AEO Audit: How to Audit Your Website for AI Search

An AEO audit is a structured review of whether AI search systems such as ChatGPT, Google AI Overviews, Perplexity, Claude and Copilot can reach, read, index, understand and cite your website. It checks crawler access, rendering, indexing, structured data, answer structure, entity and third-party evidence, and sets a measurement baseline so you can prove what changed.

This guide gives you an eight-step AEO audit (also called a GEO audit or AI visibility audit), the official crawler tokens for each AI platform, a copy-ready checklist with pass criteria, and a scoring model to turn findings into a prioritized fix list.

What is an AEO audit?

An AEO audit evaluates the technical and content conditions that decide whether AI answers use your pages. It extends a normal SEO audit with AI-specific crawler controls, answer extractability, entity accuracy and AI visibility measurement.

"AEO audit", "GEO audit" and "AI SEO audit" describe the same job with different labels. Google's own position is that optimizing for generative AI search "is optimizing for the search experience, and thus still SEO," according to its guide to generative AI features in Google Search. That is a useful reminder: most AI visibility failures are ordinary technical or content failures that a standard audit would also catch, plus a few AI-specific checks. If you are new to the concept, start with what answer engine optimization is.

Before you start: scope, inputs and tools

A useful audit is scoped to the pages and questions that drive revenue, not every URL. Set these up first:

• Priority pages: 20 to 50 URLs, including the homepage, product or service pages, pricing, comparison pages, key guides and trust pages (about, security, reviews).
• Prompt set: 50 to 100 real buyer questions across awareness, comparison and purchase intent, including unbranded category questions.
• Engines: ChatGPT, Google AI Overviews and AI Mode, Perplexity, Gemini, Claude and Copilot, weighted by where your audience actually is.
• Access: Google Search Console, Bing Webmaster Tools, Google Analytics 4, server or CDN logs, and the ability to request raw HTML with a command-line tool.

Step 1: Check AI crawler access in robots.txt and your CDN

Confirm that the crawlers which feed AI search can fetch your priority pages, and that you have made deliberate choices about training crawlers. Each major platform now separates search, training and user-triggered fetching into different tokens.

The sources are OpenAI's crawler overview, Anthropic's crawler help article, Perplexity's crawler documentation and Google's list of common crawlers. OpenAI and Perplexity both note that robots.txt changes can take about 24 hours to take effect.

Run these checks:

1. Resolve every token against your robots.txt. Lantern's free AI Crawler Checker shows whether each AI and search crawler is allowed and the exact rule that decided it. Our guide to reading AI crawler checker results explains edge cases such as wildcard groups.
2. Separate search from training on purpose. A common, coherent policy is to allow OAI-SearchBot, Claude-SearchBot and PerplexityBot while making an explicit decision on GPTBot, ClaudeBot and Google-Extended.
3. Check the CDN and firewall. Bot-protection rules can block crawlers that robots.txt allows. Perplexity's docs recommend allowing its bots by user agent plus published IP ranges, and OpenAI and Anthropic also publish IP lists.
4. Check logs. Confirm you see successful (200) responses to search crawlers on priority pages, not 403s or challenge pages.

Blocking Googlebot to stay out of AI Overviews backfires: Google states that robots.txt rules for Googlebot are the control for Search crawling, and Google-Extended does not affect Search. Blocking Googlebot removes you from Search entirely. Use snippet controls or the Search Console setting in Step 3 instead.

Step 2: Check rendering and what crawlers actually see

Make sure your key content is in the raw HTML your server returns, because most AI crawlers do not run JavaScript. Google is the exception: it renders JavaScript as long as resources are not blocked.

Vercel's analysis of AI crawler traffic found that none of the major AI crawlers it measured rendered JavaScript, including GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot and PerplexityBot. They fetched some JavaScript files but did not execute them. Gemini, which uses Googlebot's infrastructure, and Applebot did render.

How to test:

• Request a priority page with a command-line tool (for example curl -A "OAI-SearchBot" https://www.example.com/pricing) and search the response for your main heading, prices, product names and FAQ answers.
• Compare that output with Google's URL Inspection tool, which shows the rendered HTML Googlebot received.
• Flag any page where body copy, pricing tables, reviews or FAQs only appear after JavaScript runs, then fix with server-side rendering, static generation or prerendering.

Our deep dive on why AI crawlers do not render JavaScript covers frameworks and fixes in detail.

Step 3: Check indexing and eligibility in Google and Bing

A page cannot be cited from a search index it is not in. Verify indexing, snippet eligibility and AI feature settings in both Google and Bing.

For Google:

• Index and snippet eligibility: Google says a page must be indexed and eligible to show with a snippet to appear in AI Overviews or AI Mode. Use URL Inspection on each priority page and confirm it is indexed with the canonical you intended.
• Snippet controls: Check for nosnippet, max-snippet, data-nosnippet and noindex. These limit what Google shows in AI features, so make sure none are applied by accident through templates or the X-Robots-Tag header.
• Search generative AI setting: Search Console has a control under Settings that includes your site by default; Google says it rolled out to all websites worldwide as of 31 August 2026. Choosing to exclude your site from Search generative AI features removes it from AI Overviews, AI Mode and generative features in Discover, so confirm the setting (and any inherited parent setting) matches your intent.
• Canonicals and duplicates: Conflicting canonicals and near-duplicate pages split signals and waste crawl budget.

For Bing (whose AI Performance report covers citations in Copilot):

• Verify the site in Bing Webmaster Tools, submit your XML sitemap and inspect priority URLs.
• Use IndexNow, the notification protocol supported by Bing and several other engines, to report new or updated pages quickly.

For ChatGPT, OpenAI's publisher FAQ says any public site can appear in ChatGPT search if OAI-SearchBot is not blocked, and that a noindex tag keeps a page out, provided the crawler can fetch the page to read it.

Step 4: Review structured data and entity markup

Structured data helps machines identify what a page is about and who publishes it, but it is not an AI ranking switch. Google states that structured data is not required for generative AI search and that there is no special schema for it.

Audit for accuracy and consistency rather than volume:

• Organization on the homepage or about page: legal name, logo, URL and sameAs links to official profiles.
• Product, Service or SoftwareApplication on commercial pages, with prices and availability that match the visible page.
• Article with author and dates on editorial content.
• Breadcrumb for clear site hierarchy.
• FAQPage only where the questions and answers are visible on the page.

Validate with Google's Rich Results Test and the Schema Markup Validator, and fail any markup that contradicts visible text. Google's AI features guidance specifically asks that structured data match what users see.

Step 5: Audit content and answer structure

Check whether each priority page answers its main question clearly, early and in passages that make sense when quoted on their own. This is where most AEO audit findings come from.

Score each page on:

• Answer-first opening: does the first paragraph directly answer the query the page targets?
• Question-shaped headings: do H2s and H3s match how buyers ask (for example "How much does X cost?")?
• Self-contained passages: can a paragraph, list or table be lifted out and still be accurate without the surrounding text?
• Specifics: prices, limits, dates, comparisons and named sources instead of generic claims.
• Non-commodity value: Google's guidance favors unique, experience-based content over summaries of what is already online.
• Freshness: visible "updated" dates and current facts on pages about fast-changing topics.
• Coverage of follow-up questions: map the related sub-questions an AI system might search (our explainer on query fan-out shows how), and check whether you answer them.

Do not split pages into fragments or write a separate page for every phrasing. Google says chunking is not required and warns that mass-producing pages for fan-out queries can violate its scaled content abuse policy.

Step 6: Audit entity clarity and third-party evidence

Check that AI systems describe your brand correctly and that independent sources support the claims you want repeated. AI answers lean heavily on what others say about you.

Lantern's AI Recognition Gap research found that third-party sites made up 98.9% of ChatGPT source links and 99.5% of Google AI source links for the emerging brands studied. In the same research, ChatGPT correctly identified a brand in 95.0% of answers that named it, while Google AI did so in 56.8%, so entity accuracy cannot be assumed.

Run these checks:

1. Entity accuracy: ask each engine "What is [brand]?" and "Who is [brand] for?" and record errors in category, pricing, location or product names.
2. Name consistency: confirm the same brand name, description and key facts on your site, social profiles, directories and review platforms.
3. Third-party presence: list the domains cited in answers to your priority prompts (review sites, publications, forums, comparison lists) and note where you are absent.
4. Owned page coverage: check whether product, comparison and trust pages exist for the questions buyers ask. Lantern's Beyond the Homepage report shows which owned page types ChatGPT and Google AI cite most.

Google also notes that seeking inauthentic mentions is not helpful, so the remedy is genuine coverage, reviews and partnerships rather than manufactured posts.

Step 7: Set a measurement baseline

An audit without a baseline cannot show improvement. Record current AI visibility, citations and referral traffic before you change anything.

Build the baseline from four sources:

• Google Search Console: the Generative AI performance report shows impressions from AI Overviews and AI Mode by page, country and device; Google says it rolled out to all sites worldwide as of 31 August 2026.
• Bing Webmaster Tools: the AI Performance report, launched in public preview in February 2026, shows citations across Copilot and Bing AI summaries, cited pages and grounding queries. A June 2026 update added intents, topics and citation share.
• Analytics: segment AI referrals in GA4. ChatGPT adds utm_source=chatgpt.com to referral URLs; see our guide to tracking AI referral traffic.
• Prompt tracking: run your prompt set across engines and record mentions, citations, sentiment and competitor share of voice.

Use enough prompts. In the AI Recognition Gap research, expanding a tracked prompt set from 10 to 100 questions cut variation in reported brand coverage by about 73%. Lantern's free AI Visibility Report is a quick way to get a first snapshot of how ChatGPT, Perplexity, Gemini and Claude surface your brand. For a full method, see how to measure AI search visibility step by step.

Step 8: Score findings and build the fix list

Turn findings into a short, ordered backlog. Score each checklist item 0 (fail), 1 (partial) or 2 (pass), then rank fixes by impact and effort.

A simple priority order works for most sites:

1. Blockers first: anything that stops access or indexing (blocked search crawlers, firewall challenges, noindex or nosnippet mistakes, JavaScript-only content).
2. Accuracy next: wrong facts in AI answers, mismatched structured data, outdated pricing.
3. Coverage gaps: missing pages or sections for priority questions and comparisons.
4. Evidence gaps: third-party sources where competitors appear and you do not.

Re-run the technical checks after every major release and the full audit quarterly. Keep the prompt tracking running monthly so you can tie changes to results.

AEO audit checklist

Use this checklist as your working audit sheet. Copy it into a spreadsheet, add a score column (0, 1 or 2) and an owner for each row.

Common AEO audit mistakes

The most expensive mistakes are scoping and interpretation errors, not missed checks:

• Auditing the homepage only. AI answers often cite product, comparison and editorial pages, and third-party sites far more than homepages.
• Testing a handful of prompts once. AI answers vary between runs, so small prompt sets give noisy results.
• Treating llms.txt as a fix. Google says its Search features ignore it; see our llms.txt guide for when it is still worth publishing.
• Blocking every AI bot by default. Blocking search crawlers along with training crawlers removes you from the answers you want to appear in.
• Trusting a single score. Tools that return one "AI readiness" number hide which issue matters. Keep the item-level evidence.
• Skipping Bing. Bing Webmaster Tools now reports how often your pages are cited in Copilot and Bing's AI summaries.

If you want the monitoring side handled continuously, Lantern tracks mentions, citations, sentiment and share of voice across major AI engines, connects them to Search Console and analytics data, and has agents turn gaps into review-ready content and technical tasks. Our AI search visibility tools buyer's guide compares monitoring options if you are evaluating vendors.

Frequently asked questions

How do I check the AEO of my website?

Start with three quick checks: run your domain through an AI crawler checker, fetch a key page's raw HTML to confirm content is visible without JavaScript, and ask ChatGPT, Perplexity and Google AI Mode five of your buyers' questions. Then work through the full checklist above for your priority pages.

Is a GEO audit different from an SEO audit?

It overlaps heavily. A GEO audit keeps the SEO fundamentals (crawling, indexing, content quality) and adds AI-specific checks: separate crawler tokens, non-JavaScript readability, entity accuracy in AI answers, third-party citation sources and AI visibility measurement.

Should I block GPTBot or ClaudeBot?

That is a business decision about model training, not search visibility. OpenAI and Anthropic document that blocking GPTBot or ClaudeBot does not block their search crawlers (OAI-SearchBot and Claude-SearchBot), so you can opt out of training while staying eligible for AI search answers.

How often should you run an AEO audit?

Run the full audit quarterly and the technical checks after any redesign, migration, CMS change or new bot-protection rule. Track prompt-level visibility monthly, because AI answers and cited sources shift faster than rankings.

Can I do an AEO audit for free?

Yes. Google Search Console, Bing Webmaster Tools, Google's Rich Results Test, Lighthouse, a command-line fetch and manual prompt testing cover most of the checklist. Paid tools mainly save time on prompt tracking across engines and competitor citation analysis.

What should an AEO audit report include?

It should include the scored checklist, evidence for each failed item (screenshots, rules, URLs), the baseline metrics, and a prioritized fix list with owners and dates. Leadership needs the top five blockers and the baseline; implementers need the item-level detail.

AI search readiness is not a separate discipline from good technical SEO, but it does add new crawler rules, new reports and new ways to fail quietly. Run this audit once to fix the blockers, keep the baseline current, and repeat the technical checks whenever your site or the engines change.