Organic sessions no longer track demand the way they did. What to report instead, when the graph goes down but the pipeline does not.
The uncomfortable meeting
Picture it. Organic sessions are down. Qualified leads are flat or up. Somebody senior asks which number is telling the truth, and the answer is that both are, because they stopped measuring the same thing.
What actually changed
When an AI Overview answers the question, fewer people click. Seer Interactive tracked 53 brands through 2025. Where an AI Overview appeared and the brand was not cited, organic click-through rate fell 67%. AI Overviews show up mostly on informational queries (36% of them in Seer's data, against 5% of transactional), so the visit that disappears is usually someone establishing what a term meant. What survives is the visit from someone who has already been briefed by the answer and is now choosing between options. Fewer sessions, better sessions.
Seer's data points the same way. On queries with no AI Overview, click-through rate rose from 2.93% to 3.97% over 2025, which Seer puts down to the remaining searches skewing towards people ready to act. Commercial click-through rate slipped too, even without AI Overviews, so the commercial side needs watching as well. If you only watch the total, you see a decline. If you split it by intent, you see which part matters.
Four things worth reporting
Report four numbers alongside sessions: impressions on commercial terms, click-through rate split by query intent, presence in AI answers for your priority questions, and conversion rate per session. Impressions show whether buyers still see you. The intent split separates the loss you should not care about from the loss you should. The guide Four metrics to report when organic traffic falls shows how to pull each one.
Set the expectation before it happens
The worst version of this conversation is the one where nobody warned the board. Say early that sessions will probably fall and explain why, and the same data becomes evidence the strategy is working rather than evidence it is not.
