Most "AI search statistics" roundups repeat last year's numbers with a new year stamped on the title. The genuinely new data this quarter — drawn from Position Digital's July 2026 update — changes a few assumptions worth updating your GEO and AEO plan around, particularly on citation concentration and where the growth is actually happening.
What actually changed this quarter
Three shifts stand out from the rest of the noise:
AI Mode passed 1 billion monthly active users in May 2026. This is no longer an experimental surface — it's operating at the scale of a mainstream product, with the same 93% zero-click behaviour we've tracked on our GEO & AI Search Statistics page since earlier this year.
Citations are concentrating, not spreading out. The top 10 cited domains now capture around 46% of all ChatGPT citations, and the top 30 capture roughly 67%. Meanwhile Gemini's citation behaviour swung hard in the other direction — its citation rate for one tracked cohort dropped from 99% to 76% between February and March 2026, a reminder that citation share moves fast and unevenly across platforms.
Structure is outperforming schema. Pages with 120–180 words between headings receive around 70% more ChatGPT citations than densely-written pages — but schema markup alone showed no significant citation uplift in the same analysis. The lesson isn't "skip schema" (it still matters for parsing and trust), it's that structure and clarity are doing more of the actual citation-winning work than markup.
What these stats mean for B2B GEO/AEO
None of this is breaking news for anyone who has been watching zero-click search creep upward. What's changed is the shape of the opportunity:
Citation concentration cuts both ways. If a small number of domains capture most citations, getting mentioned on those domains (industry press, comparison sites, credible third-party reviews) may do more for AI visibility than publishing more content on your own site. This is the same dynamic our stats page data on citation source diversity has flagged for a while — brands are still far more likely to be cited through third parties than through their own domain.
AI Mode's scale makes the B2B research journey partly invisible to analytics. At 1 billion monthly users and 93% zero-click behaviour, a meaningful share of your prospects' vendor shortlisting is now happening inside a chat window your GA4 will never see as a referral. Freshness, structure and third-party citations are the levers that influence that invisible stage.
Content depth and refresh cadence remain the two levers that consistently move citations across every source we track, not one-off tactics. Pages refreshed within the last two to three months and pages that answer a question directly in the first few sentences keep outperforming everything else, quarter after quarter.
How doubleBaRRiL uses these stats in client strategy
We treat quarterly data drops like this one as evidence to update prioritisation, not a reason to chase every new feature announcement. In practice that means:
Tracking share of answer (how often a client is mentioned across a fixed set of prompts, relative to competitors) rather than a single one-off ChatGPT query, because citation behaviour is unstable — our own data shows only about 20% of brands stay visible across five consecutive similar answers.
Prioritising third-party mentions and earned citations over more on-site publishing volume, given how concentrated citations are among a small set of trusted domains.
Auditing structure before markup — answer-first paragraphs, question-based headings, and a genuine 3–6 month refresh cadence — because that's what the data consistently rewards, not the schema checklist alone.
You can see the full, continuously-updated dataset behind this article — including the South African adoption context — on our GEO & AI Search Statistics page, or get a read on where your own brand currently stands with the AI Visibility Checker.
The bottom line
AI search didn't become more forgiving this quarter — zero-click behaviour is as high as ever, and citations are concentrating rather than democratising. But the practical response hasn't changed: answer-first structure, genuine depth, a real refresh cadence, and a deliberate push for third-party mentions beat chasing whichever AI feature made headlines this month.