As of 13 August 2026, AI can only partly check your website design for accessibility.
This still needs a person who signs their name to it.
Can you do it?
5 minutesto a draft.
1 hourto something you’d act on.
Cost, all in£0
Skill neededpower-user
Who has to check ityou
What the alternative costsThe supplied tool list gives no price for a specialist accessibility audit.
If this goes wrong, you publish a design that appears compliant in screenshots but excludes people when they use the real interface.
What to actually do
Hand it to a person
The route this page recommends
A person who owns the outcome does this end to end, worth it when the failure is dear.
Use a tool built for this
Second choiceDo it yourself
The distant thirdA chat interface, power-user skill, and roughly 1 hour until you can act on the result.
How to actually do it
- Open the latest design in Figma or your design tool and export the key desktop, tablet and mobile screens, including menus, forms, errors, modals, focus states and empty states.
- Gather the current component notes, colour tokens, typography settings, content rules and any available HTML or CSS for the screens being reviewed.
- Write a short description of the website purpose, target users, important journeys and known accessibility needs, then paste it into the prompt with the exported designs and code.
- Run the prompt and copy each definite finding, likely risk and proposed manual test into the accessibility issues list for the project.
- Compare every colour, font size, heading, label, button, link, image and focus-state finding against the source design and mark incorrect assumptions for correction.
- Open the working site in a browser and test the listed journeys with only the keyboard, browser zoom, a reduced-motion setting and a screen reader, recording results against the AI checklist.
- Run an automated accessibility scan on the working pages, investigate any differences from the AI review, fix the confirmed issues and send unresolved interaction or user-experience questions to an accessibility specialist or representative users.
Prompt
Act as an accessibility reviewer for this website design. Use the supplied screenshots, HTML, CSS, component notes and interaction descriptions, but do not claim to have tested anything I have not supplied. Assess the design against relevant WCAG 2.2 principles, including keyboard access, focus order and visibility, headings and landmarks, colour contrast, text resizing and reflow, target size, forms and error messages, links and buttons, images and alternative text, motion, responsive states and content clarity. For each finding, give: severity as a practical priority, the exact element or state affected, the accessibility risk, the relevant WCAG success criterion if you are confident, a suggested fix, and what must be tested manually. Separate definite findings from questions or likely risks. Do not declare the design compliant. End with a test plan covering keyboard-only use, zoom or text resizing, reduced motion, a screen reader, mobile layouts and at least one automated scan. Website purpose: [describe the service and audience]. Target users and known access needs: [describe them]. Design or code to review: [paste or attach it].
Open it prefilled in ChatGPT or Claude, or copy it into Gemini, which takes no prefill link.
What it gets wrong
- AI cannot experience the finished interface through every relevant combination of browser, device, keyboard and assistive technology from a static design alone.
- AI cannot know whether the reading order, language, error handling or interaction model works for your actual users without testing those journeys with them.
- AI cannot reliably distinguish a harmless visual difference from an obstruction that prevents someone completing a task without the surrounding product context.
- AI cannot provide assurance that your whole website is accessible or transfer responsibility for an accessibility failure away from you and your organisation.
What caps this at PARTLY: verification cost, judgement under ambiguity and context depth.
How we scored this
Five axes, each scored nought to two by hand: ten means AI carries the task cleanly, and the thresholds that turn a total into YES, PARTLY or NO are published in the methodology. Each axis name links to its definition.
| Axis | Score (0–2) |
|---|---|
| Output | 2 |
| Inputs | 1 |
| Verification | 1 |
| Liability | 1 |
| Effort delta | 1 |
| Total | 6 / 10 |
The methodology and its thresholds are published in full.
FAQ
- Can ChatGPT test my website for accessibility?
- It can review screenshots, code and described interaction states and produce a useful first-pass checklist. It cannot perform every browser, keyboard, screen-reader or real-user test unless you carry those out and supply the results.
- Can AI check if my website meets WCAG?
- AI can map visible and coded features to likely WCAG issues, but it cannot honestly certify the whole site from a prompt. Check its findings with automated scans and manual tests across the important user journeys.
- What should I give AI to check my website accessibility?
- Give it current desktop, mobile and responsive designs, component states, HTML or CSS where available, colour and typography tokens, and the journeys users need to complete. Include your target users and separate supplied evidence from anything the model must infer.
- Do I still need an accessibility audit if AI checks my design?
- Yes, if the site is important, complex or used by people who could be excluded by an error. AI can reduce the first-pass effort, but working-site tests and specialist or representative-user input are still needed for issues it cannot verify from the design.
Nearby answers
Assessed by gpt-5.6-luna (gpt-5.6-luna) on 2026-08-13, second-checked by an independent model. Wrong somewhere? Email [email protected] and it gets re-checked.
The newsletter
AI news, new answers and product picks, straight to your inbox.