# Userium: Usability Checklist for the AI Age > A free, copyable prompt-based usability checklist for auditing websites. Originally created in 2013, resurrected for 2026. Website: https://www.userium.com Creator: Nina Patkai, M.Sc. Eng. Company: EVcourse (https://www.evcourse.com) License: CC BY-SA 4.0 The checklist contains 56 items across 10 categories. Each item includes a title, description, and a detailed prompt you can paste into an AI tool to audit any website against proven usability principles. --- ## 1. User Experience (7 items) ### Personalized experience. Currency, language, and region-specific content adapt to the user's context without requiring manual setup. Prompt: Audit this website for localization and personalization. Does it detect or let users set their language, currency, and region? Are delivery options, tax displays, and legal disclaimers appropriate for each market? List every gap. ### Registration provides clear value. Users understand what they gain by creating an account. Unnecessary registration walls are eliminated. Prompt: Review every point where this site asks users to register or log in. For each, does the user understand the benefit before committing? Is there content or functionality gated behind a registration wall that shouldn't be? Suggest which gates to remove. ### No dark patterns. The interface does not use deceptive design to trick users into unintended actions, subscriptions, or data sharing. Prompt: Scan this website for dark patterns as defined by the FTC and EU Digital Services Act: confirmshaming, hidden costs, forced continuity, misdirection, trick questions, disguised ads, roach motels, privacy zuckering, bait-and-switch, or fabricated urgency. List every instance with its location and which dark pattern type it represents. ### Transparent pricing. All costs, fees, and terms are clearly displayed before any commitment point. No hidden charges appear only at checkout. Prompt: Follow the complete purchase or sign-up flow on this website. Are all fees (shipping, taxes, service charges, recurring costs) shown before the final confirmation? Are subscription terms (billing frequency, cancellation, auto-renewal) clearly stated? Does the total change between any two steps? Flag every instance where the total cost is unclear. ### Consent is meaningful. Cookie banners, newsletter signups, and permission requests explain what the user is agreeing to and make declining as easy as accepting. Prompt: Review all consent mechanisms on this site: cookie banners, newsletter popups, notification requests, data sharing agreements. For each, is declining as easy as accepting? Is the language clear and non-manipulative? Does the site respect the user's choice? Are there pre-checked boxes or confusing double negatives? List every violation. ### Undo is always available. Users can reverse actions easily. Destructive actions require confirmation. There is always a clearly marked emergency exit. Prompt: Map every multi-step flow, modal, overlay, and state change on this website. For each, verify there is a clearly visible way to undo, cancel, go back, or dismiss. Test every destructive action (delete, unsubscribe, purchase). Are destructive actions confirmed before execution? Flag any flow where the user could feel trapped. ### System feedback is immediate. The interface keeps users informed about what is happening through timely, visible feedback for every action. Prompt: Analyze this website and identify every user action (clicks, form submissions, loading states, transitions) that lacks immediate visible feedback. For each, recommend a specific feedback mechanism (spinner, progress bar, confirmation message, animation) and note where the user might feel uncertain about whether the system received their input. --- ## 2. Homepage (5 items) ### Clear value proposition. A first-time visitor understands what the site offers, who it is for, and what to do next within 5 seconds. Prompt: View only the above-the-fold content of this homepage. Can you determine in under 5 seconds: what this site does, who it is for, and what the primary call to action is? If any of these are unclear, explain exactly what is missing and suggest specific copy and layout changes. ### Strong first impression. The homepage creates trust and professionalism. Visual hierarchy guides the eye from the most important element to the least. Prompt: Analyze the visual hierarchy of this homepage. Identify the intended primary, secondary, and tertiary focal points. Are they correctly ordered by visual weight (size, contrast, color, position)? Flag any competing elements that create visual noise, any CTA that is visually subordinate to less important content, and any section where the eye has no clear path. ### Contact information is visible. Company location, contact methods, and support options are easily accessible without hunting. Prompt: How easy is it to find contact information on this website? Can users locate email, phone, address, or chat support within two clicks from the homepage? Is there a clear contact page? Does the footer include basic contact details? List what's missing. ### Clear call to action. The primary action stands out visually and uses benefit-driven language instead of generic labels like "Submit". Prompt: Identify every call-to-action on this homepage. For each: is it visually prominent? Does the label communicate a benefit (e.g., "Start free trial" vs "Submit")? Is there a clear primary CTA that dominates? Are there too many competing CTAs diluting focus? Recommend improvements. ### Above-the-fold is meaningful. The most critical content and primary action appear without scrolling on both desktop (900px) and mobile (667px). Prompt: Capture the above-the-fold view of this homepage on desktop (1440x900) and mobile (375x667). Does the visible area contain: the site identity, a clear headline communicating the value proposition, and the primary call to action? Flag anything critical pushed below the fold or anything non-essential displacing important content. --- ## 3. Navigation & Information Architecture (6 items) ### Navigation is consistent. The main navigation appears in the same position on every page with consistent labels and behavior. Prompt: Visit at least 5 different pages on this website, including the homepage, a deep content page, and any transactional page. Verify that primary navigation: appears in the same position, uses identical labels, indicates the current location, and is reachable within one interaction on mobile. Document any page where navigation is missing, inconsistent, or requires more than one tap to access. ### Search works well. Search is easy to find, handles typos gracefully, shows autocomplete suggestions, and returns useful results. Prompt: Test the search functionality with: an exact term from the site, a misspelled version, a synonym, and a long-tail natural language query. Does search handle misspellings? Are results relevant and well-formatted? Does autocomplete aid discovery? Do zero-result pages offer helpful next steps instead of dead ends? Grade each aspect. ### Mobile navigation is usable. On small screens, navigation is reachable with one thumb, clearly labeled, and does not require excessive scrolling. Prompt: Test all navigation elements at a 375px viewport. Is the menu trigger clearly recognizable and tappable (minimum 48x48px)? Does the expanded menu cover the full viewport without layout conflicts? Are dropdown/sub-navigation levels accessible via tap (not hover)? Is there a visible way to close the menu? Could a bottom navigation bar improve thumb reachability? ### Content is findable. Users can reach any piece of content within 3 clicks, supported by breadcrumbs and clear wayfinding cues. Prompt: Select the 5 most important user tasks on this website. For each, count the minimum number of clicks to reach the goal from the homepage. Flag any task requiring more than 3 interactions. Check for: breadcrumbs on interior pages, related content links, clear signposting of where each link leads, and whether users can always tell where they are in the site structure. ### AI assistant is helpful, not intrusive. If the site has a chatbot or AI assistant, it offers genuine help without blocking content or auto-opening. Prompt: If this website has a chatbot or AI assistant: does it auto-open and block content? Can users dismiss it permanently? Does it provide accurate, helpful responses? Does it clearly identify itself as AI? Can users easily reach a human? Does it respect the conversation history? Can users delete their chat history? Evaluate each. ### Links are descriptive. Link text describes the destination. "Click here" and "Read more" without context are avoided. Critical for screen readers. Prompt: Scan all links on this website. Flag any with generic text ("click here", "read more", "learn more") that does not describe its destination. Check that all links: are visually distinguishable from non-link text, have descriptive text that makes sense when read in isolation (for screen readers), and indicate if they open external sites or downloads. --- ## 4. Accessibility (7 items) ### Keyboard navigation works. Every interactive element is reachable and operable with keyboard alone. Focus order is logical. No keyboard traps. Prompt: Navigate this entire website using only the keyboard (Tab, Shift+Tab, Enter, Space, Escape, arrow keys). Document: any interactive element that cannot receive focus, focus order that does not match visual order, modals or dropdowns that cannot be dismissed with Escape, missing skip-to-content link. List every failure. ### Color contrast is sufficient. Text meets WCAG AA contrast ratios: 4.5:1 for body text, 3:1 for large text and UI components. Prompt: Audit every text element and UI component for WCAG 2.2 AA color contrast compliance. Check body text (4.5:1), headings (3:1 if large), placeholder text in form fields, text over images or gradients at their lowest-contrast point, button text, link text. List every failing element with its current ratio and the minimum required. ### Images have alt text. All meaningful images have descriptive alt text that conveys purpose, not just appearance. Decorative images use alt="". Prompt: Audit every image on this website. Does each meaningful image have alt text describing its purpose/function rather than just appearance? Are decorative images marked with alt="" (not omitted entirely)? Are complex images (charts, infographics) accompanied by detailed text alternatives? Are CSS background images that convey meaning given accessible text? Flag AI-generated images that lack meaningful alt text. ### Focus indicators are visible. When tabbing through the page, every focused element has a clearly visible outline meeting WCAG 2.2 Focus Appearance requirements. Prompt: Tab through this entire website and evaluate focus indicators. Is every interactive element's focus state clearly visible? Does the focus indicator have sufficient contrast against the background? Does it meet WCAG 2.2 Focus Appearance requirements (2px outline, 3:1 contrast)? List every failure. ### Motion respects preferences. Animations are reduced or eliminated when the user has enabled "prefers-reduced-motion" in their OS settings. Prompt: Identify every animation, transition, auto-playing video, parallax effect, and moving element on this website. For each, verify it respects the prefers-reduced-motion media query. Check for auto-playing media without pause/stop controls. Ensure no element flashes more than 3 times per second. Check the CSS for a @media (prefers-reduced-motion: reduce) rule. ### Headings form a logical outline. The page uses a single H1, and heading levels (H2-H6) are nested in order without skipping levels. Prompt: Extract the heading structure (H1-H6) of every page on this website and present it as an outline. Is there exactly one H1 per page? Are heading levels sequential (no jumps from H2 to H4)? Do headings accurately summarize the content that follows? Could a user understand the page by reading only the headings? Flag every violation. ### Screen reader experience is complete. All form inputs have labels, error messages are announced, dynamic content uses ARIA live regions, and the page reads in logical order. Prompt: Simulate a screen reader experience. Does the page read in a logical order? Does each form input have a properly associated label? Do buttons and links announce meaningful text? Are dynamic content updates announced via aria-live regions? Are icon-only buttons given accessible names? Are decorative elements hidden with aria-hidden="true"? List every issue. --- ## 5. Forms & Input (6 items) ### Error messages are helpful. Validation errors appear inline next to the field, explain what went wrong in plain language, and suggest how to fix it. Prompt: Deliberately trigger every possible validation error on this website's forms. For each error: does it appear next to the relevant field? Does it explain what went wrong specifically ("Email must include @" not "Invalid input")? Does it use both color AND text/icon (not color alone)? Does focus move to the first error? Are errors announced to screen readers? List every poor error message. ### Forms support autofill. Input fields use correct autocomplete attributes and input types so browsers and password managers work correctly. Prompt: Inspect every form input on this website. Do name, email, phone, address, and payment fields have the correct HTML autocomplete attributes? Is the correct input type used (type="email", type="tel") to trigger appropriate mobile keyboards? Does the browser autofill work correctly? Are there any fields that block autofill with autocomplete="off" when they shouldn't? List every issue. ### Required fields are marked. Users can tell which fields are required before they start filling out the form. Prompt: Review every form on this website. Are required fields clearly marked (with an asterisk, "required" label, or similar)? Is the marking convention explained? Are optional fields marked instead if most fields are required? Can users identify required fields before interacting with the form? List issues. ### Input format is flexible. Phone numbers, dates, and credit cards accept multiple formats. The system adapts to the user, not the other way around. Prompt: Test every form input that expects formatted data (phone, date, credit card, postal code). Try entering the data in different valid formats. Does the form accept variations? Does it strip or format automatically? Are there format hints in the placeholder or help text? Or does it force a rigid format that the user must guess? List every inflexible field. ### Progress is preserved. If users navigate away from a partially completed form, their input is still there when they return. Prompt: Partially fill out every multi-field form on this website, navigate away, then return. Is the form data still present? For multi-step forms, can users go back to previous steps without losing data? Does the form warn users before they lose unsaved input? After a failed submission, is all previously valid input preserved? List every form that loses user data. ### Long forms are broken into steps. Forms with more than 6-8 fields are split into logical steps with a progress indicator. Prompt: Identify every form with more than 6 fields. Is each broken into logical steps with a progress indicator? Are conditionally relevant fields hidden until triggered by a prior answer? Can users review their entries before final submission? Does each step have a clear heading explaining what information is needed and why? List every form that should be multi-step but isn't. --- ## 6. Content & Readability (5 items) ### Text is scannable. Content uses short paragraphs, subheadings, bullet points, and bold keywords so users can scan before reading. Prompt: Evaluate the scannability of every content page. Are paragraphs short (3-4 sentences max)? Are there subheadings every 2-3 paragraphs? Are key terms bolded? Are lists used where appropriate? Is line height at least 1.5x font size? Is content width between 45-75 characters per line? Can a user grasp the main points by scanning for 10 seconds? Grade each page. ### Reading level is appropriate. Body text targets an 8th-grade reading level unless the audience is specifically technical. Prompt: Analyze the body copy on this website's main pages. Flag: sentences over 25 words, paragraphs over 4 lines, jargon without explanation, passive voice, double negatives, and ambiguous pronouns. Rewrite the 3 worst-offending passages at an 8th-grade reading level using short sentences and active voice. ### AI content is disclosed. Any content generated or curated by AI is labeled. Users can flag inaccuracies. Sources are cited. Prompt: Examine this website for AI-generated or AI-assisted content (text, images, recommendations, chatbot responses). For each instance: is AI involvement disclosed? Are sources cited? Is there a way for users to report inaccurate AI content? Are AI-generated images labeled as such? Does AI content include a caveat where appropriate? Is human editorial oversight evident? ### Microcopy is thoughtful. Button labels, tooltips, placeholders, and empty states communicate clearly and reduce friction. Prompt: Audit all microcopy: button labels, form placeholders, tooltips, empty states, loading messages, success/error confirmations. Flag any instance that is generic ("Submit", "Error", "OK"), written in system language rather than user language, unhelpful in an empty state, or tonally inconsistent with the brand. Provide specific rewrites for each. ### Content is up to date. No stale dates, broken links, deprecated features, or references to discontinued products. Prompt: Crawl this website and identify: pages missing publication or last-updated dates, all broken links (404s, wrong redirects), references to past events or expired promotions, outdated statistics, copyright year in the footer (is it current?), and any content referencing deprecated technologies. Provide a staleness risk score for each page. --- ## 7. Performance & Technical (6 items) ### Pages load fast. Largest Contentful Paint under 2.5s, Interaction to Next Paint under 200ms, Cumulative Layout Shift under 0.1. Prompt: Run a Core Web Vitals audit on the 5 most important pages. For each page, report: LCP (target < 2.5s), INP (target < 200ms), CLS (target < 0.1), FCP, TTFB, and total page weight. Identify the LCP element on each page and recommend the single highest-impact optimization per page. ### Mobile experience is native-feeling. The site works flawlessly on mobile: no horizontal scroll, no tiny tap targets, no desktop-only features. Prompt: Test this website on at least 3 mobile viewport sizes (320px, 375px, 428px). Is there any horizontal scrolling? Are all tap targets at least 48x48px with adequate spacing? Do all features work on mobile? Are images responsive? Is text readable without zooming (minimum 16px body)? List every mobile usability issue. ### Images are optimized. Images use modern formats (WebP/AVIF), responsive srcset, lazy loading below the fold, and width/height attributes to prevent layout shift. Prompt: Audit every image. Check: format (WebP or AVIF preferred), responsive delivery via srcset and sizes, lazy loading on below-fold images, explicit width and height attributes to prevent CLS, total image payload per page. Flag any image over 200KB, any missing srcset, and any above-fold image that is lazy-loaded when it should be eager. ### No unnecessary JavaScript. Third-party scripts are audited and justified. No render-blocking scripts. Bundle size is minimized. Prompt: Identify every JavaScript file loaded by this website including third-party scripts. For each: what is its size? Is it render-blocking? Is it necessary? Could it be deferred, lazy-loaded, or eliminated? Calculate the total JS payload. Flag any script that blocks rendering, loads synchronously in the head, is unused on the current page, or could be replaced with a lighter alternative. ### Error states are handled. Custom 404 page with navigation. Server errors show friendly messages. Broken features degrade gracefully. Prompt: Test error handling: visit non-existent URLs. Does the 404 page offer navigation, search, and helpful suggestions? Disable JavaScript. Is critical content still accessible? Simulate a slow network. Do loading states appear? Force an API failure. Does the UI show a helpful error or break silently? Rate the overall error resilience. ### SEO and LLM discoverability are solid. Unique title tags under 60 characters, meta descriptions, canonical URLs, Open Graph tags, structured data, and llms.txt for AI crawlers. Prompt: Audit the SEO and LLM discoverability of every page. Check: unique title tag under 60 characters with primary keyword, unique meta description under 160 characters, canonical URL tag, Open Graph and Twitter Card tags, structured data (JSON-LD) for the content type, XML sitemap existence, robots.txt configuration. Also check for llms.txt and llms-full.txt files that help AI crawlers understand the site. Is there a clear, structured summary of the site for LLMs? Are AI crawlers allowed or blocked in robots.txt? Flag duplicates and missing elements. --- ## 8. Trust & Privacy (6 items) ### Privacy policy is readable. Written in plain language, organized with headings, and specific about what data is collected, why, and for how long. Prompt: Read the privacy policy on this website. Is it written in plain language a non-lawyer can understand? Does it specify: what data is collected, why, how long it's stored, who it's shared with, and how to delete it? Are third-party data sharing partners named? Are user rights (access, deletion, portability) clearly explained? Does it address AI/ML use of user data? Grade its readability. ### Cookie usage is minimal. Only essential cookies are set by default. Analytics and marketing cookies require explicit opt-in with equal accept/reject buttons. Prompt: Clear all cookies, visit this website, and inspect what cookies are set before any consent interaction. Are cookies set before the user consents? Is the cookie banner genuinely user-friendly? Are accept and reject buttons equally prominent? After declining, are only essential cookies present? Can users change their preference later? List every cookie with its purpose. ### Trust indicators are present. SSL certificate, clear company identity, contact info, professional design, and relevant certifications are visible. Prompt: Audit trust indicators. Check for: valid SSL certificate, clearly visible company name and address, "About" page with real team or company history, security badges or compliance certifications, professional error-free design (no broken images, typos, lorem ipsum). List every missing trust indicator. ### Third-party services are disclosed. Users are informed about analytics, advertising, and data-sharing with third parties before it happens. Prompt: Inspect the network requests this website makes. List every third-party service contacted (analytics, ads, fonts, CDNs, APIs). Is each one disclosed in the privacy policy? Is user consent obtained before non-essential third-party requests are made? Flag every undisclosed third-party connection. ### Data collection is proportional. The site only asks for information it genuinely needs. Forms don't collect unnecessary personal data. Prompt: Review every data collection point (forms, sign-ups, checkout, chat). For each field requesting personal data: is it genuinely necessary for the stated purpose? Is the reason for collecting each piece of data explained? Are there fields that could be eliminated (e.g., phone number for a newsletter signup)? Does the site collect data through hidden means (fingerprinting, invisible trackers)? Flag any over-collection. ### AI features are disclosed. If AI powers any part of the experience (chatbots, recommendations, content, decisions), this is transparently communicated. Prompt: Identify every AI-powered feature on this website (chatbots, recommendations, personalization, automated decisions, generated content). For each: is the user informed they are interacting with AI? Is there an option to reach a human? Are AI-generated outputs labeled? Is the AI's decision-making explained where it affects users? Does the site comply with EU AI Act transparency requirements? --- ## 9. AI Readiness (4 items) ### Site works for AI agents. Personal AI agents can navigate and complete tasks on the site without getting trapped by complex JS or non-standard UI. Prompt: Audit this website for agentic reachability. If a personal AI agent were to visit this site to perform a specific task (e.g., "buy the cheapest red toaster"), could it navigate the DOM without executing complex JS? Are buttons and form actions semantically labeled? Are key actions exposed via clear semantic HTML rather than hidden behind JavaScript event handlers? Flag any agent-traps like non-standard captchas, invisible overlays, or actions that require hover states to reveal. ### AI decisions can be appealed. When AI makes decisions that affect users (content moderation, pricing, recommendations, eligibility), there is a clear path to human review. Prompt: Identify every place where AI makes or influences decisions on this website: content moderation, dynamic pricing, loan/insurance eligibility, recommendation filtering, search ranking, fraud detection, account restrictions. For each: is the user informed that AI is involved? Can the user see why the decision was made? Is there a clear, accessible process to appeal to a human reviewer? Does the appeal process meet EU AI Act requirements for high-risk systems? ### Media transparency. Since most content is now AI-enhanced, simple "AI-written" labels are not enough. Synthetic media, accountability, and verification matter. Prompt: Audit this website's content transparency. Since most content is now AI-enhanced, do not look for simple "AI-written" labels. Instead, identify: Synthetic Realism: are any "photos" of products or people actually 100% synthetic? If so, is this disclosed to prevent fraud? Verification: are user-generated sections (reviews, comments) verified as coming from real humans, or is there a high risk of automated persona generation? C2PA Compliance: do images contain Content Credentials metadata that identify the source and edit history? List any areas where the lack of disclosure could lead to deception risk under the EU AI Act. ### Engagement respects well-being. The site does not use infinite scrolls without floors, variable reward loops, ghost notifications, or other patterns designed to hijack attention. Prompt: Audit this site for dopamine dark patterns. Does it use infinite scrolling without a clear endpoint? Are notifications used for genuine utility or to bait the user back? Does the site offer a way to consume content without algorithmically-driven recommendations? Are there "time on site" maximizing features that conflict with user intent? Does the site show session duration or usage summaries? Flag any feature designed to maximize engagement at the expense of user well-being. --- ## 10. User Rights (4 items) ### Account deletion is effortless. Deleting an account is as easy as creating one. Users can export their data and request removal from AI training sets. No guilt trips, no hidden flows. Prompt: Attempt to delete an account on this website. Is the option easy to find (within 2 clicks from account settings)? Can users download/export all their data before deletion? Is the process instant or does it require waiting, emailing support, or calling a phone number? Does the site confirm what data will be deleted, including data used for AI model training? Are there confirmshaming tactics ("Are you sure you want to lose all your progress?")? Compare the number of steps to sign up vs. delete. ### Cross-device handoff works. Users can start a task on one device and continue seamlessly on another. Cart contents, form progress, and AI conversation history persist across devices. Prompt: Start a multi-step task on this website using a mobile device (add items to cart, begin a form, start an AI chat conversation). Then switch to a desktop browser logged into the same account. Is your progress preserved? Check: cart contents, partially completed forms, AI chat history, reading position, saved preferences, and notification state. List every piece of state that fails to sync across devices. ### High-stakes actions have friction. For large purchases, permanent deletions, or public posts, the site introduces intentional slowness to prevent impulsive or AI-hallucinated actions. Prompt: Identify every high-consequence action on this site: large financial transfers, permanent data deletion, public posts, account changes, subscription commitments. Does the site provide enough positive friction to ensure the user is not acting on impulse or an AI hallucination? Are there double-confirmations, cooling-off periods, or summary screens before irreversible actions? Flag any high-stakes action that can be completed in a single click without review. ### Carbon footprint is minimized. The site practices digital sobriety. Heavy AI calls are avoided when a simple heuristic would work. Auto-playing video is limited. Page weight is kept low. Prompt: Audit this website for digital sustainability. What is the total page weight per visit? Are there unnecessary AI API calls that could be replaced with simple logic or cached responses? Is auto-playing video used without user consent? Are images served in next-gen formats (WebP/AVIF)? Are third-party scripts loaded that serve no clear user benefit? Estimate the carbon cost per page view using websitecarbon.com methodology and recommend reductions. --- ## References - Jakob Nielsen: 10 Usability Heuristics (https://www.nngroup.com/articles/ten-usability-heuristics/) - Don Norman: The Design of Everyday Things (https://www.nngroup.com/books/design-everyday-things-revised/) - Steve Krug: Don't Make Me Think (https://sensible.com/dont-make-me-think/) - W3C: WCAG 2.2 Quick Reference (https://www.w3.org/WAI/WCAG22/quickref/)