GEO Audit: Ten Checks, the Free Google and Bing Reports, and the Proof Each Needs

A GEO audit (generative engine optimization audit) checks whether AI answer engines such as Google AI Overviews, ChatGPT search, Microsoft Copilot and Perplexity can reach your pages, quote them and name your brand. Start with free first-party sources: your robots.txt, Search Console's URL Inspection and generative AI performance report, and Bing Webmaster Tools' AI Performance report. Then log a fixed set of prompts for engines that publish no report.

A trustworthy GEO audit backs every finding with evidence you can open again: a robots.txt line, a raw HTML response, a report export, a logged AI answer with its date. Below are the ten checks and the proof each needs, the two free reports to read first, the crawler mix-up many audits get wrong, and where any audit stops. (If you searched "geo audit" for a location audit of local listings, that is a local SEO job; GEO here means generative engine optimization.)

What does a GEO audit check?

A GEO audit splits into checks you can settle on your own site in an afternoon and checks that need data from the engines or sampled AI answers. Each row names the evidence that settles it. If a finding in an audit has no evidence behind it, treat it as an opinion until someone produces one.

Check	Evidence that settles it	Pass line
AI search crawlers can reach the pages	robots.txt, CDN and firewall bot rules, server log lines per user agent	No block on the agents that feed search answers; logs show successful responses to them
The answer text is in the raw HTML	The page source fetched without running JavaScript	The main answer and key facts appear in that source, not only after scripts run
Google can index and snippet the page	URL Inspection in Search Console, robots meta tags on the template	Indexed, with no nosnippet or tight max-snippet on pages you want quoted
Pages lead with a quotable answer	The opening paragraph of each priority page	The page's main question is answered in its first paragraph
Structured data matches the visible page	Validator output set beside the rendered page	Every marked-up fact (price, author, address) also appears in the visible text
Entity facts agree everywhere	A side-by-side list of name, description and key facts from the site, profiles and directories	No conflicting facts
Google AI impressions	Search Console's generative AI performance report	Impressions per page recorded with the date range used
Copilot and Bing citations	AI Performance in Bing Webmaster Tools	Citations per page recorded with the date range used
Appearance and citation rate in other engines	A log of every answer collected, per prompt and engine	Stated as answers naming or linking you, divided by answers collected, with the run count
Source map	Every domain cited in the logged answers, counted	Top cited domains listed, with whether you appear on each

The first six rows are deterministic: run them twice and you get the same answer. The next two come from the engines' own reports. The last two are samples, and they need the extra discipline covered further down. This post is about proving each row, not about fixing it: for the structured data row, our guide to schema markup testing tools covers which validator answers which question, and the content and entity fixes are in our guide to optimizing your website for AI search.

Which free reports should a GEO audit start with?

Start with the two reports that come from the engines themselves: Search Console's generative AI performance report for Google, and AI Performance in Bing Webmaster Tools for Copilot and Bing. Both are free for a site you have verified, and they are the only sources in a GEO audit that are not a third party's estimate. Our roundup of free SEO tools from Google and Bing covers where they sit among other no-cost tools.

Search Console's generative AI performance report

Google's Search Central announcement says this report reached all websites worldwide as of August 31, 2026. Per Google's help page for the report, it counts impressions in AI Overviews and AI Mode by page, with no queries, so it tells you which URLs Google surfaced, not which searches did it.

For an audit, export it by page and keep the export with its date range as the proof for the row. The URLs that already earn AI impressions show what Google is willing to surface from your site, and priority pages missing from that list are your gap. If the report is not there at all, Google's help page says this "usually happens when your site hasn't received enough impressions in generative AI features," which is itself a finding.

AI Performance in Bing Webmaster Tools

Microsoft introduced AI Performance in Bing Webmaster Tools as a public preview on February 10, 2026, covering "Microsoft Copilot, AI-generated summaries in Bing, and select partner integrations", per its Bing Webmaster blog announcement. Total Citations counts how often your content is shown as a source, Average Cited Pages gives the daily average of your unique pages cited, and page-level counts show which URLs are cited most. Microsoft says the grounding queries it lists (the phrases the AI used to retrieve your content) are "a sample of overall citation activity," and that citation counts do not indicate placement within an answer.

For the audit file, export the page-level citations with their date range. Our own export is the reason this row is in the table: it showed one page collecting most of crawltide.com's citations, and the numbers, with why we do not claim the page's format caused them, are in what we learned auditing crawltide.com for AI visibility.

Which AI crawlers does a GEO audit need to check?

Check the search crawlers, not the training crawlers. OpenAI documents three separate agents, and only OAI-SearchBot controls whether you can appear in ChatGPT search. According to OpenAI's crawler documentation, "OAI-SearchBot is used to surface websites in search results in ChatGPT's search features," while GPTBot crawls content "that may be used in training our generative AI foundation models."

That split changes how an audit should read your robots.txt. A rule that disallows GPTBot is a training decision, not a visibility problem, and an audit that reports it as "blocking ChatGPT" has merged two different agents. OpenAI says sites opted out of OAI-SearchBot "will not be shown in ChatGPT search answers, though can still appear as navigational links." It also notes "it can take ~24 hours from a site's robots.txt update for our systems to adjust," so a re-check the same afternoon proves nothing.

ChatGPT-User is the awkward one. It visits a page when a user asks ChatGPT to, and OpenAI says that "because these actions are initiated by a user, robots.txt rules may not apply." A clean audit lists that as a gap in your control, not as something a robots.txt edit fixes.

Google works differently again. Its list of common crawlers describes Google-Extended as a token for Gemini training and grounding, and states that "Google-Extended does not impact a site's inclusion in Google Search nor is it used as a ranking signal in Google Search." AI Overviews depend on ordinary Googlebot access and indexing.

Robots.txt is only half of access. Many CDNs and firewalls ship bot rules that challenge or block unfamiliar user agents, and those blocks never show up in robots.txt. If your server logs show AI search crawlers getting challenge pages or errors, fix the bot rule first. To confirm robots.txt itself loads and parses, the free robots.txt and sitemap checker does it in one pass.

Does Google need special markup or an AI text file for AI Overviews?

No. Google's guide to AI features in Search says a page "must be indexed and eligible to be shown in Google Search with a snippet," and that "there are no additional technical requirements." It adds: "You don't need to create new machine readable files, AI text files, or markup to appear in these features," and "there's also no special schema.org structured data that you need to add."

So the Google rows of a GEO audit are Googlebot access, indexing and snippet controls. Run URL Inspection on your five or ten most important pages and look for an indexed status and no nosnippet directive; the free indexability checker flags noindex and nosnippet on a single URL. An audit that fails you on Google's AI features for lacking an AI text file is grading something Google says it does not need. Our post on llms.txt covers what that file can and cannot do. Structured data still helps search results in general, so keep it accurate; just do not treat it as an AI Overviews switch.

How do you log AI answers for engines that publish no report?

For ChatGPT, Perplexity, Claude and Gemini, the proof is a raw prompt log: one row per answer, with the date, the engine, the prompt word for word, whether the answer named you, whether it linked you, and every URL it cited. Choosing the prompts is covered in the find stage of our find, fix, verify loop; the audit's job is to keep the list identical between audits and hand over the log itself, not a summary of it.

From that log the audit reports the last two rows of the table: an appearance or citation rate stated as answers that named or linked you, divided by answers collected, and a source map counting which domains the answers cite. The source map is where an audit finds the review sites, directories and forums engines rely on for your topic. One run per prompt is an anecdote, and a rate with no denominator cannot be re-tested; our guide to running an AI visibility test without fooling yourself covers how many repeats you need and how to read the spread.

Are free GEO audit scanners worth running?

Yes, once, for their technical checks: crawler rules, page fetchability, raw HTML content and whether structured data is present. Use a scanner to catch an obvious block or a page that renders blank without JavaScript, then verify each finding yourself with the rows above.

Ignore the headline score for audit purposes: it is the vendor's formula, and scores from two tools do not compare (our post on how an AI visibility score is calculated explains why). Some scanners ask a language model to answer "as ChatGPT would"; that is a simulation, not ChatGPT's real answer, and it belongs in a different column from the answers you logged. Our free AI visibility check is a readiness scan of this kind: it scores one page's signals and sends no prompts to any AI engine.

What should a paid GEO audit hand you?

A paid GEO audit should deliver evidence for every finding, a baseline with its method written down, and a ranked backlog with an owner per item; a score without those is a slide, not an audit. Five questions separate a real audit from a presentation:

1. Which user agents did you test, and what does each block in my robots.txt actually prevent?
2. How many prompts, which engines, how many runs per prompt, and can I have the raw answer log?
3. What did you divide by for each rate?
4. Do your Google numbers reconcile with my Search Console generative AI report, and your Copilot numbers with Bing's AI Performance report?
5. If you recommend structured data or an AI text file for AI Overviews, which problem does it solve, given Google's statement that neither is needed?

A vendor who answers all five with files rather than adjectives is selling an audit you can re-run without them.

What a GEO audit cannot tell you

A GEO audit cannot guarantee that any engine will cite you, because the engines choose sources and do not publish how. It finds the obstacles you control and measures where you stand on a given date; it does not predict next month's answers.

• Clicks from AI answers: Google's AI features guide says this traffic is counted in the overall Performance report under the "Web" search type, and neither free report described here splits out clicks from AI answers.
• Personalized answers: sampled prompts run logged out, so they do not show what a signed-in user with history sees.
• User-triggered fetches: per OpenAI, robots.txt may not apply to ChatGPT-User, so access control is never complete.
• Engines with no report: for ChatGPT, Perplexity, Claude and Gemini you only know what your own prompt log recorded.
• Cause and effect: if your citations rise after a fix, the audit shows correlation. The engines, their indexes and your competitors moved in the same period.

Where does CrawlTide fit in a GEO audit?

CrawlTide automates the repeatable rows; the human rows stay human. Its site audit crawls your pages for technical issues, and the AI Readiness checklist in AI Visibility checks that search crawlers such as Googlebot, Bingbot and OAI-SearchBot are allowed, that pages are not noindexed and that snippets are allowed. Blocking training crawlers like GPTBot does not count against you.

For the answer-side rows, CrawlTide AI Visibility checks your tracked prompts in Google AI Overviews, ChatGPT Search, Claude, Gemini and Perplexity on a fixed schedule (every 7 days on Pro), records whether each answer mentioned you, cited you or left you out, and shows which competitors appear instead. With Search Console connected, it brings Google's own data in beside it. Neither replaces judgment on rows four to six: whether your opening paragraphs answer the question, and whether your facts agree across the web, still needs a person reading the pages. After the audit, the fixes belong in a cycle; the find, fix, verify loop for generative engine optimization sets out that order.

Run the first hour of your GEO audit today

Open your robots.txt and write, next to every AI user agent it mentions, the decision that rule reflects: training, search or user fetches. Then open Search Console's generative AI report and Bing's AI Performance report and note which pages each one lists. If you cannot say what a robots.txt rule is for, or your priority pages appear in neither report, those are your first two findings.