GEO audit
GEO Audit
Led growth at cnvrg.io before its acquisition by Intel.
A GEO audit reviews page readiness and records how selected AI-search engines answer agreed buyer questions. It separates verified technical findings from answer observations and unexplained absences.
Your buyers stopped scrolling through ten blue links. They ask ChatGPT, Perplexity, Gemini and Google AI Overviews a question and read the answer. That answer cites three or four sources. A GEO audit tells you whether you are one of them.
GEO, generative engine optimization, is the practice of getting an AI engine to retrieve, trust, and cite your pages inside its answer. It is not the same as classic SEO. A page can rank on page one of Google and still be invisible to an AI engine, because the model decides what to quote on a different set of signals: structured facts, entity clarity, source authority, and machine-readable access.
The historical June 23, 2026 report scored 61 SaaS and AI sites using the v1.12.0 AI Search Readiness rubric; 56% scored 60 or below. That is a dated readiness-rubric result, not observed AI citation share, recommendations, referral traffic or revenue.
The seven dimensions
What a GEO audit actually checks
Citability: can a model lift a clean answer from your page?
A GEO audit reviews whether key pages present clear answers and supporting evidence that readers can find. Answer-first sections and relevant facts can be useful editorial choices. A number in the first 60 words is a heuristic, not a documented Google AI requirement or a proven highest-impact citation fix.
Crawler access: can the AI bots reach you at all?
The audit checks relevant crawler access and robots.txt product controls separately. It flags technical access blocks or content unavailable in initial HTML. Google-Extended is a product control token with no separate HTTP request user agent; it does not control Google Search inclusion or ranking. Passing these checks does not establish that an engine will retrieve or cite a page.
Structured data: are your facts machine-readable?
The audit checks whether Organization, Person, Service, FAQPage and Article schema describes the visible page accurately and whether existing sameAs links are consistent. Contradictory claims need review, but repeated @id values are not automatically errors. Google does not require special schema or speakable markup for AI Overviews or AI Mode; structured data alone does not establish that a page will be retrieved or cited.
llms.txt and machine access surface
An llms.txt file is an optional way to publish a selected content map. The audit checks whether an existing file is current and points to the intended pages. Mine is public at /llms.txt as a working reference. Google does not require AI text files for AI Overviews or AI Mode; this file alone does not establish crawler access, indexing or citation.
Entity presence: does the model know you exist?
The audit checks that existing references to your company are consistent across Wikidata, Crunchbase, LinkedIn and relevant knowledge-graph entries. This is an identity-consistency check, not a requirement to create a listing on every platform or evidence that an AI engine will prefer your brand. Missing third-party profiles do not, by themselves, explain why a competitor was cited.
Source authority and corroboration
The audit maps which key claims are supported by primary sources, third-party citations or original data, and flags assertions that need evidence. Original data can give readers something useful to evaluate when its method and limitations are clear. It is not a proven universal citation-ranking factor, and missing evidence does not establish how a specific model selected its sources.
Live citation test across five engines
The audit runs your real buyer queries through ChatGPT search, Perplexity, Gemini, Google AI Overviews and Bing Copilot, and records who gets cited today. That is your baseline. See the GEO consultant overview for how the ongoing work moves that baseline.
Scored on the same rubric as the State of AI Search benchmark
The GEO audit does not score on opinion. It uses the rubric I built for the State of AI Search 2026 benchmark, the same one I ran across 61 SaaS sites.
The historical June 23, 2026 report scored 61 SaaS and AI sites using the v1.12.0 AI Search Readiness rubric; 56% scored 60 or below. That is a dated readiness-rubric result, not observed AI citation share, recommendations, referral traffic or revenue.
The output is not a vanity grade. Each verified finding maps to a prioritized fix, using buyer relevance, dependencies and estimated implementation effort. You get the score, the gap, and the build order. From there the work runs as the Citation Loop, the four-line delivery system behind GEO services: the audit is the entry point, not the product.
GEO audit deliverables
A named readiness score. The 0–100 score is based on the named readiness rubric, with version, date, checked inputs and limitations. It is not observed citation share or a forecast.
A per-dimension readiness report. The report records the inspected pages, evidence and limitations; it does not assign a proven citation loss to each missing check.
A recorded answer baseline. For an illustrative record, a buyer question answered on one recorded engine links to your page: mark a source citation. A brand named without a supporting link is a mention. A failed run is unavailable, not zero visibility. Repeat comparable question-and-engine sets, state the observed-run denominator, and keep referrals and qualified outcomes separate.
A prioritized fix list. Priorities follow the documented buyer problem, supporting evidence, dependencies and estimated implementation effort.
An optional Weekly AI Answer Audit. The same query set re-run every week across all five engines, each answer checked for wrong pricing, wrong facts and competitor leakage. Most clients add it after seeing what the engines already get wrong about them.
The audit pairs with the technical side of the work. If your tracking cannot yet prove which channel drives revenue, start with server-side tracking so the GEO wins are measurable. To see the operator track record behind the rubric, read the case studies.
Evaluating consultants instead of running this yourself? Start with how to vet a GEO expert: a five-check rubric that works on anyone, including me.
For an instant self-serve check before the full audit, the free growth leak audit scores AI crawler access, citability, schema and llms.txt in 30 seconds.
FAQ
Frequently asked questions about a GEO audit
What is a GEO audit?
The audit reports a named, versioned readiness rubric separately from live answer observations. For each agreed buyer question, record the engine and mode, capture date, market and language, answer, brand mention and linked source URL. Failed or unavailable runs remain separate from observed absence. The output includes the readiness findings, the recorded answer sample and a scoped implementation order.
How is a GEO audit different from an SEO audit?
An SEO audit asks whether you rank in Google's blue links. A GEO audit asks whether an AI engine quotes you inside its answer. They overlap on technical health but diverge on what matters: GEO weights machine-readable facts, entity clarity, citable answer blocks, and direct crawler access over keyword density and backlinks. A page can rank on page one and still be invisible to ChatGPT.
Which AI engines does the audit test?
ChatGPT search, Perplexity, Google Gemini, Google AI Overviews, and Bing Copilot. The audit runs your real buyer queries through each one and records who gets cited today, which becomes the baseline you measure improvement against.
How long does a GEO audit take and what do I do with it?
I personally inspect the agreed pages, technical access, visible entity facts and recorded answer sample. You receive the named readiness rubric, evidence, gap report and prioritized build order. This hands-on investigation is part of the audit; we separately agree which implementation changes belong in the scope. Advisory is the sole advice-only offer. Delivery and later citation observations are reviewed separately.
What is the State of AI Search 2026 benchmark?
The historical June 23, 2026 report scored 61 SaaS and AI sites using the v1.12.0 AI Search Readiness rubric; 56% scored 60 or below. That is a dated readiness-rubric result, not observed AI citation share, recommendations, referral traffic or revenue.
Does passing the audit checks mean AI engines will cite my site?
No. Passing the crawler access checks does not establish that an engine will retrieve or cite a page, and structured data alone does not establish that either. That is why the audit also runs a live citation test, which records who gets cited today across five engines and becomes your baseline.
Do I need an llms.txt file or special schema to appear in Google AI Overviews?
No. Google does not require special schema or speakable markup for AI Overviews or AI Mode, and it does not require AI text files for them either. An llms.txt file is an optional way to publish a selected content map. If you already have one, the audit checks whether it is current and points to the intended pages.
Do I need profiles on Wikidata, Crunchbase and LinkedIn before the audit?
No. The entity check looks at whether existing references to your company are consistent across Wikidata, Crunchbase, LinkedIn and relevant knowledge-graph entries. It is not a requirement to create a listing on every platform. Missing third-party profiles do not, by themselves, explain why a competitor was cited.
What is the optional Weekly AI Answer Audit?
It is an optional add-on to the GEO audit. It re-runs the same query set every week across all five engines, and each answer is checked for wrong pricing, wrong facts and competitor leakage. Most clients add it after seeing what the engines already get wrong about them.
Is there a quick way to check my AI visibility before the full audit?
Yes. For an instant self-serve check before the full audit, the free growth leak audit scores AI crawler access, citability, schema and llms.txt in 30 seconds. The full audit then runs your real buyer queries through ChatGPT search, Perplexity, Gemini, Google AI Overviews and Bing Copilot.
Next step
Find out if AI is citing you or your competitor
The GEO audit gives you the score, the gaps, and the build order. The window is open now because almost no one in your category has done the work. Book the audit and find your number.
“SEO asked whether Google ranks you. GEO asks whether the AI even mentions you, and most audits still don’t check the second question.”
Yaniv Goldenberg
Sources: Google Search: helpful content, Pew Research: AI.
Work with Yaniv
Fractional Head of Growth for B2B SaaS, B2C, and e-commerce. Revenue that compounds, not spikes. Any channel, any motion, any stage.
Explore the Generative Engine Optimization hub