How to check your AI visibility each month, without paying for a tool

Why does a single check mislead?

Because the answer changes almost every time you ask. In a study by SparkToro, 600 volunteers ran 12 questions through ChatGPT, Claude and Google’s AI 2,961 times in November and December 2025. They found less than a 1 in 100 chance that ChatGPT or Google’s AI would give the same list of brands twice in 100 tries, and less than 1 in 1,000 that two lists would come in the same order (SparkToro, 27 Jan 2026). If ChatGPT named you on Tuesday, it may leave you out on Wednesday.

Under the noise sat something steadier. Across 994 answers to 142 differently worded questions about headphones, the leading brands showed up in 55 to 77% of them. Rand Fishkin, who led the study, now counts the share of runs that name a brand, across many questions each run several times, as a reasonable measure, and dismisses any tool that sells a ranking position in AI. Fishkin runs SparkToro, which sells audience research software, and his research partner Patrick O’Donnell works at Gumshoe.ai, which sells AI visibility tracking. The report names that conflict itself.

Our guide to why AI assistants don’t recommend your business explains how assistants pick names.

What goes in the spreadsheet each month?

1. Fix your list of questions

   Write 10 to 20 questions the way buyers ask you on calls, on WhatsApp and by email. Include the city and the kind of customer, and leave your business name out: “Who installs solar panels for factories in Sharjah?” Keep the wording the same every month, because a reworded question starts its history again, and add new questions at the bottom. SparkToro found people word the same need very differently and the assistants still caught the intent (SparkToro), so keep your buyers’ own words.

2. Run each one three times in four places

   Ask in ChatGPT, Gemini and Perplexity, and search on Google to see whether an AI Overview appears. Spread the runs over different days, in a private window or logged out where you can: Gemini, for one, can use memory of your past chats to personalize its answers (Gemini Apps Help, Get personalization with memory of your past Gemini chats). Google designs AI Overviews to appear where they add something to the usual results (Google Search Central, AI features and your website), so write “no Overview” when there isn’t one. Fifteen questions make 180 runs a month; split them over a few sittings.

3. Record three things per run

   Whether you were named, which competitors were, and which sites the answer cited. ChatGPT answers that searched the web may carry citations and a Sources button (OpenAI Help Center, Searching the web with ChatGPT). Gemini has a Sources button, though some answers have none (Gemini Apps Help, View related sources). Perplexity numbers its citations in every answer (Perplexity Help Center, How does Perplexity work?), and AI Overviews show links to relevant pages (Google Search Central).

At the end of the month, divide the runs that named you by all the runs that produced an answer, for each assistant and overall. Do the same for each rival. That’s your share, and theirs.

What counts as progress?

A share of runs that rises and holds, read per assistant and next to your rivals’ shares. The sources column tells you why it moved: when your own service pages, or the directories and articles that list you, start appearing as cited sources, the assistants are finding you.

Be modest about small moves. SparkToro says that to know an assistant’s set of recommendations you usually need to ask at least 60 to 100 times (SparkToro). Three runs a month is far fewer, so read your sheet for direction. Our rule is to trust a change once it lasts two months running and shows in more than one assistant, and never to report a rank.

In our view, a paid tool earns its fee once you need hundreds of questions run many times. Before you buy, ask it to show its math, as SparkToro advises.

Which free reports show AI visibility and traffic?

Three, each seeing a different slice. Copy their monthly figures onto a second tab.

• Google Search Console’s Generative AI performance report, open to all sites since 31 August 2026, shows how often links to your site appeared in AI Overviews and AI Mode, by page, country, device and date. It counts impressions only, without the searches, and may not appear until your site has enough of them (Search Console Help, Generative AI performance report (Search); Google Search Central Blog, 3 Jun 2026). Clicks from those features sit in the ordinary Performance report, under the “Web” search type (Google Search Central, AI features and your website).
• Google Analytics 4 added an AI Assistant channel on 13 May 2026 (What’s new in Google Analytics). It groups visits from sources such as ChatGPT, Gemini, Deepseek, Copilot and Grok, and leaves out Google’s AI Overviews and AI Mode (Analytics Help, Default channel group). OpenAI says ChatGPT adds utm_source=chatgpt.com to the links it sends (OpenAI Help Center, Publishers and Developers FAQ). Google’s examples don’t name Perplexity, so check the Referral channel for perplexity.ai too.
• Bing Webmaster Tools’ AI Performance report, in public preview since 10 February 2026, counts how often your pages are cited in Microsoft Copilot and Bing’s AI summaries, which pages, and a sample of the phrases the AI searched with (Bing Webmaster Blog, 10 Feb 2026).

Why ask buyers how they heard about you?

Because some AI answers never send a click. Gemini says some of its responses carry no sources at all (Gemini Apps Help, View related sources). A buyer who reads your name there may later search for you or message you on WhatsApp, and your analytics can’t tie that to the assistant.

So add “An AI assistant, such as ChatGPT or Gemini” to the “How did you hear about us?” question on your forms, in your call script and in your first WhatsApp reply, and give it a column in the sheet. When those answers rise in the same months as your share of runs, you can trust the change more.