126 million prompts say the same thing: your client's category, not their content, sets the ceiling on AI visibility

What the study is

The primary source is Semrush's announcement of the 2026 AI Visibility Index, dated June 26, 2026. It scales an earlier 2,500-prompt study up to 126 million U.S. AI search prompts analyzed from January through April 2026, across four environments — ChatGPT, Gemini, Google AI Mode and Google AI Overviews — with benchmarks across 22 industries. Declare the obvious interest up front: Semrush sells an AI visibility product, and so do we. Read the numbers, not the conclusions. We are quoting the observed figures and flagging which ones come from a survey instead.

Concentration is the finding

Semrush measured how much of a category's total AI visibility is captured by its top three brands. The spread is enormous:

Semrush's own reading is that the less concentrated categories “may offer greater opportunities for brands to gain visibility over time.” That is the polite version. The blunt version, which matters if you are the one signing a retainer: in an 82.9%-concentrated category, a challenger's realistic goal is to be mentioned at all, not to be recommended first — and a GEO proposal that promises otherwise is writing a cheque the category will not cash.

The scarcity is confirmed from the other direction: only 36 global brands held top-100 visibility across all four platforms in every month of the study. Semrush calls them the “Universal 36,” and they are exactly who you would guess — YouTube, Google, Reddit, Amazon, Facebook, Apple, Walmart, Disney, Nintendo. Cross-platform, always-on AI visibility is currently a 36-brand club. Nobody else has it. That is a useful thing to be able to tell a client who thinks they are failing.

The engines are not one channel — and they are not the same size

The single most actionable number in the study: “ChatGPT cites an average of 15 sources per response... while Gemini cites an average of 3 sources per response.” Both draw heavily on community and reference platforms — Reddit, Wikipedia, and for Gemini, YouTube. Do the arithmetic an agency has to do. A ChatGPT answer has roughly five times as many citation slots as a Gemini answer. Gemini is close to winner-take-all; ChatGPT has room in the middle of the pack. The same client, with the same content, can be a visible brand in one and invisible in the other — and Semrush says exactly that: these differences “help explain why a brand can perform strongly in one AI environment while having limited visibility in another.” Which means a single-engine screenshot is not a measurement, and “we checked ChatGPT” is not an answer to “are we visible in AI search.”

Two numbers that are survey, not data — labelled honestly

The two figures the trade press is quoting most come from a companion survey of marketers, not from the 126-million-prompt corpus, and they should carry that label: 45% of marketing leaders say they cannot accurately measure their brand's visibility within AI-generated answers, and only 9% say they have the tools to track all relevant metrics across platforms. Semrush also reports that among organizations fully integrating SEO and AI visibility into one workflow, 81% reported increased traffic or leads from AI platforms, versus 36% of those managing them separately. Self-reported, from a vendor whose product sells into exactly that gap. Directionally believable; not evidence. Treat them as such, and do not put them in a client deck as though they were measured.

What to do with this on Monday

• Check the category's concentration before you scope the engagement. The realistic ceiling for a challenger in a top-3-holds-82.9% category is fundamentally different from one where the top three hold 41.4%. Scope, price, and — most importantly — the promise you make should follow the category, not the client's ambition.
• Measure per engine, and weight by how many slots the engine actually has. With ChatGPT at ~15 citations per answer and Gemini at ~3, an identical “not visible” result means very different things in each. Report them separately or you are averaging away the finding.
• Set the expectation with the Universal 36. If only 36 brands on earth hold visibility across all four platforms every month, then “we are not cited everywhere” is the normal condition, not a failure. Clients who understand that stop asking for the impossible and start funding the achievable.
• Re-measure on a cadence, because the corpus is a snapshot. These figures cover January to April 2026. The engines have shipped substantial changes since. A benchmark you cannot refresh is a fact with an expiry date.

The takeaway

The useful thing about a 126-million-prompt corpus is that it kills the fantasy of a universal playbook. There isn't one. There is a category structure, four engines of very different citation appetite, and a 36-brand club that is visible everywhere. Everyone else is fighting for a handful of slots whose number is set by the engine and whose availability is set by the category — which is why the first honest deliverable for any client is a measurement, not a strategy.

You can run the free 60-second check on any business and see its mention rate across ChatGPT and Perplexity (the $79 audit adds Gemini and Claude, 25 questions sampled three times each — because one run is a sample, not a score). If you do this for clients, the $249 agency 5-pack runs the full repeated-sample audit across five of them, white-labeled, so you can put the category's real ceiling in front of a client before you promise them the top of it.