AI search in 2026, by the numbers

This is the most robust finding in the whole category, because it comes from observed behaviour rather than modelled estimates.

That last figure deserves more attention than it gets. Even when you win the citation, the click is unlikely. Being cited is worth having for influence and mention. It is not a traffic strategy, and anyone selling it as one is misreading the data.

Dissent worth recording: Google disputed the Pew methodology, arguing the measurement window overlapped with unrelated algorithm testing. The direction is corroborated by other studies; the precise magnitude is contested.

If you are planning to become a cited source, the distribution is the first thing to understand.

The strategic conclusion is uncomfortable but clear: for most organisations, the achievable goal is not to become one of the fifteen domains. It is to be accurately represented within them, in the community threads, reference entries, and trade coverage that the engines already quote.

Gartner's 2024 forecast of a 25% decline in traditional search volume by 2026 is quoted constantly. It is worth flagging that it was a forecast, not a measurement, and that referral data showing Google's continued dominance sits awkwardly beside it. Both can be true, query volume and referral share are different things, but they are frequently conflated.

The licensing and litigation picture matters for anyone planning long-term. The corpus these systems draw on is increasingly shaped by commercial agreements and court outcomes rather than open crawling, which means the set of sources an engine is willing to cite is itself becoming a negotiated artefact.