The GEO Audit Checklist: 15 Things to Check Before You Optimize

Quick Summary

A GEO audit checks whether AI answer engines can reach your pages, read them, trust them, and quote them. Run it before you write a single new word. Most brands that feel invisible in ChatGPT, Perplexity, Gemini, and Google AI Overviews have a retrieval problem, not a content problem.

The 15 checks, grouped by what they fix:

• Retrieval eligibility: AI crawler access, indexation and snippet rules, server-rendered content, and your own crawl log baseline.
• Extractability: Answer placement, fact density, question-shaped headings, and honest freshness signals.
• Entity and authority: Consistent brand naming, third-party presence, and verifiable author credentials.
• Measurement: A fixed prompt set, citation accuracy, competitor share of voice, and AI referral tracking in analytics.

Three facts that shape how you should run this audit in 2026:

• On 15 May 2026, Google published its first official guide to AI features in Search. It said AEO and GEO are still SEO. It also names tactics you can skip: llms.txt, content chunking, and AI-specific rewriting. (Google Search Central)
• FAQ rich results stopped appearing in Google Search on 7 May 2026. FAQ content still helps AI engines. FAQ schema is no longer a search appearance lever. (Search Engine Land)
• Most AI crawling is for training, not search. Search crawling sat at under 10 percent of AI bot requests in May 2026, per Cloudflare Radar. So a blanket bot block quietly cuts you out of citations. (Cloudflare)

Why does the audit have to come first?

Most GEO projects start backwards. A brand sees ChatGPT name three rivals and skips them. The reaction is to publish more. Six months on, the content library has doubled and the citation count has not moved.

The reason is usually dull. The pages were blocked, or thin, or built so the answer never lands until paragraph nine. New content cannot fix a retrieval problem. An audit tells you which one it is before you spend the budget. It also gives you a baseline. That matters more than most teams expect. Without a starting number, you cannot prove any of it worked.

This checklist is what we run at White Bunnie across B2B IT services, SaaS, and ITES accounts. The order matters. The cheapest, highest-impact checks come first.

What is a GEO audit?

A GEO audit is a structured review of one thing whether AI engines can find, read, trust, and cite your pages. It covers four layers. Retrieval eligibility. Content extractability. Entity and authority signals. Measurement. An SEO audit asks where a page ranks. A GEO audit asks a different question: can a passage on that page be lifted into an answer and credited to you?

It does not replace a traditional SEO audit. It sits on top of one. The two share most of their technical layer and diverge sharply at the measurement layer.

How a GEO audit differs from an SEO audit

• Unit of analysis: SEO audits score pages. GEO audits score passages, since AI engines cite fragments.
• Access: SEO audits check Googlebot. GEO audits also check OAI-SearchBot, PerplexityBot, and Claude-SearchBot.
• Success metric: SEO audits end in rankings and clicks. GEO audits end in citation rate, mention rate, and share of answer.
• Overlap: Crawlability, indexation, semantic HTML, and page experience carry over. Google says its AI features run on the same core ranking systems.

The GEO audit checklist: 15 checks

Part 1: Retrieval eligibility

1. Confirm AI crawlers are not blocked in robots.txt

Open yoursite.com/robots.txt and read it line by line. Two groups matter, and they are not the same. Training crawlers feed future models: GPTBot, ClaudeBot, Google-Extended, CCBot. Search crawlers build the live indexes that answers are drawn from: OAI-SearchBot, ChatGPT-User, PerplexityBot, Claude-SearchBot, and Googlebot.

Blocking the training group is a fair business call. Blocking the search group cuts you out of citations for good. Many sites did this by accident when a security vendor pushed a default rule. Check it first. Nothing else here matters if this answer is wrong.

2. Verify indexation and snippet eligibility

Google builds AI Overviews and AI Mode answers from its existing Search index. If a page is not indexed or is cut from snippets, it cannot show up. Audit three things: noindex tags on pages that should rank, blanket nosnippet rules, and tight max-snippet values set site-wide by a CMS plugin.

Run a coverage check in Google Search Console and compare indexed URLs against your priority list. Gaps here are usually the fastest win in the whole audit.

3. Check that key content renders without JavaScript

Most AI crawlers handle client-side rendering poorly. Does your pricing, product name, service copy, or spec sheet only appear after JavaScript runs? Then assume some engines never see it.

Test it. Turn off JavaScript in a browser, or fetch the raw HTML with curl. Look for your core facts. If the page comes back an empty shell, move that content into the server-rendered HTML. This one fix often lifts several checks at once.

4. Pull your own crawl baseline from server logs

Before you trust any dashboard, look at your logs. Filter for AI user agents over the last 90 days. You want to know which engines visit, how often, which sections they favour, and whether any are receiving errors or redirect chains.

Then compare crawl volume against referral sessions. That gives you your own crawl-to-refer ratio. Cloudflare Radar publishes network-wide numbers for context. Those ratios swing hard between windows, so read them as direction, not a target.

Part 2: Content extractability

5. Check where the answer sits on the page

Generative engines read the opening of a page to decide what it covers. Older SEO copy builds toward the point. That structure works against you now.

For each priority page, ask one question: does a direct, self-contained answer appear within the first screen? Add a 40 to 60-word summary near the top. Write it so it still makes sense out of context. This is the most repeatable content fix in the whole audit.

6. Measure fact density

The core academic study on GEO came from Princeton, Georgia Tech, and IIT Delhi. It ran at ACM KDD 2024 and tested methods across thousands of queries. Quotations, statistics, and cited sources lifted visibility by up to roughly 40 percent. Keyword stuffing did not help.

So audit your top 10 pages. Count the data points, named sources, and figures per 500 words. Vague claims get paraphrased. Sourced claims get quoted.

7. Rewrite headings as the questions buyers actually ask

Engines read headings as signals about what a section answers. “GEO Audit Metrics” tells a model very little. “What does a GEO audit measure?” tells it where to look.

Go through your H2 and H3 tags. Rewrite keyword strings as plain questions or plain statements. Keep one idea per section. Do not break the page into fake chunks. Google says its systems handle multi-topic pages fine. This is standard on-page SEO work with a different brief.

8. Audit freshness signals for accuracy, not theatre

Check that dateModified in your schema matches a real content change. Check that prices, product names, statistics, and screenshots are current. Then check for the opposite failure: pages where a script bumps the date monthly while the content sits untouched.

Stale facts hurt more than old dates. A page that cites a 2023 figure as current hands an engine a reason to pick the competitor who updated theirs.

Part 3: Entity and authority

9. Test entity consistency across every surface

Models build a picture of your brand from many sources. Gaps between them weaken it. Search your brand name. Compare how it reads on your site, LinkedIn, Crunchbase, G2, your Google Business Profile, and trade directories.

Look for mismatched legal names, old addresses, abandoned social profiles, different service names, and founder names spelled three ways. Fix the highest-authority sources first.

10. Map your third-party footprint

Most in-house teams skip this one. It is where the leverage sits. A large share of brand mentions in AI answers starts on pages you do not own. Think review platforms, listicles, forums, trade press, and comparison sites.

Run your category prompts. Note which domains keep showing up as sources. If the same five review sites feed every answer in your niche, your presence there matters more than your next blog post.

One caution. Google has warned that manufactured or inauthentic mentions carry real risk. Earn the placements.

11. Verify E-E-A-T signals are provable

Author bylines that link to a real profile. Credentials a reader can check. An about page that names actual people. Organization and Person schema that matches what is on the page.

Ask one blunt question of each priority page. Could a doubting reader confirm that the writer knows this subject? If not, an engine judging trust lands in the same place.

Part 4: Measurement and governance

12. Build and freeze a baseline prompt set

Write 20 to 50 prompts your buyers would actually type. Mix category prompts (“best AEO tools for B2B SaaS”), comparison prompts, problem prompts, and branded prompts.

Run them across ChatGPT, Perplexity, Gemini, Claude, and Google AI Mode. Record which brands and URLs get cited. Then freeze the list. Swapping prompts between runs kills comparability. It is the most common measurement mistake we see.

13. Check what engines say about you, not just whether they mention you

Mention count is a shallow metric. Ask each engine what your company does, who it serves, and what it costs. Note every error.

Wrong pricing, dead products, and old positioning all trace back to a stale source. Finding that source often beats winning one more citation.

14. Benchmark competitor share of voice

Across the frozen prompt set, count how often each rival appears. Note which of their pages get cited, and what those pages do that yours do not. The answer is usually dull. A clearer summary. Better numbers. A comparison section you never wrote.

15. Confirm AI referral traffic is actually being tracked

Most analytics setups bury AI referrals. Build a GA4 segment or channel group for chatgpt.com, perplexity.ai, gemini.google.com, claude.ai, and copilot.microsoft.com before you start optimising.

Expect a small number. AI referral share of web traffic stays low even as usage grows. Much AI-led discovery lands as direct traffic, not a referral. Track it anyway. You cannot report on a channel you never set up.

What your audit should stop you from doing

A useful audit removes work as well as adds it. Google’s May 2026 guidance included a myth-busting section that retires several tactics still sold as GEO essentials.

• txt: Google may crawl the file, but it gives it no special treatment. Other agents may read it, so it is cheap to keep, but it is not a lever.
• Content chunking: No need to fragment articles. Google says its systems extract the relevant passage from multi-topic pages on their own.
• AI-specific rewriting: Engines understand synonyms and intent. You do not need to chase every long-tail variation.
• FAQ schema as a traffic tactic: Rich results ended on 7 May 2026. Search Console reporting followed in June. API support ends this month. Keep FAQ content that answers real questions. Drop the idea that the markup is a shortcut.
• Manufactured brand mentions: Google flagged inauthentic mentions as risky. Paid mention farms are a liability.

For the full technical layer, our Technical AEO Checklist goes deeper into schema and crawler configuration, and our breakdown of what changed in Google’s AI search guidance covers the May update in detail.

GEO audit tools that consistently rank in Google results

Search “best GEO tools” or “AI visibility tools” in 2026. The same names show up across the top-ranking roundups. That consistency signals market adoption. It is not proof of quality. Pricing here also changes monthly, so check before you buy.

Named most often across top-ranking comparisons

• Profound: the reference point for enterprise AI visibility analytics. Deepest crawler and conversation data. Priced for teams with a dedicated function.
• Peec AI: The mid-market favourite. Strong prompt-level reporting and multi-project setup, which suit agencies.
• AI: The cheapest credible paid entry point, widely cited at around 29 dollars per month for basic tracking.
• Semrush AI Visibility Toolkit: The first step if you already pay for Semrush. AI data sits beside your keyword and content reporting.
• Ahrefs Brand Radar: Checks your brand against a large set of search-backed prompts across several AI surfaces. History goes back to 2025.
• Scrunch AI: Built for enterprise and agency programmes managing multiple client brands.
• SE Ranking and AthenaHQ: Mid-tier options that show up repeatedly in agency-focused lists.
• HubSpot AI Search Grader: Free, useful for a first snapshot, not a monitoring solution.

Free tools that belong in every audit

• Google Search Console: Indexation, coverage, and the impression baseline everything else compares against.
• Google Rich Results Test and Schema Markup Validator: Structured data validation.
• PageSpeed Insights: Core Web Vitals, including INP.
• Cloudflare Radar: Public crawler and crawl-to-refer benchmarks.
• Manual prompt testing: Run 20 prompts by hand across four engines. It costs nothing and shows if you have a problem worth paying to fix.

A tool measures the gap. It does not close it. We compared the main tracking platforms in detail in our Ahrefs vs Semrush vs Profound comparison.

How to sequence the fixes

A full audit on a mid-sized site turns up 30 to 60 findings. You cannot fix them all this quarter, and you should not try. Sort by impact against effort.

• Tier 1, fix this week: Crawler blocks, noindex on priority pages, JavaScript-only core content, missing analytics segments. High impact, low effort. Everything else depends on them.
• Tier 2, fix this quarter: Answer summaries on top pages, fact density on commercial content, heading rewrites, entity cleanup on major profiles.
• Tier 3, run continuously: Third-party placement, author authority, prompt monitoring, competitor benchmarking.

Here is why order matters. A B2B services client came to us with 18 months of content and no AI citations at all. More content was not the fix. We corrected crawler access, added server-rendered answer blocks to existing service pages, and cleaned up entity errors on third-party profiles. Citations followed within a quarter. The content was never the bottleneck.

Frequently asked questions

How long does a GEO audit take?

A focused audit on a 50 to 200-page site takes 15 to 25 hours. Technical checks move fast. The prompt set takes the longest. You run 20 to 50 prompts across four or five engines and log every source. Big sites with many regions or product lines run longer.

Do I need a GEO audit if my SEO is already strong?

Usually, yes, and the audit is shorter. Strong SEO covers most retrieval work, so you often pass checks 1 to 4 as is. The gaps show up later. Answers buried mid-page. Thin fact density. No baseline. No AI referral tracking. Standard SEO reports do not surface any of that.

How often should I re-run it?

Run the full audit once a year or after a migration or redesign. Re-run the prompt set monthly against the same frozen list. Check crawler access every quarter. CDN and security defaults change without notice. Cloudflare shifted its default AI traffic handling in September 2026, and that is exactly the kind of change that alters eligibility in silence.

Is GEO separate from SEO, or the same thing?

Google says AEO and GEO are still SEO. Its AI features run on the same ranking systems, so that holds for Google Search. It is incomplete elsewhere. ChatGPT, Perplexity, and Claude run their own crawlers, their own retrieval, and their own source preferences. Treat GEO as an extra surface with extra checks, built on the same base.

How soon do fixes show up in AI answers?

Technical fixes land fast. Unblock a crawler and it can register within days once the engine returns. Content restructuring usually takes four to eight weeks. Entity and authority work takes a quarter or more, since it waits on third-party sources. Anyone promising faster is guessing.

Start with the diagnosis

The brands winning AI citations in 2026 do not publish more than everyone else. They fixed the plumbing first. Then they wrote fewer, better pages, with the answer up top and real numbers under it.

Work through these 15 checks in order. Most teams find their problem in the first four. That is good news. Those are the cheapest ones to fix.

At White Bunnie, we run this audit first on every AI SEO engagement because the findings decide where the budget goes. If your organic numbers look stable but the pipeline from search has thinned, the audit usually explains the gap.

Neha Garg

Neha founded White Bunnie in 2018. The agency specializes in SEO, AEO, and GEO for B2B IT services, SaaS, and ITES companies targeting global markets. Their work has helped brands move from zero visibility to consistent AI citations across various popular AI platforms like ChatGPT, Perplexity, and Gemini.