Semrush’s most widely shared claim about AI search — that 81% of teams which merged their SEO and AI-visibility work saw more traffic — did not come from the 126 million prompts its new index is named after. It came from a separate survey the company has not detailed.

The distinction matters because the 2026 AI Visibility Index, published on 26 June by Semrush, now an Adobe company after the $1.9bn acquisition closed in April, is landing in budget conversations across B2B marketing. Teams under pressure to prove they can be found inside ChatGPT, Gemini and Google’s AI answers are being handed a single, quotable number. Whether it measures what it appears to measure decides how much weight it can bear.

Where does the 81% figure come from?

Semrush’s 81% figure comes from a companion survey run alongside the prompt analysis, not from the prompt analysis itself. In the company’s account, among organisations that fully integrate SEO and AI visibility into one workflow, 81% reported increased traffic or leads from AI platforms; among those managing the two separately, 36% reported the same. Two things are worth holding onto. The result is self-reported: it records what marketers said happened, not what Semrush measured happening. And the company has not published the survey’s sample size, respondent profile or method, so the gap between 81% and 36% cannot be independently checked. The operative verb is reported, not gained.

That provenance has largely gone unmentioned as the figure has travelled. Trade write-ups have carried the 81% under headlines about brands gaining traffic, folding a self-reported survey answer and a 126-million-prompt measurement into one impressive-sounding whole. The two are not the same class of evidence. A prompt-scale analysis observes what AI engines actually did; a survey records what respondents believe happened to them, shaped by who was asked and how. None of this makes the figure false. It makes it unverified, and softer than a 126-million-prompt headline invites a reader to assume.

What did the 126-million-prompt study measure?

Semrush’s prompt study measured how visible brands are inside AI answers across ChatGPT, Gemini, Google AI Mode and Google AI Overviews, using 126 million US search prompts gathered between January and April 2026 and benchmarked across 22 industries. Its sturdier findings are structural rather than promotional. Only 36 brands — Semrush calls them the Universal 36, among them YouTube, Reddit, Amazon and Apple — held top-100 visibility on all four platforms every month, evidence that strength on one engine rarely carries to another. Visibility also concentrates unevenly: in news and media the three most-visible brands took 82.9% of category visibility, against 41.4% in finance.

The engines differ in how they cite, too — Semrush found ChatGPT drew on about 15 sources per answer while Gemini used around three — and being mentioned in an answer is not the same as having a brand’s own site cited as the source, a distinction The Helm has reported before, when B2B brands that ranked for thousands of keywords surfaced as the cited source in roughly 3% of AI answers. The release also leans on Adobe’s own numbers, reporting that AI traffic to US retail sites rose 1,324% between October 2024 and May 2026 and travel traffic 2,215% — growth that explains why the index exists, though it sizes the channel rather than any one brand’s return from it.

What should marketing teams take from it?

The index’s firmest and most useful finding is a gap, not a win. Semrush reports that 45% of marketing leaders cannot accurately measure their brand’s visibility in AI-generated answers, and only 9% have the tools to track every relevant metric across platforms — a measurement problem sitting underneath every visibility claim, including its own. For a marketing manager, that points to a sequence rather than a slogan: measure visibility engine by engine before assuming a tactic worked, treat the 81% as a directional survey signal rather than a target to hit, and keep self-reported results apart from measured ones in any internal case.

In practice, integration is less a reorganisation than a reporting habit: putting AI-answer citations and mentions on the same dashboard as rankings and organic traffic, so one team can see where a brand is described, cited or missing across engines and act on it. That is worth doing on its own merits, whatever weight the 81% ultimately bears.

The integrated-versus-siloed argument may well be sound — but the evidence offered for it is softer than the index’s scale implies, and recent scrutiny of the wider field has run the same way: a review of 45 generative-engine-optimisation studies found no technique that reliably lifts AI visibility across engines.

The 2026 AI Visibility Index is free to browse at ai-visibility-index.semrush.com. The survey behind its most-quoted number is not, and Semrush has yet to publish the sample or method that produced it.