How to check whether AI assistants mention your brand

When someone asks an assistant which product to use for a job, the answer names a few. Those answers don't show up in your analytics, and a rank tracker can't follow them. What you can do is sample them: count how often your brand comes up across a set of answers, and read what the assistants say on the questions your buyers ask.

For Google's own AI features there is one measured signal as well. Google says sites appearing in AI Overviews and AI Mode are included in the overall search traffic in Search Console, in the Performance report. So clicks from them are already in the numbers get_search_console_performance reads, mixed in with the rest.

get_brand_visibility takes a brand, product or domain and up to 9 competitors. It returns total mentions in ChatGPT and in Google's AI Overviews, a share-of-voice table for the set, and up to 20 cited sources, the top 10 from each platform, ranked by the AI search volume behind them.

Check how often ChatGPT and Google's AI Overviews mention [brand] compared with [competitor 1], [competitor 2] and [competitor 3]. Show share of voice, and list the top 10 cited sources.

Without competitors you get a raw mention count and no share of voice, and a raw count is hard to judge on its own. With them you can see whether you are a twentieth of the conversation or half of it.

To see the kind of result before connecting an agent, the free AI visibility checker runs one brand lookup with no account.

The counts come from a third-party data set of assistant answers, not from every conversation anyone has had. Treat them like search volume: useful for comparing brands with each other and for watching a trend, not as a count of real people. ChatGPT data covers the United States in English only.

The sources list is the most useful part of the result. These are the pages the answers draw on for your category: review sites, comparison articles, forum threads, documentation, and sometimes a competitor's blog. Ask four questions of each:

• Is your brand on the page at all?
• If it is, is what the page says accurate and current?
• If it isn't, is there an honest way on? A comparison that takes submissions, a directory you qualify for, a thread where you can answer a real question.
• If it's your own page, does it answer the question as directly as the pages beside it?

For the 10 most-cited sources, tell me which are about our category and whether [brand] appears on each. Use your own web search to read the pages and say that's where the answer came from. Then list the ones we could reasonably get onto, and how.

IndexZero returns the list of sources but doesn't fetch other sites' pages, so the agent reads them with its own browsing. The prompt asks it to say so, which keeps the IndexZero data and the agent's reading apart.

run_ai_prompt sends one prompt to any of ChatGPT, Claude, Gemini and Perplexity and returns each answer with its citations. highlightBrand marks your brand wherever it appears. Web search is on by default, which is closer to how people use these assistants, and webSearchCountryCode localizes it. A model that fails is reported as failed without hiding the others.

Write prompts the way a buyer would, not the way you'd describe yourself. "What's the best invoicing tool for a two-person design studio?" tells you something. "Is [brand] good?" mostly doesn't.

Ask ChatGPT, Claude, Gemini and Perplexity: "What's the best [category] tool for [customer type]?" Highlight [brand]. Tell me who each one recommends, whether we're mentioned, and which sources each one cited.

Pick five to ten questions that match how your customers describe the problem, and keep them fixed so readings compare over time. Prompts don't re-run on a schedule; you or the agent run them when you want a reading. Every brand lookup and prompt run is saved on the project, and list_ai_visibility_runs and get_ai_visibility_run reopen past runs for free, so the history builds as you go and a question asked last week doesn't have to be bought again.

The assistants will disagree, and the disagreement tells you what to do: an assistant that recommends three competitors and not you, citing a comparison page you aren't on, has told you what to do next.

Google is direct about its own AI features: "There are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary." A page has to be indexed and eligible to show in Search with a snippet, and "You don't need to create new machine readable files, AI text files, or markup to appear in these features" (Google Search Central).

That leaves ordinary work, aimed by what the measurements showed:

• Pages that answer your buyers' questions in plain language, on URLs Google has indexed. With Search Console connected, inspect_urls checks index status for up to 10 URLs at a time, at no cost.
• An accurate presence on the sources the answers cite.
• Crawler access. Assistants that search the web have to be able to fetch your pages, and OpenAI says that sites opted out of OAI-SearchBot won't be shown in ChatGPT search answers, though they can still appear as navigational links. The AI crawler checklist separates search crawlers from training crawlers, so you can block training without dropping out of answers, and the free robots.txt tester shows what your live file allows each one.
• A site that isn't broken where it matters. See how to run a site audit with an AI agent.

Then measure again with the same competitors and the same prompts. Compare each reading with the last one, which the saved runs make free; a single reading says little on its own.

Both tools cost credits and need Starter ($29 a month) or above. On the free plan they return a plan-upgrade error. run_ai_prompt charges per model selected, so one model costs less than four, and its answers are cached for a week: re-running the same prompt with the same models and settings inside that window costs nothing. Reopening a saved run with get_ai_visibility_run is always free.

AI visibility is one of four workflows in SEO with AI agents, which also covers keeping an agent's spending in check and checking what it reports.