A website audit for AI search checks whether crawlers can fetch your pages, whether the content is in the HTML they receive, and whether each page answers a real question clearly. The bottom line: most sites that fail do so on access or rendering, which take an hour to find and fix, so audit those first, then check structure, evidence, and freshness page by page for your most important pages.
See also: answer engine optimization, explained step by step.
The checks, with pass criteria
| Check | Pass criteria | How to test |
|---|---|---|
| Search crawler access | OAI-SearchBot, Claude-SearchBot, and PerplexityBot are not blocked by robots.txt, CDN, or firewall | Read robots.txt, review bot protection settings, check server logs |
| Indexing | Key pages are indexed in Google and Bing | Search Console and Bing Webmaster Tools |
| Raw HTML content | Body text is present without JavaScript. Our scan fails under 30 raw words as JavaScript rendered and under 450 as thin | curl the page and count words |
| Headings | One H1 and at least three H2 sections | View source |
| Structured data | JSON-LD with a business entity type plus a supporting type, matching visible content | validator.schema.org |
| Meta and canonical | Description of 40 or more characters, canonical link, Open Graph title, Twitter card, and a sitemap declared in robots.txt | View source and robots.txt |
| Answer readiness | Questions answered directly, with FAQ content or question headings | Read the page |
| Concrete evidence | Numbers, prices or ranges, named entities, dates, and sources | Count and read |
| Freshness | Substance updated within a sensible period, with an accurate date | Check dates and content |
| Internal links | Important pages linked with real anchor tags | Crawl or inspect |
Steps
- List your 10 most important pages. Services, products, pricing, and the pages that answer customer questions.
- Run the free scan on the homepage and two others. It applies eight weighted checks and lists crawler access for 16 AI bots.
- Fetch each page with curl. Compare the raw HTML with what you see in the browser.
- Read robots.txt and any firewall rules. Look specifically for blocks on the search and user fetch bots.
- Verify Bing and Google indexing. Submit the sitemap in both.
- Score each page against the table. Pass, fail, or partial.
- Fix in order. Access, then rendering, then structure, then evidence, then freshness.
- Re run the scan. Confirm the fixes.
Where audits go wrong
- Treating a high readiness score as a result. Google says AI Overviews and AI Mode need no special optimization, and Ahrefs found no citation lift from adding schema to already cited pages.
- Auditing the homepage only. Most questions are answered by inner pages.
- Ignoring a JavaScript rendering problem because the page looks fine in a browser. Vercel found the major AI crawlers do not run JavaScript.
After the audit
A site audit covers the part you control. Whether AI names you also depends on corroboration on other sites and on how your pages compare with competitors', so pair it with a test of real customer questions run many times.
Questions people ask
How do I audit my website for AI search visibility?
Check crawler access, indexing, whether content is in the raw HTML, headings, structured data, answer readiness, concrete evidence, and freshness on your most important pages, and fix in that order.
What is the most common AI visibility problem?
Blocked crawlers or content that only appears after JavaScript runs. Vercel found the major AI crawlers do not run JavaScript.
Which pages should I audit first?
Your 10 most important pages: services, products, pricing, and the pages that answer customer questions, not only the homepage.
Is a good technical score enough?
No. It measures readiness, not outcomes. Only repeated runs of customer questions show whether AI names you.