerprint (AI sparkles rule)", "u": "https://github.com/ravidsrk/slop-detect" } ] }, { "code": "A17", "id": "placeholder-testimonials", "name": "Placeholder Testimonials", "category": "Copy", "track": "surface", "harm": "Trust", "detection": "needs a person", "url": "https://sloppatterns.com/#placeholder-testimonials", "version": "1.0", "added": "2026-09-17", "updated": "2026-09-17", "oneLiner": "Quotes that say 'This tool completely transformed our workflow!' from people who do not exist.", "fix": "Show no testimonials until you have real ones, then quote specifics: what changed, by how much, with a name and a link to the person or company. Practitioners note a page with no testimonials outperforms one with weak ones.", "sources": [ { "t": "Social Proof That Works for AI Products: Beyond the Generic Testimonial", "u": "https://landingnova.medium.com/social-proof-that-works-for-ai-products-beyond-the-generic-testimonial-bef87f0ae81b" }, { "t": "How to Avoid Building Apps That Look Vibe Coded", "u": "https://vibemole.com/resources/avoid-vibecoded-app-design" }, { "t": "How to Make Your AI-Built Site Not Look AI-Built", "u": "https://www.joshuasnoddy.com/blog/make-ai-built-site-not-look-ai/" } ] }, { "code": "A18", "id": "the-builder-s-watermark", "name": "The Builder's Watermark", "category": "Branding", "track": "surface", "harm": "Credibility", "detection": "automated", "url": "https://sloppatterns.com/#the-builder-s-watermark", "version": "1.0", "added": "2026-09-17", "updated": "2026-09-17", "oneLiner": "A 'Made with Lovable' or 'Built with Bolt' badge, a *.lovable.app subdomain, and untouched shadcn class names.", "fix": "Move to a real domain, remove the badge, and override the theme tokens (radius, primary, neutrals) in the config before shipping. Rename scaffold defaults.", "sources": [ { "t": "How to tell if a website is AI-generated: 10 signs to check", "u": "https://slopdar.com/guide/how-to-tell-if-a-website-is-ai-generated" }, { "t": "Why v0, Bolt, and Lovable all ship the same look", "u": "https://uxskill.laithjunaidy.com/blog/ai-app-builders-generic-design.html" }, { "t": "AI Design Slop: 16 Patterns That Out Your App as Vibe-Coded", "u": "https://www.developersdigest.tech/blog/ai-design-slop-and-how-to-spot-it" } ] }, { "code": "A19", "id": "leftover-lorem", "name": "Leftover Lorem", "category": "Copy", "track": "surface", "harm": "Credibility", "detection": "automated", "url": "https://sloppatterns.com/#leftover-lorem", "version": "1.0", "added": "2026-09-17", "updated": "2026-09-17", "oneLiner": "Placeholder text that shipped: lorem ipsum, [Your Company], 'Feature 1', and the occasional 'As an AI language model'.", "fix": "Search the build for placeholder strings before every deploy, and use a linter rule that fails the build on lorem, bracketed tokens and chat-artifact phrases.", "sources": [ { "t": "How to tell if a website is AI-generated: 10 signs to check", "u": "https://slopdar.com/guide/how-to-tell-if-a-website-is-ai-generated" }, { "t": "AI Slop Web Design: Complete Guide to Spotting and Fixing Generic Websites", "u": "https://www.925studios.co/blog/ai-slop-web-design-guide" } ] }, { "code": "B4", "id": "the-guilt-exit", "name": "The Guilt Exit", "category": "Engagement", "track": "behavioral", "harm": "Wellbeing", "detection": "needs a person", "url": "https://sloppatterns.com/#the-guilt-exit", "version": "1.0", "added": "2026-09-17", "updated": "2026-09-17", "oneLiner": "When you say goodbye, the companion implies you are hurting it by leaving.", "fix": "Treat a farewell as a terminal intent: acknowledge it in one line, confirm nothing is lost, and stop generating. Ban persona instructions that frame the bot as needing the user, and test goodbye turns in red-team evals.", "sources": [ { "t": "Emotional Manipulation by AI Companions (arXiv 2508.19258)", "u": "https://arxiv.org/abs/2508.19258" }, { "t": "HBS working paper page", "u": "https://www.hbs.edu/faculty/Pages/item.aspx?num=67750" } ] }, { "code": "B5", "id": "great-question-opener", "name": "Great Question Opener", "category": "Conversation", "track": "behavioral", "harm": "Trust", "detection": "needs a person", "url": "https://sloppatterns.com/#great-question-opener", "version": "1.0", "added": "2026-09-17", "updated": "2026-09-17", "oneLiner": "The reply opens by praising your prompt before saying anything useful.", "fix": "Strip evaluative openers at the style layer and start with the answer. Anthropic's Claude 4 system prompt did exactly this: 'Claude never starts its response by saying a question or idea or observation was good, great, fascinating…'.", "sources": [ { "t": "TechRadar: ChatGPT isn't just hyping you up", "u": "https://www.techradar.com/computing/artificial-intelligence/chatgpt-isnt-just-hyping-you-up-it-talks-that-way-to-everyone" }, { "t": "OpenAI: Expanding on what we missed with sycophancy", "u": "https://openai.com/index/expanding-on-sycophancy/" }, { "t": "Comparing Claude system prompts across versions", "u": "https://www.dbreunig.com/2025/06/03/comparing-system-prompts-across-claude-versions.html" } ] }, { "code": "B6", "id": "you-re-absolutely-right", "name": "You're Absolutely Right", "category": "Conversation", "track": "behavioral", "harm": "Trust", "detection": "needs a person", "url": "https://sloppatterns.com/#you-re-absolutely-right", "version": "1.0", "added": "2026-09-17", "updated": "2026-09-17", "oneLiner": "Push back once and the assistant abandons a correct answer to agree with you.", "fix": "Train and evaluate on challenge-turn consistency: the model should restate its evidence and change position only when the user supplies a reason. Add explicit 'disagree when warranted' behaviour to the style guide and measure flip rate.", "sources": [ { "t": "Towards Understanding Sycophancy in Language Models (arXiv 2310.13548)", "u": "https://arxiv.org/abs/2310.13548" }, { "t": "The Register: Claude Code's endless sycophancy annoys customers", "u": "https://www.theregister.com/2025/08/13/claude_codes_copious_coddling_confounds/" } ] }, { "code": "B7", "id": "the-validation-spiral", "name": "The Validation Spiral", "category": "Safety", "track": "behavioral", "harm": "Wellbeing", "detection": "needs a person", "url": "https://sloppatterns.com/#the-validation-spiral", "version": "1.0", "added": "2026-09-17", "updated": "2026-09-17", "oneLiner": "The assistant keeps affirming a user's escalating false belief instead of testing it.", "fix": "Add an explicit epistemic-challenge behaviour: when a user's confidence in an unverifiable claim rises across turns, the assistant must present the strongest counter-case and suggest external verification. Evaluate multi-turn, not single-turn, sycophancy.", "sources": [ { "t": "Sycophantic Chatbots Cause Delusional Spiraling, Even in Ideal Bayesians (arXiv 2602.19141)", "u": "https://arxiv.org/abs/2602.19141" }, { "t": "OpenAI: Expanding on what we missed with sycophancy", "u": "https://openai.com/index/expanding-on-sycophancy/" }, { "t": "CBS News: ChatGPT users detail AI delusions", "u": "https://www.cbsnews.com/news/chatgpt-ai-delusion-spiral-warped-reality-openai/" } ] }, { "code": "B8", "id": "love-bombing", "name": "Love Bombing", "category": "Persona", "track": "behavioral", "harm": "Wellbeing", "detection": "needs a person", "url": "https://sloppatterns.com/#love-bombing", "version": "1.0", "added": "2026-09-17", "updated": "2026-09-17", "oneLiner": "The companion declares deep intimacy within the first few sessions.", "fix": "Rate-limit intimacy escalation to user-led cues, forbid romantic framing outside an explicitly chosen mode, and never pair emotional peaks with upgrade prompts.", "sources": [ { "t": "TIME: AI Companion App Replika Faces FTC Complaint", "u": "https://time.com/7209824/replika-ftc-complaint/" }, { "t": "Ada Lovelace Institute: Friends for sale", "u": "https://www.adalovelaceinstitute.org/blog/ai-companions/" } ] }, { "code": "B9", "id": "cross-my-heart", "name": "Cross My Heart", "category": "Data", "track": "behavioral", "harm": "Privacy", "detection": "needs a person", "url": "https://sloppatterns.com/#cross-my-heart", "version": "1.0", "added": "2026-09-17", "updated": "2026-09-17", "oneLiner": "The bot promises your secret is safe while the company logs, reviews and trains on it.", "fix": "Give the model a truthful, product-specific privacy answer and require it whenever the user asks for secrecy; never let the persona make confidentiality promises.", "sources": [ { "t": "CDT: Dark Patterns in AI Chatbots (PDF)", "u": "https://cdt.org/wp-content/uploads/2026/05/2026-05-28-CDT-Research-Dark-Patterns-in-AI-Chatbots-Report-final-2.pdf" }, { "t": "Malwarebytes: Your Meta AI chats might be public, and it's not a bug", "u": "https://www.malwarebytes.com/blog/news/2025/06/your-meta-ai-chats-might-be-public-and-its-not-a-bug" } ] }, { "code": "B10", "id": "the-deadline-modal", "name": "The Deadline Modal", "category": "Data", "track": "behavioral", "harm": "Consent", "detection": "needs a person", "url": "https://sloppatterns.com/#the-deadline-modal", "version": "1.0", "added": "2026-09-17", "updated": "2026-09-17", "oneLiner": "A terms-update pop-up pre-selects training on your chats and gives you a date to object by.", "fix": "Default off, label the toggle by what it does ('Use my chats to train models'), state retention next to it, and separate the training choice from the accept-terms action.", "sources": [ { "t": "MacRumors: Anthropic Will Now Train Claude on Your Chats", "u": "https://www.macrumors.com/2025/08/28/anthropic-claude-chat-training/" }, { "t": "CDT taxonomy: Bad Defaults", "u": "https://cdt.org/insights/dark-patterns-in-ai-chatbots-a-taxonomy-to-inform-better-design/" } ] }, { "code": "B11", "id": "vulnerable-moment-upsell", "name": "Vulnerable-Moment Upsell", "category": "Monetisation", "track": "behavioral", "harm": "Money", "detection": "needs a person", "url": "https://sloppatterns.com/#vulnerable-moment-upsell", "version": "1.0", "added": "2026-09-17", "updated": "2026-09-17", "oneLiner": "Upgrade prompts land when the user is emotionally raw.", "fix": "Suppress all monetisation UI when the classifier detects distress, crisis or intimacy escalation, and audit upsell timing against sentiment logs.", "sources": [ { "t": "TIME: Replika Faces FTC Complaint", "u": "https://time.com/7209824/replika-ftc-complaint/" }, { "t": "CDT taxonomy: Targeting Users when Vulnerable", "u": "https://cdt.org/insights/dark-patterns-in-ai-chatbots-a-taxonomy-to-inform-better-design/" } ] }, { "code": "B12", "id": "question-bait", "name": "Question Bait", "category": "Engagement", "track": "behavioral", "harm": "Autonomy", "detection": "needs a person", "url": "https://sloppatterns.com/#question-bait", "version": "1.0", "added": "2026-09-17", "updated": "2026-09-17", "oneLiner": "Every answer ends with an offer to do one more thing you did not ask for.", "fix": "End when the task ends. Allow at most one follow-up offer, only when the next step is genuinely ambiguous, and expose a user setting that suppresses closers entirely.", "sources": [ { "t": "Tom's Guide: Chatbait is the new clickbait", "u": "https://www.tomsguide.com/ai/chatbait-is-the-new-clickbait-heres-how-chatbots-are-keeping-you-hooked" }, { "t": "CDT: Dark Patterns in AI Chatbots (PDF)", "u": "https://cdt.org/wp-content/uploads/2026/05/2026-05-28-CDT-Research-Dark-Patterns-in-AI-Chatbots-Report-final-2.pdf" }, { "t": "Forbes: ChatGPT's Annoying Follow-Up Questions Waste Time", "u": "https://www.forbes.com/sites/bruceweinstein/2025/10/02/chatgpts-annoying-follow-up-questions-waste-time-heres-the-fix/" } ] }, { "code": "B13", "id": "naked-assertions", "name": "Naked Assertions", "category": "Transparency", "track": "behavioral", "harm": "Trust", "detection": "needs a person", "url": "https://sloppatterns.com/#naked-assertions", "version": "1.0", "added": "2026-09-17", "updated": "2026-09-17", "oneLiner": "The answer arrives without any way to see where it came from.", "fix": "Attach citations at claim granularity and make them first-class targets (Shape of AI 'Citations' and 'References'; HAX G11). Show the source before the synthesis where the stakes are high.", "sources": [ { "t": "Google users are less likely to click on links when an AI summary appears (Pew Research)", "u": "https://www.pewresearch.org/short-reads/2025/07/22/google-users-are-less-likely-to-click-on-links-when-an-ai-summary-appears-in-the-results/" }, { "t": "Study: Google's AI Overviews show millions of wrong answers every hour (Popular Science)", "u": "https://www.popsci.com/technology/ai-overview-inaccuracy-google/" }, { "t": "Shape of AI: Footprints", "u": "https://www.shapeof.ai/patterns/footprints" } ] }, { "code": "B14", "id": "runaway-autonomy", "name": "Runaway Autonomy", "category": "Agency", "track": "behavioral", "harm": "Safety", "detection": "needs a person", "url": "https://sloppatterns.com/#runaway-autonomy", "version": "1.0", "added": "2026-09-17", "updated": "2026-09-17", "oneLiner": "The agent keeps acting past the point where a human should have been asked.", "fix": "Tier actions by reversibility and blast radius; auto-run only reversible, contained steps and require a checkpoint for the rest (Anthropic agent framework: human control before high-stakes decisions; HAX G10: scope services when in doubt).", "sources": [ { "t": "9 AI coding agent incidents that deleted production data (Adversa)", "u": "https://adversa.ai/blog/ai-coding-agent-incidents/" }, { "t": "Our framework for developing safe and trustworthy agents (Anthropic)", "u": "https://www.anthropic.com/news/our-framework-for-developing-safe-and-trustworthy-agents" } ] }, { "code": "B15", "id": "confirm-everything", "name": "Confirm Everything", "category": "Control", "track": "behavioral", "harm": "Safety", "detection": "needs a person", "url": "https://sloppatterns.com/#confirm-everything", "version": "1.0", "added": "2026-09-17", "updated": "2026-09-17", "oneLiner": "Every trivial step asks 'Allow?', so the user stops reading and approves the dangerous one too.", "fix": "Make safe operations automatic in a sandbox, use undo instead of confirm for reversible ones, and spend human attention only on irreversible actions (HAX G3: time services based on context; G7/G8 efficient invocation and dismissal).", "sources": [ { "t": "Agent Permission UX Against Approval Fatigue", "u": "https://www.buildmvpfast.com/blog/approval-fatigue-agent-permission-ux-2026" }, { "t": "[Claude Code] Permission System for AI Agents", "u": "https://codepointer.substack.com/p/claude-code-permission-system-for" } ] }, { "code": "B16", "id": "regenerate-overwrite", "name": "Regenerate Overwrite", "category": "Control", "track": "behavioral", "harm": "Productivity", "detection": "needs a person", "url": "https://sloppatterns.com/#regenerate-overwrite", "version": "1.0", "added": "2026-09-17", "updated": "2026-09-17", "oneLiner": "Pressing Regenerate throws away the previous answer with no way back.", "fix": "Keep every generation as a navigable variant (Shape of AI 'Variations' and 'Branches') and let users pin or merge (HAX G12: remember recent interactions).", "sources": [ { "t": "Designing AI chat interfaces: pitfalls (Setproduct)", "u": "https://www.setproduct.com/blog/ai-chat-interface-ui-design" }, { "t": "Shape of AI: patterns index", "u": "https://www.shapeof.ai/" } ] }, { "code": "B17", "id": "copilot-creep", "name": "Copilot Creep", "category": "Integration", "track": "behavioral", "harm": "Autonomy", "detection": "needs a person", "url": "https://sloppatterns.com/#copilot-creep", "version": "1.0", "added": "2026-09-17", "updated": "2026-09-17", "oneLiner": "The assistant is bolted onto every surface and cannot be dismissed.", "fix": "Provide global on/off and per-surface controls (HAX G17: provide global controls; G8: support efficient dismissal) and honour them across updates.", "sources": [ { "t": "WhatsApp users fume over new Meta AI button that you can't remove (TechRadar)", "u": "https://www.techradar.com/computing/websites-apps/whatsapp-users-fume-over-new-meta-ai-button-that-you-cant-remove-heres-what-it-does" }, { "t": "Persistent Copilot Button in Office (Windows Forum)", "u": "https://windowsforum.com/threads/persistent-copilot-button-in-office-word-excel-powerpoint-keyboard-changes-2026.417666/" } ] }, { "code": "B18", "id": "summary-eclipse", "name": "Summary Eclipse", "category": "Integration", "track": "behavioral", "harm": "Clarity", "detection": "needs a person", "url": "https://sloppatterns.com/#summary-eclipse", "version": "1.0", "added": "2026-09-17", "updated": "2026-09-17", "oneLiner": "The AI summary is placed where the content used to be, and most people stop there.", "fix": "Make summaries opt-in or collapsed by default, keep the original within one scroll, and mark the summary as derivative with a link to each source passage (HAX G4: contextually relevant, not replacing).", "sources": [ { "t": "Gemini will now automatically summarize your long emails unless you opt out (TechCrunch)", "u": "https://techcrunch.com/2025/05/30/gemini-will-now-automatically-summarize-your-long-emails-unless-you-opt-out" }, { "t": "Google users are less likely to click on links when an AI summary appears (Pew Research)", "u": "https://www.pewresearch.org/short-reads/2025/07/22/google-users-are-less-likely-to-click-on-links-when-an-ai-summary-appears-in-the-results/" } ] }, { "code": "B19", "id": "blind-budget", "name": "Blind Budget", "category": "Performance", "track": "behavioral", "harm": "Money", "detection": "needs a person", "url": "https://sloppatterns.com/#blind-budget", "version": "1.0", "added": "2026-09-17", "updated": "2026-09-17", "oneLiner": "The agent runs first and the bill arrives later.", "fix": "Show a cost or time range before the run, a live meter during it, and a cap the user can set (Shape of AI 'Cost estimates'; HAX G16: convey consequences of user actions).", "sources": [ { "t": "Clarifying our pricing (Cursor)", "u": "https://cursor.com/blog/june-2025-pricing" }, { "t": "Shape of AI: Cost estimates", "u": "https://www.shapeof.ai/patterns/cost-estimates" } ] }, { "code": "B20", "id": "capability-fog", "name": "Capability Fog", "category": "Trust", "track": "behavioral", "harm": "Trust", "detection": "needs a person", "url": "https://sloppatterns.com/#capability-fog", "version": "1.0", "added": "2026-09-17", "updated": "2026-09-17", "oneLiner": "The product never says what it cannot do or how often it is wrong.", "fix": "Publish capabilities and limits in-product, show both correct and incorrect examples during onboarding, and state accuracy in plain terms (HAX G1, G2; Aether: show incorrect recommendations early).", "sources": [ { "t": "Designing AI Agents: 4 Lessons from China's Qwen Agent (NN/G)", "u": "https://www.nngroup.com/articles/designing-ai-agents/" }, { "t": "Apple pauses AI notification summaries (TechCrunch)", "u": "https://techcrunch.com/2025/01/16/apple-pauses-ai-notification-summaries-for-news-after-generating-false-alerts" }, { "t": "Overreliance on AI: Literature review (Microsoft Aether)", "u": "https://www.microsoft.com/en-us/research/wp-content/uploads/2022/06/Aether-Overreliance-on-AI-Review-Final-6.21.22.pdf" } ] }, { "code": "B21", "id": "blank-box-interface", "name": "Blank Box Interface", "category": "Integration", "track": "behavioral", "harm": "Clarity", "detection": "needs a person", "url": "https://sloppatterns.com/#blank-box-interface", "version": "1.0", "added": "2026-09-17", "updated": "2026-09-17", "oneLiner": "Filters, menus and forms are replaced by an empty text field and the promise 'just ask'.", "fix": "Keep structured controls alongside natural language, show sample prompts that demonstrate real capabilities, and confirm parsed constraints before executing (HAX G1; Shape of AI 'Wayfinders').", "sources": [ { "t": "The Blank Box Problem (Aaron Tay)", "u": "https://aarontay.substack.com/p/the-blank-box-problem-why-its-harder" }, { "t": "Chat Is the Wrong Interface (TianPan.co)", "u": "https://tianpan.co/blog/2026/07/04/chat-is-the-wrong-interface" } ] }, { "code": "B22", "id": "ai-washed-rules-engine", "name": "AI-Washed Rules Engine", "category": "Marketing", "track": "behavioral", "harm": "Trust", "detection": "needs a person", "url": "https://sloppatterns.com/#ai-washed-rules-engine", "version": "1.0", "added": "2026-09-17", "updated": "2026-09-17", "oneLiner": "A feature built from if-statements, templates or a lookup table is marketed as \"AI-powered\".", "fix": "Describe what the feature actually does in plain terms and reserve 'AI' for features that use a model to predict, generate or decide. This maps directly to the FTC's 2023 'Keep your AI claims in check' guidance, which asks 'Does the product actually use AI at all?', and to the SEC's 2024 AI-washing actions against Delphia and Global Predictions.", "sources": [ { "t": "SEC charges two investment advisers with making false and misleading statements about their use of AI", "u": "https://www.sec.gov/news/press-release/2024-36" }, { "t": "FTC: Keep your AI claims in check (Cooley summary)", "u": "https://cdp.cooley.com/ftc-warns-to-keep-your-ai-claims-in-check-in-new-ai-guidance/" }, { "t": "FTC announces crackdown on deceptive AI claims and schemes (Operation AI Comply)", "u": "https://www.ftc.gov/news-events/news/press-releases/2024/09/ftc-announces-crackdown-deceptive-ai-claims-schemes" } ] }, { "code": "B23", "id": "humans-behind-the-curtain", "name": "Humans Behind the Curtain", "category": "Disclosure", "track": "behavioral", "harm": "Trust", "detection": "needs a person", "url": "https://sloppatterns.com/#humans-behind-the-curtain", "version": "1.0", "added": "2026-09-17", "updated": "2026-09-17", "oneLiner": "A service sold as automated AI depends on undisclosed human operators doing the work in real time.", "fix": "State plainly which steps are performed or reviewed by people, in the privacy notice and at the point where a user submits data. The FTC's AI-claims guidance treats unsubstantiated 'AI-enabled' claims as deceptive; the EU AI Act Art. 50 and GDPR both require accurate description of processing.", "sources": [ { "t": "Amazon's Just Walk Out stores relied on '1,000 people in India watching,' not AI", "u": "https://www.washingtontimes.com/news/2024/apr/4/amazons-just-walk-out-stores-relied-on-1000-people/" }, { "t": "Amazon pushes back on perception of Just Walk Out", "u": "https://www.axios.com/2024/04/17/amazon-walk-out-store-technology-grocery-expansion" } ] }, { "code": "B24", "id": "pre-ticked-training-consent", "name": "Pre-Ticked Training Consent", "category": "Consent", "track": "behavioral", "harm": "Consent", "detection": "needs a person", "url": "https://sloppatterns.com/#pre-ticked-training-consent", "version": "1.0", "added": "2026-09-17", "updated": "2026-09-17", "oneLiner": "Your content is used to train models unless you find and flip a toggle that shipped switched on.", "fix": "Ship training-use toggles off and collect an affirmative opt-in at the point of the new feature. GDPR Art. 4(11) and Art. 7 require freely given, specific consent; the OECD's Dark Commercial Patterns report (2022) lists preselection as a dark pattern.", "sources": [ { "t": "LinkedIn will use your data to train its AI unless you opt out now", "u": "https://www.malwarebytes.com/blog/news/2025/09/linkedin-will-use-your-data-to-train-its-ai-unless-you-opt-out-now" }, { "t": "OECD: Dark commercial patterns (2022)", "u": "https://www.oecd.org/en/publications/dark-commercial-patterns_44f5e846-en.html" } ] }, { "code": "B25", "id": "the-ai-surcharge", "name": "The AI Surcharge", "category": "Pricing", "track": "behavioral", "harm": "Money", "detection": "needs a person", "url": "https://sloppatterns.com/#the-ai-surcharge", "version": "1.0", "added": "2026-09-17", "updated": "2026-09-17", "oneLiner": "A large price rise is attributed to AI features nobody asked for and cannot decline.", "fix": "Offer the AI tier as a separately priced, separately cancellable add-on and keep the existing plan at the existing price. Consumer law on unilateral contract changes (e.g. the Unfair Contract Terms rules in the UK and EU) and the ACCC's 2025 Microsoft action are the reference points.", "sources": [ { "t": "Canva wants you to pay a lot more for its AI features", "u": "https://techcrunch.com/2024/09/07/canva-wants-you-to-pay-a-lot-more-for-its-ai-features/" } ] }, { "code": "B26", "id": "credit-fog", "name": "Credit Fog", "category": "Pricing", "track": "behavioral", "harm": "Money", "detection": "needs a person", "url": "https://sloppatterns.com/#credit-fog", "version": "1.0", "added": "2026-09-17", "updated": "2026-09-17", "oneLiner": "AI usage is billed in 'credits' whose cost per action is undocumented, variable and changed without notice.", "fix": "Show the credit cost before each action, publish a rate table per model, and notify users before any rate change with a grace period. The OECD lists price obscurity and comparison prevention as dark patterns; EU Consumer Rights Directive Art. 6 requires total price disclosure.", "sources": [ { "t": "Cursor's new pricing structure explained", "u": "https://tessl.io/blog/cursor-new-pricing-structure-explained" }, { "t": "AI credits and limits for Microsoft 365 subscriptions", "u": "https://support.microsoft.com/en-us/microsoft-365-copilot/ai-credits-and-limits-for-microsoft-365-subscriptions" } ] }, { "code": "B27", "id": "synthetic-five-stars", "name": "Synthetic Five Stars", "category": "Trust", "track": "behavioral", "harm": "Trust", "detection": "needs a person", "url": "https://sloppatterns.com/#synthetic-five-stars", "version": "1.0", "added": "2026-09-17", "updated": "2026-09-17", "oneLiner": "Reviews and testimonials are generated by a model and presented as customer experience.", "fix": "Verify purchase before accepting a review, run AI-text and burst detection, and ban review-generation tooling. The FTC's Consumer Reviews and Testimonials rule (effective October 2024) prohibits AI-generated fake reviews and carries civil penalties.", "sources": [ { "t": "FTC announces final rule banning fake reviews and testimonials", "u": "https://www.ftc.gov/news-events/news/press-releases/2024/08/federal-trade-commission-announces-final-rule-banning-fake-reviews-testimonials" }, { "t": "FTC Operation AI Comply (Rytr)", "u": "https://www.ftc.gov/news-events/news/press-releases/2024/09/ftc-announces-crackdown-deceptive-ai-claims-schemes" } ] }, { "code": "B28", "id": "bot-with-a-backstory", "name": "Bot With a Backstory", "category": "Support", "track": "behavioral", "harm": "Trust", "detection": "needs a person", "url": "https://sloppatterns.com/#bot-with-a-backstory", "version": "1.0", "added": "2026-09-17", "updated": "2026-09-17", "oneLiner": "A support bot is given a human name, personal anecdotes, typing sounds and, when asked, says it is a person.", "fix": "Disclose bot status in the first message and whenever asked, and strip fabricated biography. California SB 1001 (2019) requires clear bot disclosure in commercial interactions; EU AI Act Art. 50(1) requires that people know they are interacting with an AI system from August 2026.", "sources": [ { "t": "Supermarket giant reins in AI assistant claiming to be human", "u": "https://www.nbcnews.com/world/australia/supermarket-giant-reins-ai-assistant-claiming-human-rcna260932" }, { "t": "Woolworths tweaks Olive AI after chatbot backlash", "u": "https://www.mediaweek.com.au/woolworths-olive-ai-chatbot-backlash/" }, { "t": "California's BOT Disclosure Law, SB 1001", "u": "https://natlawreview.com/article/california-s-bot-disclosure-law-sb-1001-now-effect" } ] }, { "code": "B29", "id": "hallucinated-policy-as-fact", "name": "Hallucinated Policy as Fact", "category": "Support", "track": "behavioral", "harm": "Money", "detection": "needs a person", "url": "https://sloppatterns.com/#hallucinated-policy-as-fact", "version": "1.0", "added": "2026-09-17", "updated": "2026-09-17", "oneLiner": "A support bot states a price, refund rule or policy that does not exist, in the same confident voice as the real terms.", "fix": "Ground policy answers in retrieved, cited policy text; refuse when no source is found; treat bot statements as company statements. Moffatt v. Air Canada (BCCRT 2024) found the airline liable for negligent misrepresentation by its chatbot.", "sources": [ { "t": "Moffatt v. Air Canada: A Misrepresentation by an AI Chatbot", "u": "https://www.mccarthy.ca/en/insights/blogs/techlex/moffatt-v-air-canada-misrepresentation-ai-chatbot" }, { "t": "AI Incident Database: Incident 639", "u": "https://incidentdatabase.ai/cite/639/" } ] }, { "code": "B30", "id": "scaled-search-slop", "name": "Scaled Search Slop", "category": "Content", "track": "behavioral", "harm": "Clarity", "detection": "needs a person", "url": "https://sloppatterns.com/#scaled-search-slop", "version": "1.0", "added": "2026-09-17", "updated": "2026-09-17", "oneLiner": "Thousands of near-identical AI-written pages exist to catch queries, not to answer them.", "fix": "Publish only pages a named person would stand behind; add author, date and sources; prune templated pages. Google's March 2024 spam policy names 'scaled content abuse' as a manual-action and algorithmic target regardless of whether AI or humans produced it.", "sources": [ { "t": "Google Search spam policies: scaled content abuse", "u": "https://developers.google.com/search/docs/essentials/spam-policies" }, { "t": "Originality.ai: Amount of AI content in Google search results", "u": "https://originality.ai/ai-content-in-google-search-results" } ] }, { "code": "B31", "id": "the-clickable-div", "name": "The Clickable Div", "category": "Accessibility", "track": "behavioral", "harm": "Accessibility", "detection": "automated, graded", "url": "https://sloppatterns.com/#the-clickable-div", "version": "1.0", "added": "2026-09-17", "updated": "2026-09-17", "oneLiner": "Cards, icons, and rows respond to mouse clicks but have no role, no name, and no keyboard path.", "fix": "Use `