GEO audit: a 10-point checklist and a free instant check

What a GEO audit is

A GEO (generative engine optimization) audit answers one question: when a buyer asks ChatGPT, Perplexity, Gemini or Google AI Mode for a recommendation in your market, are you named, and if not, who is and which pages do the engines cite? It is the AI-search version of a rank check, except the output is a sentence, not ten blue links, so you measure mentions, position in the answer and cited sources.

Most brands fail it. In our study of 198 agencies (10 buyer questions × 4 engines each, 7,910 answers), 53% were not named once, and on average only 5.5% of an agency’s answers named it.

The 10-point GEO audit checklist

1. Mention rate. Share of answers that name the brand, for 10 or more real buyer questions, per engine. This is the headline number.
2. Position. How often the brand is named first, not just somewhere in the list.
3. Engine spread. Check all four engines. In our data 41% of the agencies that were named at all were named by only one engine, so a ChatGPT-only check misses most mentions.
4. Competitors named instead. The brands that appear when yours does not. This is your real competitor set in AI search.
5. Cited sources. The domains and pages the engines link. In our data 44% of titled citations were list pages ("best X", "top Y"); Perplexity cites about 8.9 domains per answer.
6. Own-site citation rate. How often your own domain is cited. Zero with high competitor citations means your pages do not answer the question the engines are asked.
7. Brand recognition. Ask each engine about the brand directly. Known, thin or unknown decides whether you need entity work (consistent name, profiles, directory listings) or content work.
8. Crawler access. Check robots.txt does not block the AI crawlers you want: OAI-SearchBot and ChatGPT-User (OpenAI), PerplexityBot, Google-Extended and ClaudeBot. GPTBot is OpenAI's training crawler; the search and user bots are what feed live answers, so blocking them can remove you from answers. Google-Extended controls Gemini's use of your content; AI Overviews follow Googlebot.
9. Answer-shaped pages. One page per buyer question, with the answer in the first paragraph, a comparison table, a price and a date. Engines quote pages that state facts plainly.
10. Fix list and re-test. Rank the gaps (missing listings, missing pages, thin entity) and re-run the same questions weekly. Answers vary run to run, so watch the trend over many questions.

Run a GEO audit three ways

Doing it by hand also works: write 10 buyer questions, ask each engine, log who is named and which links appear. It takes about two hours per brand and is hard to repeat weekly, which is why agencies automate it.

Patterns to look for

• The brand is named for its own name but never for the category question ("best X for Y").
• Competitors are cited from the same two or three directories and list posts. Getting onto those lists is often the fastest fix, though we have not tested that it changes recommendations.
• The engines disagree. In our sample Gemini names no business at all in about 30% of answers, while ChatGPT and Perplexity name about six per answer.
• Pages answer the keyword but not the question, so the engine cites a third-party summary instead.

FAQ