How do I see how often AI recommends me vs. competitors?

To see how often AI recommends you versus your competitors, ask ChatGPT, Gemini and Perplexity the same buyer questions, each in a few different wordings, count every business each answer names and every site it cites, and repeat the same set on a schedule, counting a change only once it holds for 2 weeks.

We did this for our own site on Sep 30, 2026. The result is below, with the other tools’ names hidden.

What does “how often AI recommends you versus competitors” measure?

It measures the share of answers, to a fixed set of buyer questions, that name each business. Some teams call this AI share of voice. We keep to the plain version: for every answer, who was named, and who was cited.

Those are two separate counts. Named means the business appears in the answer text. Cited means its site is listed as a source. A business can be one without the other. What it doesn’t measure is website traffic or your position in search results, and our score doesn’t either.

Here is our own example. We asked 9 buyer questions of our own, each asked 3 ways, on ChatGPT, Gemini and Perplexity, and 78 answers came back. Answersky was named in 0 of 78. The answers named 7 other tools that track AI visibility, and the most-named of them appeared in at least 29 of the 78. The record is in our log. If you want the same view for your own business, the free Quick Scan on our home page shows who AI named instead of you.

Which questions should you ask to compare yourself with competitors?

Ask the questions a buyer types before they have a shortlist, never your own brand name. A question with your name in it only tests whether the assistant recognises you; it says nothing about whether you come up while someone is still choosing.

Three kinds work well, shown here with a made-up example for a project management tool:

• “Best” questions: “What is the best project management tool for a 20-person agency?”
• “Which do teams like mine use” questions: “Which project management tools do small agencies use?”
• Alternative questions: “What’s an alternative to [a competitor you already know]?”

Then fix the set and keep it fixed. If you change a question, you are no longer comparing like with like, so start a new sheet. For the by-hand version of asking in a fresh chat and writing down what comes back, see how to check if ChatGPT mentions your business.

Why ask each question several ways, on each assistant?

Because the wording and the assistant both change who is named and which sites are cited. An academic study of AI search, Chen et al. in 2025, found that AI search services “differ significantly from each other in their domain diversity, freshness, cross-language stability, and sensitivity to phrasing.”

Our own data shows the same thing. For one question, “What is the best AI visibility tool for B2B sales and marketing teams?”, ChatGPT’s 3 wordings cited 12 different sites between them, and no site was cited in all 3. Perplexity’s 3 wordings cited 46 different sites, again none in all 3. Across all 54 ChatGPT and Perplexity answers, Perplexity’s answers cited 17 different sites on average and ChatGPT’s 2.5. These cited-site numbers cover ChatGPT and Perplexity only: Gemini returns its sources as redirect links that our scan doesn’t read yet.

So we ask each question in 3 different wordings, once each, as how a scan works explains. One wording is one view; 3 are a steadier picture.

Should you count who is named or who is cited?

Count both, and keep them apart. Named is what your buyer reads. Cited shows which pages the answer drew on, which is often where a competitor’s name comes from.

The same Chen et al. study found that, in the answers it tested, AI search leaned heavily on third-party sources such as reviews and press over companies’ own sites. In our scan, answersky.com was cited in 0 of the 54 ChatGPT and Perplexity answers. For the kinds of sites that were cited, see which websites AI cites.

One practical line: count names by reading each answer, not only by searching for a name. A business whose name is also an ordinary word, or that is written without its usual suffix, is easy to miss.

Can Google Search Console or Bing show how you compare with competitors?

No. Both show your own site only, and neither shows who was named instead of you.

Google’s page on AI features says sites appearing in AI Overviews and AI Mode “are included in the overall search traffic in Search Console”, reported in the Performance report under the Web search type. You can’t separate AI traffic out, and you can’t see anyone else’s. Bing’s AI Performance report, in public preview since Feb 2026, shows “citation counts for specific URLs from your site” across Copilot and Bing’s AI summaries.

Both are free and worth turning on. But who is named instead of you only comes from asking the questions.

How do you set up AI visibility monitoring against competitors?

Set it up as a sheet with one row per question, wording and assistant, and the same columns every time:

1. Date
2. Assistant
3. Question
4. Wording (1, 2 or 3)
5. Named you? (yes or no)
6. Which competitors were named
7. Your site cited? (yes or no)
8. Which competitor sites were cited (sites the competitors own)
9. Which third-party sources were cited (review sites, trade blogs, press)

Count each business once per answer, even if the answer names it several times, and divide by the usable answers in your fixed set. Record any missing or empty answer and leave it out of that denominator. Being named is not the same as being recommended, so read how the answer presents you. In our own scan, 3 of our 81 calls came back empty (all from Gemini), so our base is 78.

Ask on the same weekday each time, each question in a fresh chat. We ask the models behind ChatGPT, Gemini and Perplexity via their APIs with web search, and the apps can answer differently, so pick one way and stay with it. Check weekly while you are changing things, monthly otherwise. Save the answer text, not only your tick marks, so you can recheck a row later.

If you would rather a tool did this, how to choose an AI visibility tool is a checklist to hold any of them against, ours included.

When does a change in AI visibility count?

A change counts only once it holds. One answer is noise and one week is a hint. Our own rule, on our log, is that a change counts only after it holds 2 weeks, and on how it works that a change counts when 2 scans agree.

So keep the questions and wordings fixed, add dated rows, and wait for the second check before you act on a difference. Be plain about what you’ve seen, too: a change you observe doesn’t prove that your edit caused it, because answers also move on their own. We don’t promise any result, and no one controls what an assistant says.

Is AI visibility monitoring different for marketing agencies?

The method is the same, repeated for each client: a fixed question set for the client’s category, each question asked 3 ways, named and cited counted separately, and the raw answers kept as evidence.

What changes is the reporting. Show a client the dated answers and who was named instead, not a promised position. If you are an agency, keep each client’s sheet separate and note the date and assistants on every page, so a number can always be traced back to the answers it came from.

To start, write a short list of buyer questions, ask each 3 ways on each assistant, fill in the first dated sheet, and put the next check in your calendar. That first sheet is your baseline. For a first look without the sheet, the free Quick Scan does a version of this for you.

Free Quick Scan: 6 of your buyers’ questions on ChatGPT and Perplexity, 12 answers, and who AI named instead of you. Report within 3 business days.