GEO audit and AI visibility audit checklist: what should it include?

What is a GEO audit, and how is it different from an SEO audit?

A GEO audit is a check of how AI engines see, describe and recommend your brand when buyers ask them what to buy. An SEO audit asks whether Google can crawl and rank your pages. A GEO audit asks the question that now decides the sale: when a shopper asks ChatGPT for the best product in your category, are you in the answer, are you first, and which websites convinced the engine to name a rival instead?

The two overlap more than GEO sellers admit. Google says the SEO best practices still apply to AI Overviews and AI Mode, and that there are no additional requirements to appear in them (Google Search Central). What changes is the scoreboard. A GEO audit measures outputs an SEO audit never looks at: named brands, cited sources, share of voice and AI-referred orders.

What are the 9 checks in a GEO audit?

Every check ends in the same place: a gap you can see and a fix you can ship. If an audit you are quoted skips one, ask why.

1. Prompt baseline: are you named, and how often?

Write 30 to 50 prompts your buyers actually ask: category prompts (best magnesium for sleep), problem prompts (why am I bloated after every meal), comparison prompts (your brand vs your top rival) and brand prompts (is your brand legit). Run each one several times in fresh sessions on ChatGPT, Perplexity, Gemini and Google AI Overviews. Record whether you are named, your position, how you are described and which sources are cited.

Repetition is not optional. AI answers change between runs, and one supplement testing firm told New Hope Network that only dozens of repetitions give a representative average. One screenshot of ChatGPT naming you proves nothing, and neither does one screenshot of it ignoring you. Our guide to checking if ChatGPT recommends your brand has the full prompt template.

2. Citation source map: who is deciding for the engine?

For every prompt, list the URLs the answer cites. This is the most valuable page in any audit, because it tells you exactly where the recommendation is being won. In our October 2026 tests of a supplement buying question, ChatGPT cited only research sources (the World Gastroenterology Organisation, PubMed and NCCIH), while Perplexity cited Healthline, Forbes, Fortune, Everyday Health and Prevention. Same question, two engines, two completely different battlefields.

The pattern holds across categories: brands get named through pages they do not own. So the output of this check is a target list, the publishers, roundups, Reddit threads and review sites to get into, ranked by how often the engines cite them for your prompts.

3. Brand facts: what does AI say about you?

Ask each engine about your brand directly: what it is, what it costs, what is in it, who it is for, whether it is worth it. Log every wrong fact, every outdated price and every claim you would never make. Watch for look-alike stores and resellers: AI engines sometimes cite copycat domains that use a brand's name, and every answer that sends a buyer there is a sale you paid to create and somebody else collected.

4. Crawler access: can AI engines reach your store?

Check robots.txt, and any firewall in front of your store, for the AI search crawlers. OpenAI says sites that opt out of OAI-SearchBot will not be shown in ChatGPT search answers, though they can still appear as navigational links (OpenAI). Perplexity says PerplexityBot is designed to surface and link websites in its search results and recommends allowing it (Perplexity). Training is a separate switch: OpenAI treats GPTBot and OAI-SearchBot independently, so you can block training and still appear in ChatGPT search.

Robots.txt is only half the check. A CDN or bot firewall can block crawlers your robots.txt allows, and Cloudflare announced on July 1, 2025 that it was changing its default to block AI crawlers (Cloudflare). Then confirm in your logs that the real bots reach you, checked against the IP ranges OpenAI publishes. Most requests claiming to be AI bots in our own server logs were scanners faking the user agent.

5. Rendering: do your facts exist without JavaScript?

View the raw source of your top product pages, not the rendered page, and search for your price, review count, key ingredients and FAQ answers. Anything that only appears after scripts run, or only inside an image, is invisible to any crawler that does not run scripts. The classic miss is an animated counter that reads 0 in the raw HTML. We found and fixed exactly that on our own site in September 2026: crawlers were reading $0 results and a 0.0 rating until we rendered the real values on the server.

6. Page answers: do your pages answer buying questions?

AI engines quote specifics, not slogans. Audit your top product, collection and comparison pages against your prompt list: who the product is for, what is in it and how much, how fast it works, how it compares with the obvious alternatives, what it costs per serving or per month, and what proof backs it. Then check that Product, Organization and FAQ structured data matches what is visible on the page. Our llms.txt and schema guide separates what matters from the noise.

7. Shopping data: are your products in the AI catalogs?

AI shopping results run on product data, not blog posts. OpenAI says Shopify product data is already integrated into ChatGPT through Shopify Catalog, and Shopify's ChatGPT channel has its own eligibility rules: you sell to the US, qualify for Shopify Catalog, accept the agentic terms and complete your store policies (Shopify Help). OpenAI says merchants in ChatGPT shopping results are ranked on signals such as availability, price, quality and whether the merchant is the maker or primary seller, and that the results are not ads (OpenAI Help). The audit checks eligibility, titles, variants, prices and the product facts an engine needs to compare you. Our Shopify ChatGPT checklist covers the setup.

8. Off-site proof: do the sources AI trusts know you exist?

Take the source map from check 2 and score your presence on every cited domain. Are you in the roundup, the Reddit thread, the review site, the publisher piece? For supplements, the 5W Supplements AI Visibility Index reports that third-party certifications such as NSF Certified for Sport, USP Verified and Informed Sport were the strongest observed signal. Off-site proof is where most brands lose, and an audit that does not turn it into a named target list has skipped the hardest part.

9. Measurement: can you see AI revenue?

If you cannot count it, you cannot defend the budget. ChatGPT adds utm_source=chatgpt.com to the links it sends, so organic ChatGPT visits are easy to isolate in GA4 and Shopify. Paid ChatGPT ad clicks carry your own ad UTMs and belong in a separate channel, or they inflate your AI number. Then add the platform reports: Google Search Console now has a Generative AI features report (beta) for impressions in AI features, and Bing Webmaster Tools has an AI Performance report (beta) for citations. Our AI traffic tracking guide has the full setup.

What should you get at the end of a GEO audit?

Not a score. An AI visibility score with no prompt list behind it is a number you cannot check and cannot improve. A real deliverable has four parts:

1. The prompt sheet. Every prompt, every engine, every run: named or not, position, description and cited sources. This is your baseline, and every future audit is measured against it.
2. The source map. The domains and URLs the engines cite for your category, ranked by how often they appear, with your presence on each one.
3. The technical findings. Crawler access, rendering and structured data issues, each with the exact page and the exact fix.
4. A ranked fix list. Every fix ranked by the buying prompts it can win and the revenue behind them, with an owner and a date. Three fixes shipped beat thirty fixes filed.

Read across a row and the fix list writes itself. In the example, the brand wins its own comparison prompt, loses the category prompt to a publisher roundup it is not in, and leaks brand prompts to a reseller and an outdated price.

How much does a GEO audit cost, and is a free one worth it?

Published ecommerce pricing puts a one-time GEO audit at about $2,200 to $5,000, scaled by the number of product categories covered (Mentionable). Our GEO agency cost guide breaks down retainers, tools and what drives the price.

A free audit is worth exactly what it shows you. If it covers your real buying prompts, repeated runs on every engine, a cited-source map and a ranked fix list, it beats most paid decks. If it is a tool screenshot followed by a pitch, walk.

Can you run a GEO audit yourself in 60 minutes?

Yes, a rough one. It will not replace repeated runs or a full source map, but in an hour it tells you whether you have a problem and how big it is.

1. Write the 10 questions your customers ask before they buy, in their words, not your keyword list.
2. Run each one in ChatGPT, Perplexity, Gemini and Google. Note every brand named and every site cited.
3. Ask each engine what your brand is and whether it is worth it. Log every wrong fact and every reseller link.
4. Open yourdomain.com/robots.txt and search for OAI-SearchBot, GPTBot, PerplexityBot and ClaudeBot. A Disallow on a search crawler is a fix for today.
5. View the raw source of your best-selling product page and search for your price, review count and hero ingredient. If they are missing, crawlers that do not run scripts cannot see them.
6. In GA4, filter the last 90 days of sessions by source for chatgpt, perplexity, gemini and copilot. If you have no AI channel, build one this week.
7. Count how many of the sites cited in step 2 mention you at all. That number is your off-site gap, and it is usually the biggest one.

How we run a GEO audit at Succession Media

We run the audit on the first call, free, and we run the same checks on our own site every week. Omar Ghouat, our AI Search Lead, owns the method.

• Weekly AI Answer Lab. A fixed library of buying questions for your category, asked in ChatGPT, Gemini and Perplexity from a US location, scored on named, named first, cited sources and share of voice against your top rivals.
• Verified bot logs. AI crawler hits checked against the IP ranges OpenAI publishes, so spoofed bots never inflate the numbers.
• Revenue, not screenshots. Organic AI referrals split from paid ChatGPT ads and reported as sessions, orders and revenue next to your ad numbers.
• Fixes shipped, not filed. Product page answers, structured data, rendering and robots fixes go live, and the off-site target list gets worked every week.

We hold our own domain to the same standard. In the 10 days after our own GEO pass in September 2026, verified ChatGPT-User requests to succession.media went from under 2 a day to about 37 a day, and in October Google's AI Overview cited our guide for the question "what's the best GEO agency for DTC supplements". At one brand we work with, ChatGPT traffic converted at 4.99%. AI search is a revenue channel. Audit it like one.

FAQ