Your site ranks on page one of Google for the exact question your best customers ask, and ChatGPT still recommends a competitor instead. That gap is not a ranking problem. It is a structural one: the schema, the answer format, and the crawler permissions that let an AI agent read your page with confidence are not in place yet. Ignore it long enough and the traffic that used to click through starts arriving pre-answered, with someone else's name attached.
An AEO audit is a structured review of the technical and content signals that let AI agents such as ChatGPT, Perplexity, Claude, and Google AI Overviews cite a page with confidence: schema markup, crawler permissions, heading structure, and direct-answer formatting. It is the AEO equivalent of a technical SEO audit, built around citation instead of ranking. Most site owners have never seen one, so it is worth walking through what actually happens during the four stages.
How Is an AEO Audit Different From a Regular SEO Audit?
A technical SEO audit checks whether Googlebot can crawl, index, and rank a page: sitemaps, canonical tags, backlink profile, Core Web Vitals. An AEO audit checks whether an AI agent can lift an answer from that same page with confidence: schema type coverage, whether headings are phrased as the question a person would type, and whether GPTBot, ClaudeBot, and PerplexityBot are even allowed to reach the page. The two audits overlap on clean HTML and fast loading, but a site can pass every SEO check and still fail every AEO check, because ranking and citation are graded on different signals.
A single robots.txt line illustrates the gap: User-agent: GPTBot / Disallow: /blog/ keeps a blog fully indexed in Google while making every post invisible to ChatGPT's crawler. Nothing about that line is an SEO problem. It only shows up on an audit built to check AI agent access specifically.
What Are the Four Stages of an AEO Audit?
A structured AEO audit moves through four stages, each with its own deliverable: Audit, Architect, Implement, and Monitor. Skipping a stage is the most common reason a homegrown audit produces a list of ideas instead of a finished result.
- Audit. Crawl every indexable page and log the gaps: missing schema types, blocked AI crawlers, headings that are not phrased as questions, and pages with no direct-answer paragraph in the first 150 words.
- Architect. Turn the gap list into a prioritized plan, ranking fixes by expected citation lift against the effort to ship them, so the highest-traffic pages get fixed first instead of whichever page is easiest.
- Implement. Write and deploy the actual files and markup: llms.txt at the domain root, AGENTS.md for crawler-readable structure, JSON-LD schema blocks in the head, and an updated robots.txt that allows the major AI crawlers.
- Monitor. Track citation events across AI engines on a weekly cadence — running the site's top category questions through ChatGPT, Perplexity, Claude, and AI Overviews — and feed what changes back into the next round of fixes.
What Does the Audit Report Actually Contain?
A finished audit report is a page-by-page findings table, not a generic score. Each row names the check, what it looked for, and the specific gap found on that page, so the reader can act on it without guessing. Five schema types anchor the report: Organization for the whole site, WebSite with SearchAction on the homepage, BreadcrumbList on every non-home page, FAQPage on any page with a question-and-answer section, and Article on every blog post.
| Report section | What it checks | Example finding |
|---|---|---|
| Crawler permissions | robots.txt directives for GPTBot, ClaudeBot, PerplexityBot, and Google-Extended | GPTBot disallowed on /blog/* |
| Schema coverage | Organization, WebSite, BreadcrumbList, FAQPage, and Article JSON-LD | FAQPage schema present on 2 of 14 eligible pages |
| Heading structure | Share of H2s phrased as a literal question | 0 of 12 H2s on the pricing page are question-shaped |
| File stack | llms.txt, AGENTS.md, and skill.md presence at the webroot | No llms.txt found at the domain root |
How Long Does an AEO Audit Take?
A full AEO implementation for a mid-size site runs 7 to 14 days end to end: 2 to 3 days for the audit itself, 3 to 5 days to produce the schema, llms.txt, and AGENTS.md files, and 2 to 3 days for QA and deployment. A free homepage audit is faster by design, with a 24-hour turnaround, because it checks a single page against the same four report categories instead of the whole site.
How Do You Know the Audit Fixes Worked?
You check citation rate directly, the same way the Monitor stage does: pick your top 10 category questions, run each one through ChatGPT, Perplexity, Claude, and Google AI Overviews, and log which sites get cited. Google Search Console reports referrals from Google's own AI Overviews, but it shows nothing from ChatGPT, Perplexity, or Claude, so a dedicated AI-visibility tool such as AthenaHQ, Profound, Otterly, or SE Ranking's AI Visibility module fills that gap. A newer brand should expect a 5 to 15 percent citation rate across major engines within the first six months; a category leader can reach 30 to 60 percent once the full file stack is in place.
Where a Manual Audit Breaks Down
Everything in the report above is checkable by hand: open each page, read the robots.txt file, inspect the head for JSON-LD, and count question-shaped headings. The work is not complicated. It is repetitive across every page on the site, and it has to be redone every time a crawler changes its behavior or a new schema type becomes standard, which is exactly the kind of maintenance that quietly stops happening once the initial project ends.
That is the gap a done-for-you audit closes: someone runs the same four-stage checklist on a schedule instead of once. Our AEO FAQ page walks through every file type and schema check in the report in more detail, including which crawlers to allow and why FAQPage schema tends to move citation rate the most.
Do You Need a Professional Audit, or Can You Run One Yourself?
You can run this checklist yourself, and plenty of teams do. A developer who is comfortable editing a robots.txt file and adding a JSON-LD script tag can close most of the gaps in the report above in an afternoon per page. The honest trade-off is time and consistency, not capability: a self-run audit tends to happen once and then drift, because there is no standing process forcing a recheck after the next site redesign or crawler policy change.
A paid audit buys back that time and adds a second set of eyes on a stack most teams only touch once. Our four-stage methodology exists for teams that want it done faster, do not have the technical bandwidth this quarter, or want confirmation before shipping schema changes across a large site. If the cost of a full audit is the sticking point, a self-serve Pro membership priced at $49 a month gives your own team the checklist and templates and lets you run the four stages in-house on your own schedule.
What's the Next Step?
Request a free homepage audit and you will get the same four-category report described above, with a 24-hour turnaround, before you commit to anything larger. You will see, page by page, which schema types are missing, which AI crawlers are blocked, and how many of your headings are already phrased as questions, so the decision to go further is based on your own findings instead of a sales pitch. Get your free AEO audit and see which of the four stages your site is missing first.