TL;DR

For two years, “AI visibility” mostly meant “do we show up in ChatGPT.” That shortcut is getting expensive. Comscore’s Q2 2026 AI Intelligence Report, covered by Digiday, shows the market splitting fast.

Share of AI assistant prompt volume (Comscore)
AssistantJanuary 2026June 2026
ChatGPT70%50%
Gemini17%30%
Claude2%11%

Comscore also found Google showing an AI Overview on 39.4% of desktop searches in June, up from 25.8% a year earlier. iPullRank’s Michael King put the takeaway plainly:

“Publishers should stop asking ‘how do we show up in ChatGPT’ and start asking ‘where is our audience actually asking questions.’”

— Michael King, iPullRank, quoted in Digiday

Market share tells you where people ask. It doesn’t tell you whether the engines give them the same answer. We had the data to check.

What we measured

The sample is seven businesses: five local service firms and two software companies. Each one has a frozen set of buying questions, like “fee-only financial advisor in San Diego” or “best appointment scheduling software.” On September 15 we put all 260 questions to ChatGPT, Claude, Gemini and Perplexity, and kept only the questions all four engines answered. Then we asked one thing of each answer: does it name the business?

The engines disagreed on a third of the questions

How many engines named the business, per question (260 questions)
Engines naming the businessQuestionsShare
All four5722%
Three228%
Two2510%
One4116%
None11544%

The engines agreed on 172 questions: all four named the business, or none did. On the other 88 they split. On 41 questions a single engine named the business and three didn’t. If that one engine isn’t the one you track, those wins don’t exist in your report.

The biggest engine wasn’t the most generous one

Share of answers naming the business, by engine
EngineAnswers naming the businessRate
Claude121 of 26047%
Gemini98 of 26038%
ChatGPT84 of 26032%
Perplexity82 of 26032%

ChatGPT has half the market but tied for the lowest rate in our sample. Claude, at 11% of prompt volume, named these businesses most often. That doesn’t make Claude “better.” It means each engine has its own habits, and a ChatGPT-only number would have understated all seven businesses.

They rarely agree on who comes first

Being named is one question. Being named first is another, because buyers read the top of the list. On 195 questions where every engine listed at least one brand, all four put the same brand first only 44 times, or 23%. On the rest, the “top answer” depended on which assistant the buyer opened.

They read different sources, too

The disagreement starts upstream, with what each engine reads. Digiday’s companion piece on AI visibility measurement gathers several examples:

We see the same thing at a local scale. Several of the businesses we track have a feature page in a city magazine. In the latest round, Perplexity cited those pages in 32 answers and ChatGPT in 6. The same placement looks strong or weak depending on the engine you check.

Gemini needs its own handling

Gemini is the fastest-growing engine in Comscore’s data, and robots.txt can switch it off. Google says the Google-Extended token also controls whether Gemini can use a site’s content in its answers. Many sites block it thinking it only affects training. We saw a site that blocks it get no Gemini citations at all.

What to do about it

  1. Track every engine your buyers use. At current shares, ChatGPT-only tracking misses half the prompts.
  2. Report each engine on its own line. A single blended score averages away exactly the disagreement shown above.
  3. Weight by your audience, not the headlines. Comscore’s shares are national. B2B buyers, local shoppers and developers split differently.
  4. Fix gaps engine by engine. If Perplexity cites you and ChatGPT doesn’t, look at what ChatGPT reads for that question. Often it’s a Reddit thread or a directory you’re missing from.

Limits of our data. This is seven businesses in one round, mostly in one metro area. It shows that engines disagree and roughly how much. It is not a national estimate of each engine’s behavior. Answers were read by a language model against a fixed rubric.

Frequently asked questions

Is it enough to track ChatGPT?

No. Comscore puts ChatGPT at about half of AI assistant prompt volume as of June 2026. In our data, the four engines disagreed on whether to name the business on 34% of questions, and on 41 of 260 questions only one engine named it.

Which AI engine mentions businesses most often?

In our September 2026 sample of 260 buying questions, Claude named the tracked business in 47% of answers, Gemini in 38%, and ChatGPT and Perplexity in 32% each. The ranking will differ by industry and market.

Do the engines agree on the top recommendation?

Rarely. All four engines listed the same brand first on only 23% of the questions where each engine listed a brand.

Every AI tool, one report

See where each AI tool names you, and where it doesn’t.

A BeCited audit reports every AI tool separately. See what ChatGPT and Google’s AI tell buyers in your city about your business. The $20 Snapshot asks about 25 of their questions.

See where you stand: $20 See a sample report

Sources. Market share, conversation and AI Overview figures are from Comscore’s Q2 2026 AI Intelligence Report as reported by Sara Guaglione, Digiday, September 22, 2026. Source-mix figures from Adobe/Semrush, Tinuiti and Sitecore Scrunch are as reported by Seb Joseph, Digiday, September 15, 2026. Agreement, per-engine and citation figures are BeCited tracking data from the September 15, 2026 round.