Answer
What does an AI search audit tool check?
Short answer
An AI search audit tool fetches your pages the way an assistant crawler does and checks four things: whether AI crawlers are allowed in robots.txt, whether each page leads with an extractable answer passage, whether the served HTML carries valid JSON-LD, and whether an llms.txt maps your site. It reports pass or fail per page, then drafts the missing markup.
The four checks that matter
A conventional SEO crawler grades things an assistant does not weigh heavily, and skips the things it does. An AI search audit narrows to what decides retrieval:
- Crawler access. Is GPTBot, OAI-SearchBot, Google-Extended, PerplexityBot, or ClaudeBot blocked in robots.txt or at the CDN? This is the most common hard blocker and it is nearly always accidental.
- Extractable passages. Does each important page answer its question in a self-contained opening paragraph, or does the claim only make sense after three paragraphs of build-up? Assistants lift sentences, not arguments.
- Machine-readable entity data. Is there valid Organization, Product, Article, or FAQPage JSON-LD in the HTML the server returns — not markup injected after hydration, which most assistant crawlers never see.
- llms.txt. Is there a parseable map at your site root telling assistants what you cover and which pages matter?
Why the served HTML is the thing being audited
Plenty of sites pass an audit in the browser and fail for assistants, because the browser runs JavaScript and the crawler often doesn't. Title tags, canonical links, and JSON-LD written by a client-side framework after page load are invisible to a fetch that reads the raw response.
So the audit that counts requests each URL without executing scripts and grades what came back. If a page's body is a near-empty shell in that response, no amount of on-page copy will get cited.
Audit, then fix — in one pass
A findings list is the easy half. The reason AI search audits stall is that every finding resolves to writing markup: a title tag, a meta description, a JSON-LD block, an llms.txt entry, an FAQ answer in the buyer's phrasing.
AgenticSEO produces those artifacts as part of the audit and publishes the approved ones to WordPress, Webflow, HubSpot, or a Git repository, then re-audits on a schedule so the fix is verified rather than assumed. The free check runs the full sequence on three pages with no sign-up.
Real output
A findings block from a real audit
Three findings and the exact output that closes each one. Pass/fail, not a score out of 100 — you can't publish a score.
FAIL robots.txt blocks assistant crawlers
GPTBot disallowed
PerplexityBot disallowed
Google-Extended not mentioned (inherits Disallow: /)
FIX (append to /robots.txt)
User-agent: GPTBot
Allow: /
User-agent: OAI-SearchBot
Allow: /
User-agent: PerplexityBot
Allow: /
User-agent: Google-Extended
Allow: /Highest-severity finding: while this fails, nothing else you fix can be retrieved.
FAIL /pricing served HTML contains no JSON-LD
FIX (insert into <head> of the served response)
{
"@context": "https://schema.org",
"@type": "Product",
"name": "AgenticSEO",
"brand": { "@type": "Organization", "name": "Traction Gap Partners, LLC" },
"offers": {
"@type": "Offer",
"priceCurrency": "USD",
"price": "29.00",
"availability": "https://schema.org/InStock"
}
}Generated from the page's real content, then validated before it is offered for publishing.
# llms.txt > AgenticSEO — AI visibility audits and automated publishing. ## Core pages - [Pricing](https://www.agenticseo.live/pricing): plans, free check, trial terms - [How it works](https://www.agenticseo.live/how-it-works): audit to publish loop - [FAQ](https://www.agenticseo.live/faq): direct answers to product questions
Written site-wide from the crawl, so the map matches the pages that actually exist.
Abbreviated from an AgenticSEO audit of agenticseo.live. Every line below is output the app writes for you.
Related questions
What is an AI search audit?
A check of whether AI assistants can retrieve and cite your site. It grades crawler access in robots.txt, whether pages lead with extractable answer passages, whether the served HTML carries valid JSON-LD, and whether an llms.txt maps the site — page by page, as pass or fail.
How is an AI search audit different from an SEO audit?
A classical SEO audit grades indexability, links, Core Web Vitals, and keyword targeting. An AI search audit grades retrievability by assistants: crawler permissions for GPTBot and peers, passage extractability, entity markup in the raw HTML response, and llms.txt. They overlap but neither replaces the other.
Is there a free AI search audit tool?
Yes. AgenticSEO's free check audits three pages of your website with no sign-up and no card, and hands back the publishable output — title tags, meta descriptions, JSON-LD, and llms.txt — for the gaps it finds.
Which crawlers should an AI search audit check for?
At minimum GPTBot and OAI-SearchBot (OpenAI), Google-Extended (Gemini training and grounding), PerplexityBot, and ClaudeBot. A blocked or unlisted crawler is the most common reason a well-optimized site never appears in assistant answers.
Related: How to rank in ChatGPT · Free llms.txt generator · Answer engine optimization
More answers: How do you rank in ChatGPT? · What tools are best for keyword research for AEO? · What app should agencies use for AEO recommendations? · How is Profound different from other AEO tools? · What is AI visibility tracking software? · What belongs on an answer engine optimization checklist?
Check whether AI names you — or your competitor
Free check with downloadable output. No card, no sign-up.