TL;DR
When an answer appears above the results, impressions can climb while clicks fall. Judge the channel on qualified visits and pipeline, not on clicks alone. Track brand and non brand queries separately, watch which queries still produce clicks, and treat being cited in the answer as a placement you can earn.
Why clicks and impressions move apart
AI features in Google Search now answer more questions directly. A visitor who reads the answer on the results page never reaches your site, but the impression still counts. A dashboard that only reports clicks will show a decline with no explanation.
The mistake is to read that as a ranking loss. Check position and impressions first, which is the opening move of a technical SEO audit. If both are stable and clicks dropped, the query is being answered on the results page.
What to measure instead
- Qualified sessions, meaning visits that reach a service or pricing page
- Assisted conversions and form starts, not only last click conversions
- Brand versus non brand queries in Search Console, tracked as separate groups
- Queries where your page still earns a click, which are usually specific and commercial
- Citations in AI answers, recorded manually per topic until a reliable tool exists
Split brand from non brand
Brand queries behave differently from discovery queries, which is why the Search Console performance report lets you filter them separately. Someone searching your company name intends to reach you, and an AI answer does not replace that visit. Discovery queries are where answer boxes take the click.
Segmenting the two, which is ordinary keyword research on your own query list, shows which decline is structural and which is a real problem. A drop concentrated in informational queries is expected. A drop in commercial queries, where the visitor was ready to act, is the one to investigate.
What we tell clients
Change the reporting before changing the strategy, which is what reporting dashboard setup is for. A channel judged only on sessions will look worse every quarter as more answers appear on the results page, even when the work is sound.
Then write for the answer. Clear structure, direct statements and specific detail make a page easier to cite, and structured data makes the facts machine readable too. That is the same discipline that made pages rank in the first place, applied to a different surface. The University at Buffalo case study is the reporting side of that work.