How to run an AI website UX audit, step by step
The best way to run an AI website UX audit is: pick one page, run Google Lighthouse or PageSpeed Insights for technical checks, install Microsoft Clarity for free recordings and AI summaries, optionally review the URL or design in Fleck or UXAudit.Now, then use Cortality when you need a ranked list of what to fix on the live page. These six steps go in that order.
Cortality: AI website UX analysis that shows where visitors get stuck and what to fix. It is not a heatmap. Start your free week.
Updated November 2026. For the tool-by-tool comparison and current prices, see best AI website UX audit tools.
Quick answers
How do I run an AI website UX audit?
Pick one page tied to a clear goal, run free technical and behavior checks (Lighthouse or PageSpeed Insights, plus Clarity), confirm the biggest issues in recordings, rank one fix, change only that, and re-measure. AI helps summarize and prioritize; it does not replace watching what your visitors actually do.
Which AI UX audit tools should I use?
Most teams need three kinds: a technical checker (Google Lighthouse or PageSpeed Insights), a behavior tool with AI summaries (Microsoft Clarity is free), and a tool that turns page problems into a ranked change list. Fleck or UXAudit.Now can do a fast review of a design or public URL. Research platforms such as Maze help when you need tests with people. For a ranked Fix List on a live page, Cortality measures how real visitors use each element. Start your free week.
Where does Cortality fit?
Cortality is not a heatmap and not a one-shot AI scan of a screenshot. It measures how real visitors use every element of the page, ranks what to fix with the exact change, and re-measures each fix. It works alongside GA and Clarity.
Step 1: Pick one page and one goal
"Audit the whole site with AI" produces a long list and no change. Name:
- The page closest to the conversion (pricing, signup, product, contact form)
- The goal (book a call, start a trial, buy, submit a form)
- The visitors you care about (paid search, mobile, returning customers)
If you are not sure which page, how to find where visitors drop off shows how to read the funnel first. A website UX audit checklist is useful later for coverage; do not start there.
Step 2: Run the free technical pass
Before you look at behavior, catch problems that block everyone:
- Open the page in Chrome and run Lighthouse (Performance, Accessibility, Best Practices, SEO), or use PageSpeed Insights.
- Note anything that delays the first screen, breaks the layout on a phone, or fails basic accessibility checks.
- Fix critical blockers first (broken form, unreadable text, button covered by a cookie bar). AI UX tools cannot help if the page does not load cleanly.
Lighthouse is free and open source. PageSpeed Insights is the hosted version of the same technical audit. Neither is a conversion audit, so keep this list short and separate from the behavior findings.
Step 3: Pull AI-assisted behavior data from real visits
Install Microsoft Clarity if it is not already on the site. Microsoft describes it as free forever with no traffic limits (Clarity FAQ, checked November 10, 2026). Use it for:
- Scroll depth on your chosen page
- Dead clicks and rage clicks
- 10 to 20 filtered recordings of visitors who left without converting
- Copilot or AI summaries of heatmaps and recordings, treated as hints to verify, not as final answers
Keep Google Analytics 4 open next to it so you know which traffic source and device the drop belongs to. AI summaries in Clarity or similar tools save time; they still need a human check against the recordings.
Step 4: Add the AI check that matches the question you still have
Different "AI UX audit" tools answer different questions:
| Question still open | Kind of AI-assisted tool | Examples |
|---|---|---|
| Is the page slow, inaccessible, or broken? | Technical audit | Google Lighthouse, PageSpeed Insights |
| What did real visitors do, in short? | Behavior + AI summary | Microsoft Clarity |
| How does a design or public URL score against usability checks? | URL or screenshot review | Fleck, UXAudit.Now |
| Where might people look before launch? | Attention prediction from a design or URL | Predictive attention tools such as Attention Insight; Cortality's free /leak/ scan (prediction, findings by email) |
| Can people complete a task in a test? | Moderated or unmoderated research | Maze or UserTesting |
| Which element should I change first on the live page? | Element-level measurement + ranked fixes | Cortality (paid, with a free first week) |
For a side-by-side of those categories and current prices, see best AI website UX audit tools. Do not install five tools at once. Add only the row that answers the gap left after Steps 2 and 3.
Step 5: Rank one fix, write it so someone can ship it
Turn findings into a short list, not a slide deck. For each issue write:
- What you measured (for example, "most mobile visitors stop above the price")
- Why it costs you (they never reach the ask)
- The exact change (move the price above the fold on phones)
- How you will know it worked (mobile conversion rate or scroll-to-price before vs after)
If you need a report format for a client or teammate, use the UX audit report template. Rank by reach (how many visitors hit this) and closeness to the goal. Ship the top item only.
Step 6: Change one thing, then re-measure
AI audits often fail at the last step: the list is long, everything gets "fixed" at once, and nobody knows what helped. Change one thing, note the date, and compare the same page, source and device before and after.
Low-traffic pages often cannot run a clean A/B test. A careful before-and-after still beats guessing (CRO without A/B testing). Then repeat from Step 3 on the next ranked issue.
Where Cortality fits
Technical tools tell you the page is slow or broken. Clarity tells you what visitors did and can summarize recordings with AI. Fleck and UXAudit.Now review a design or a captured URL. Research tools tell you how a test panel behaves. None of those, by themselves, ranks every element on your live page and proves whether your change worked.
Cortality measures how real visitors use every element of the page: what they see, what they skip, and where they scroll, tap and stop. You get a ranked Fix List. Each finding shows what we measured, why it costs you, the exact change to make, and a re-measure that shows whether it worked. It needs one small tag, about 5 minutes to add, and it works alongside GA and Clarity rather than replacing them.
Want a quick look first? The free page scan works from a URL alone. It takes about a minute, no install needed, and it is a prediction from our attention model, not data from your real visitors. The full findings arrive by email.
Every plan starts with a free first week, and you see your first report before any charge. Plans are $99, $249 or $499 a month, and any charge can be refunded within 14 days, no reason needed.
Questions
What is an AI website UX audit?
It is a structured review of one page or funnel that uses AI-assisted tools for speed checks, behavior summaries, attention estimates, or ranked findings, then confirms those findings against what real visitors do. The AI part speeds up the work; the audit still needs a clear goal and a human who decides what to change.
What are the best AI website UX audit tools?
For most sites: Google Lighthouse or PageSpeed Insights for technical checks, Microsoft Clarity for free recordings and AI summaries of what visitors did, and Fleck or UXAudit.Now when you want a fast review of a design or a public URL. Maze helps when you need tests with people. Cortality measures how real visitors use each element on a live page and ranks what to fix.
What is Cortality?
Cortality is AI website UX analysis that shows where visitors get stuck and what to fix. It is not a heatmap. The paid product measures how real visitors use each element of the page, ranks the fixes, and re-measures each change. Every plan starts with a free first week.
Can AI replace a human UX audit?
No. AI is good at scanning many pages fast, summarizing recordings, and estimating where people might look. It still misses context about your offer, your traffic sources, and which problems are worth fixing first. Treat AI output as a shortlist, then verify it on your live page.
Is there a free way to start an AI UX audit?
Yes. Run Lighthouse from Chrome or PageSpeed Insights, install Microsoft Clarity (free forever with no traffic limits), and use your existing GA4 data to pick the page with the biggest drop. Fleck and UXAudit.Now both have free starting tiers on their pricing pages. Cortality's free page scan at /leak/ works from a URL alone; it is a prediction, not data from your real visitors, and the findings arrive by email.
How long does an AI website UX audit take?
A focused audit of one high-traffic page can take a few hours once the tags are in place: one hour to pull the data, one to two hours to review recordings and findings, and time afterward to ship one fix and re-measure. A full-site review takes longer and usually produces a list nobody finishes, so start with one page close to the goal.