AI Search Audit: How to Check Your Website in Five Steps

An AI search audit checks whether AI assistants such as ChatGPT, Perplexity and Google's AI Overviews can reach your website, read it, quote it and describe your business accurately. It runs in five steps, in order: crawler access, indexing, content, entity consistency and measurement. Fixing a later step is wasted effort until the earlier ones pass.

The order is the whole point. Most AI search audits sold today start with content scores or "AI readiness" grades and skip the plumbing. A page with perfect answers is invisible if a firewall is turning the crawler away. Below is the audit Alchelyst runs, step by step, with what to look for and what to fix first.

What is an AI search audit?

An AI search audit is a structured review of how visible and citable a website is to AI-driven answer engines. It covers the technical access those systems need, the structure of the content they quote, and the consistency of the facts they combine from across the web.

It overlaps heavily with a technical SEO audit, because the foundations are the same. Google says pages need to be "indexed and eligible to be shown in Google Search with a snippet" to appear as supporting links in AI Overviews, and that no special markup or AI files are required (Google Search Central). The difference is in what an AI search audit checks on top: whether individual sentences can be quoted, and whether assistants describe the business correctly.

Step 1: Can AI crawlers reach your site?

Start with access, because nothing else matters if crawlers can't get in.

Check:

• Your robots.txt allows Googlebot, Bingbot, OAI-SearchBot (ChatGPT search) and PerplexityBot (Perplexity search).
• You've made a deliberate decision about training crawlers such as GPTBot. Blocking them is fine; blocking search crawlers by accident isn't.
• Your CDN, host or security plugin isn't blocking these crawlers. Bot protection often does this silently, so check its logs or settings, not only robots.txt.
• Your important content is in the HTML. Text that only appears after scripts run, or that lives inside images, is harder for crawlers to read.

What usually breaks: a robots.txt copied from a template that blocks "all bots except Google", or a bot-protection setting switched on during a spam problem and never revisited.

Step 2: Are your pages indexed on one clean address?

AI features built on search indexes can only use pages those indexes contain. This step checks that your pages are in them, once, on the right address.

Check:

• Your site answers on one host, either www or the bare domain, and permanently redirects the other.
• Every page's canonical tag points to that host.
• Your XML sitemap is submitted in Google Search Console and Bing Webmaster Tools.
• Search Console's page indexing report doesn't show important pages as excluded.
• No important page carries a noindex or nosnippet tag by mistake.

What usually breaks: a site that redirects one host to the other while its canonical tags point at the host that redirects. Alchelyst found exactly this on its own site and fixed it before publishing anything here.

Step 3: Does your content answer questions in quotable sentences?

This is where most of the work is, and where most sites fail. AI assistants quote sentences that answer part of a question on their own and say who's making the claim.

Check each key page for:

• A 40 to 60 word direct answer at the top, before any background.
• Headings phrased as the questions buyers actually ask.
• One claim per sentence, with the business named instead of "we".
• A source named and linked for every statistic.
• A plain-text statement of what you do, for whom and where.
• At least one claim a competitor would disagree with.

A quick test: copy any sentence from your homepage into a blank document. Does it still make sense, and does it say who it's about? If not, an assistant can't use it either.

What usually breaks: pages written to sound impressive rather than to answer anything. "Innovative, tailored solutions" survives a brand review and fails this step every time. If that's your problem, it's a positioning issue more than a writing one, covered in why marketing ends up sounding like competitors.

Step 4: Is your business described the same way everywhere?

Assistants describe your business by combining what they find across the web. This step checks that every source says the same thing.

Check:

• Name, address, phone and email are identical on your site, Google Business Profile, LinkedIn and directories.
• The founder or key people are named on your site and connected to the business where they appear elsewhere.
• One plain sentence describing the business is used everywhere, not five different taglines.
• Organization schema is on every page, and article pages carry Article and BreadcrumbList schema with a named author.
• All structured data matches visible text on the page.

What usually breaks: an old address on a directory, a former business description on LinkedIn, or schema describing services the page no longer mentions.

Step 5: Are you measuring what AI assistants say?

The last step sets up the measurement you'll use after the fixes. No tool captures every AI answer about your business, so the audit establishes a baseline you can repeat.

Set up:

• A list of 10 to 20 questions buyers ask before hiring you, in their words.
• A record of what ChatGPT, Perplexity, Google and Claude say in answer to each one today: mentioned, linked, accurate.
• Referral tracking in your analytics for visits from AI assistants.
• A monthly reminder to run the same questions again.

What should you fix first after an AI search audit?

Fix in the order of the audit. Access and indexing problems are fast to fix and block everything else, so they always go first. Content is slower and usually delivers the most, so start with the pages closest to a buying decision: services, pricing and comparisons, not the blog. Entity work runs alongside and compounds over months.

The mistake to avoid is rewriting fifty pages before checking whether crawlers can reach any of them.

Can you run an AI search audit yourself?

Yes. Every check above can be done with free tools: your robots.txt file, Google Search Console, Bing Webmaster Tools, a browser and the assistants themselves. Expect a few hours for a small site. The parts that take judgement are step 3, deciding which claims are worth making, and step 4, finding every stale listing.

The free AI Search Visibility Checklist below turns these five steps into 30 checks you can tick off. If you'd rather have Alchelyst run it, that's the starting point of the SEO / AEO service, and the background is in what answer engine optimisation is.

Frequently asked questions

What is an AI search audit?

An AI search audit reviews whether AI assistants like ChatGPT, Perplexity and Google's AI Overviews can reach a website, read and quote its content, and describe the business accurately. It covers crawler access, indexing, content structure, entity consistency and measurement.

How is an AI search audit different from an SEO audit?

An AI search audit includes the technical checks of an SEO audit, such as crawling, indexing and canonicals, and adds checks specific to AI answers: whether AI search crawlers are allowed, whether sentences can be quoted on their own, and whether assistants describe the business accurately.

How often should I audit my website for AI search?

Run a full audit once, then repeat the access and indexing checks after any site migration, redesign or hosting change. Check what AI assistants say about your business monthly, using the same question list, so you can see the trend.

Do I need special tools for an AI search audit?

No. Google Search Console, Bing Webmaster Tools, your robots.txt file and the AI assistants themselves cover every check. Paid tools can speed up crawling large sites, but they don't replace asking the assistants your buyers' questions directly.