A Google AI Overviews barometer on B2B buying queries
Fifty questions that B2B buyers actually type, put to Google AI Overviews, measured every month and published every quarter. The question set is frozen, which is what makes comparison over time possible at all.
This barometer does not measure Fast Growth Advisors. It measures what Google writes, and whom it cites, across five sectors: human resources, artificial intelligence, security, data and analytics, plus one rotating sector in each edition.
What is a Google AI Overview, and why does it hit revenue?
An AI Overview is the answer block Google writes itself, above the results. It does not hand over ten links: it gives the answer, then cites two to five sources. In France it now appears on more than half of all searches.
The effect on traffic has been measured. Ahrefs tracked 963 domains in Search Console, twenty-eight days before the French launch and nine days after. The most exposed sites lost 23.1 % of their click-through rate in nine days. Their impressions did not fall.
This is where it stops being an SEO topic and becomes a management one. The dashboard reports nothing: rankings hold, impressions hold, the curve stays flat. What drops is the number of people arriving on the site, so the top of the funnel, so the cost per lead. A team steering by impressions buys traffic to offset a loss it cannot see.
And the lost click is only half of it. When Google writes the answer, it describes your category in its own words, cites whom it chooses, and recommends vendors. A buyer can form a complete view of your market without opening a single site. The question is no longer whether you rank, it is whether you are the source the machine repeats.
Sources: Ahrefs, “1 mois d’AI Overviews en France et déjà −23,1 % de taux de clics”, 19 August 2026, ahrefs.com. SE Ranking, “AI Overviews Already Appear on Over 52% of French Searches”, data collected 7 and 8 August 2026, seranking.com.
Which edition is available?
The first edition covers the French market and was measured on 1 September 2026. Google wrote an answer to all fifty questions. The 312 sources it cited are spread across 192 distinct domains, and the most cited domain is YouTube, with 47 citations where the runner-up has eight.
A UK edition is in preparation. It will rest on a question set built for that market rather than translated from the French one, because a translated query measures a market nobody searches in. Until it exists, we publish no UK figures.
How to read this Google AI Overviews barometer?
Three states, never merged. Google cites a page, Google answers without citing that page, or Google writes no answer at all on that question. The third is not invisibility: it is an absence of measurement, and writing it otherwise would invent a fault.
Two kinds of question, two criteria. On a comparison query, success means Google names vendors. On a needs query, success means vendor pages enter the cited sources. The two are never added into a single score.
The scope is Google AI Overviews and nothing else. Not ChatGPT, not Perplexity, not browsing agents.
The cost is published with every edition. It runs at four tenths of a cent per question.
What does Google answer in your category?
The free diagnostic puts five questions from your category to ChatGPT, Perplexity and Google AI Overviews. Answers archived, verification link included, causes read on both the message and the pages.
You will know which of your questions Google writes an answer to, whom it cites instead of you, and which ones deserve no budget at all.