How to audit your website for AI search visibility: a step by step checklist

Run five checks: confirm AI crawlers can reach the site, confirm each page answers its core question in the opening lines, confirm schema markup describes the content correctly, confirm a llms.txt file exists, and ask the assistants directly what they know about the business. Each check takes minutes and together they show the real gap.

Worth doing this month rather than filing under someday: Ahrefs recently found that visitors arriving via an AI assistant convert into leads at around 23 times the rate of a typical Google visitor. A site an assistant cannot read properly never gets the chance to earn that traffic.

Step 1: Confirm the crawlers can actually reach the page

Open the site's robots.txt file and check it is not quietly blocking the crawlers that feed today's assistants. This single file causes more missed citations than any content problem.

• GPTBot, which feeds ChatGPT.
• PerplexityBot, which feeds Perplexity.
• Google-Extended, which feeds Gemini and AI Overviews.
• ClaudeBot, which feeds Claude.

Step 2: Check whether the page answers its own question early

Pick the ten pages most likely to be asked about and read only the first two sentences of each. If the core question is not answered by the end of the second sentence, an assistant summarising the page will struggle to quote it cleanly, and will often reach for a competitor's page instead.

Step 3: Check the structured data actually matches the page

Structured data tells an assistant what kind of content it is looking at rather than making it guess. Confirm the relevant schema types are present and, more importantly, that the values inside them match what is actually on the page.

• Article or BlogPosting, matched to the real headline and date.
• FAQPage, matched to the real questions and answers shown on the page.
• Organization, with a consistent name and description across every page that uses it.

Step 4: Publish and check a llms.txt file

A llms.txt file at the site root gives an assistant a short, plain summary of what the business does and where to find the pages that matter most. It will not force a citation on its own, but for a site an assistant is already unsure about, it removes an easy reason to skip it.

Step 5: Ask the assistants directly what they know

Open ChatGPT, Gemini and Perplexity and ask each one what the business does and who it competes with. Josh Harris, founder of The Sparked Group, runs this step on every audit: "The answer an assistant gives back is the most honest audit result you will get. It shows you exactly what it thinks it knows, and where it is guessing."

An assistant cannot cite a page it cannot read or does not understand.