AI visibility audit: a checklist you can run in one afternoon

An AI visibility audit checks whether ChatGPT, Perplexity, Gemini, Claude and Google AI Overviews name your product when buyers ask about your category. You run 15 to 25 real buyer prompts, three times per engine, and log who is named and which pages are cited. Then you check crawler access, indexing, your content and the lists AI quotes.

The first run takes about an afternoon with free accounts and a spreadsheet. Each later run takes less, because the prompts stay the same. People also call this a GEO audit or an AEO audit: the steps are the same. For the bigger picture, read our LLM SEO guide and SEO vs GEO vs AEO.

What do you need before you start?

• A spreadsheet with the columns from step 3 below.
• Accounts on ChatGPT, Perplexity, Gemini and Claude. Free plans are enough for a first run.
• A private or logged-out browser window for Google, so AI Overviews are not shaped by your own search history.
• Access to Google Search Console and Bing Webmaster Tools for your site.
• A list of 3 to 5 competitors buyers compare you with.

Step 1: which prompts should you test?

Test the questions a buyer asks before they know your name. Write 15 to 25 prompts in total, in the words your buyers use, and save them. You will reuse the exact same set every month.

Step 2: how do you run the prompts in each engine?

Run each prompt three times per engine, each time in a new chat. AI answers change from run to run, so one run is a sample, not a result. In Prefer's study of 960 answers (measured 13 September 2026), two runs of the same question shared on average only 37% of their cited sites in ChatGPT and 35% in Gemini. Perplexity was the most stable at 96%, Claude was at 68%.

1. ChatGPT: start a new chat for every run and turn search on, so the answer shows its sources.
2. Perplexity: use the default search. Every answer lists its sources.
3. Gemini: start a new chat and open the sources under the answer when it shows them.
4. Claude: turn web search on, otherwise the answer comes from training data only and has no sources.
5. Google AI Overviews: search the prompt in a private window and note whether an overview appears at all. Not every search shows one.

Step 3: how do you record mentions and citations?

Use one row per prompt, engine and run. With 20 prompts, five engines and three runs, that is 300 rows. Copy these columns into your sheet:

At the end, count three numbers: the share of runs that named you, the share that named each competitor, and the share that cited your own site. These are your baseline for next month.

Step 4: which cited sources should you list?

List every URL the engines cited, then sort them by how often they appear. Most of those pages are not yours. In a Foundation and AirOps study of 5.1 million AI responses for 50 B2B brands, only 2.24% of citations on unbranded questions went to the brand's own pages. In Prefer's July 2026 data, 40% of sources for buyer questions were independent lists.

• Independent "best X" list articles and roundups.
• Review and comparison sites.
• Reddit threads and other forums.
• LinkedIn posts and articles.
• Your own pages: which ones, and for which prompts.

Mark whether each source names you, names competitors, and is a page you could get onto. Sources cited by several engines come first.

Step 5: which technical checks matter?

Robots.txt: are AI crawlers allowed?

Open yoursite.com/robots.txt and check the search bots below. Each company separates search from training, so you can allow search and still decide on training. Blocking a search bot can keep you out of that engine's answers.

Also check your firewall or CDN. Some bot protection settings block AI crawlers even when robots.txt allows them. OpenAI also publishes the IP ranges of OAI-SearchBot, so you can allow them.

Is your site indexed in Google and Bing?

1. In Google Search Console, inspect your 10 most important URLs with the URL Inspection tool. Each should say it is on Google. Google says pages need to be indexed and eligible for a snippet to appear in AI Overviews and AI Mode, and that this traffic is reported in the Performance report under the Web search type.
2. In Bing Webmaster Tools, check the same URLs and submit your sitemap. Bing's AI Performance report (public preview since 10 February 2026) shows how often your pages are cited in Microsoft Copilot and Bing's AI summaries.
3. Set up IndexNow so Bing and other participating engines hear about new and changed pages right away.

Structured data and llms.txt

Google says there is no special structured data needed for AI features, and no need for AI text files. It does ask that structured data matches the visible page. So check your Organization, Product and Article markup with the Rich Results Test for errors, not for extra tricks. An llms.txt file is a proposed Markdown file at your site root that points AI agents to your key pages. It is cheap to add and keep accurate, but treat it as hygiene, not a lever.

Step 6: which content checks should you run?

• Does your homepage say in one plain sentence what you sell and who it is for?
• Is there a page for each main use case, answering the question in the first 40 to 60 words?
• Do you have honest comparison and alternatives pages for the competitors that come up in the answers?
• Are key facts in HTML text, not only in images, videos or scripts?
• Is your company described the same way on your site, LinkedIn, review sites and directories?
• When an answer described you wrongly, which page did it cite? Fix that page, or ask its owner to.

Step 7: which third-party lists should you check?

Take the "best X" articles and roundups from step 4 and check each one. Note whether you are listed, at which position, with a correct description, and whether the list shows when it was last updated. Then sort them into three groups: lists you are on, lists you could earn a spot on, and lists that sell spots.

1. Lists you are on: check the description and the link. Send the editor a correction if something is out of date.
2. Lists you could earn: email the editor with a short, factual pitch. Say what you do, who uses it and why it fits the list. Many editors add tools when the pitch is useful.
3. Lists that sell spots: ask how the spot is labeled, how long it stays live and what you pay. Look at our listicle placement price index and the listicle placements guide before you agree to anything.

If you decide to pay, Rank Monthly is one option: it books brands into list articles from publishers who joined, and you pay monthly only while the spot is live. It does not track AI answers, and Google and the assistants decide what they cite. The free routes above come first.

Step 8: how do you repeat the audit every month?

1. Run the same prompts on the same engines, three times each, in the same week of the month.
2. Add a new tab per month and compare: share of runs named, share per competitor, share citing your site.
3. Compare the cited sources list. A new source that names competitors and not you is a work item.
4. Check Search Console and Bing's AI Performance report for the same period.
5. Change the prompt set at most once a quarter, and note the change, or the months stop being comparable.

Expect noise: one prompt moving from 1 in 3 runs to 2 in 3 can be chance. Look for trends across many prompts and months.

Should you use an AI visibility tool or do it by hand?

Do the first audit by hand. Move to a tool when the manual run takes too long or you need daily data. For example, Otterly.AI starts at $29 a month for 15 prompts on ChatGPT, AI Overviews, Perplexity and Copilot, and Peec AI starts at $95 a month for 50 prompts run daily, with 3 engines active at once (both checked 1 November 2026). Tools save time and show trends, but they still sample answers and they do not change what AI cites. We compare ten trackers in the best AI visibility tools and more in the best AEO tools.

The full checklist (copy this)

1. Write 15 to 25 buyer prompts and save them as a fixed set.
2. Run each prompt 3 times in ChatGPT, Perplexity, Gemini, Claude and Google AI Overviews, each in a new chat.
3. Log named, position, described correctly, linked, competitors and cited sources for every run.
4. Count your share of runs named, each competitor's share and your share of citations.
5. List every cited source and sort by how often it appears.
6. Check robots.txt for OAI-SearchBot, Claude-SearchBot, PerplexityBot and Googlebot, and your firewall.
7. Inspect your top 10 URLs in Google Search Console and Bing Webmaster Tools, and submit sitemaps.
8. Set up IndexNow.
9. Test your structured data for errors and keep llms.txt accurate.
10. Check your key pages: one-line description, answer first, comparison pages, facts in text.
11. Check each cited list: on it, could earn it, or sells spots. Act on each group.
12. Repeat monthly with the same prompts and compare.