Why AI Overviews quote someone else on your revenue queries

When Google shows an AI Overview, people click through to a regular search result on about 8 percent of those searches, against 15 percent when no Overview appears (Pew Research Center, July 2025, drawn from the browsing records of 900 US adults). Rank still exists, but on those queries it matters less than whether Google quotes you inside the Overview. This audit answers one question across the queries that carry your revenue: can a page you control become one of the cited sources, and if so, what has to be on it. What the engagement decides: Three things, and we commit to all three in writing. First, which of your queries actually trigger an Overview in the markets you sell in, measured rather than assumed. Second, whether the citation slot is reachable by a vendor page at all, or held by a class of source you cannot join. Third, for the reachable ones, the exact passage that would have to exist on your page for it to be quotable. How presence is measured: An Overview firing once is not evidence. The same query can produce an Overview on Tuesday and none on Thursday, and the cited set shifts with it. We check every query in the set five times across ten days, from at least two locales, logged out, and count a query as Overview-present only when it fires in three of the five checks. We also collect the sub-questions Google fans the query out into, because citations attach to those rather than to the head term. A brand that loses the head term often already wins two of its sub-questions and nobody internally knows it. Who does the work: Both partners run their own engagements. One spent six years leading organic at a US insurance marketplace and killed a content programme two quarters after launching it, against the wishes of the CMO who commissioned it. The other ran content for a developer tools company through a documentation migration with a fixed cutover date. We have told three clients in the past year that the query set they wanted audited was not worth auditing, and returned the deposit. When the answer is inconvenient: Roughly a third of the query sets we look at contain a block no commercial page will ever be quoted for. Dosage questions cite NIH and Examine. Regulatory questions cite the regulator. Best-of questions frequently cite a Reddit thread and a publisher roundup, and your own page is structurally the wrong kind of document for that slot. When that is what we find, we say the set is closed, we say which portion of your traffic plan depends on it, and we cut our own scope rather than sell a rewrite that cannot work. You get that answer in week one, not in a final deck. What you receive: A query ledger as CSV: query, Overview presence rate across the five checks, the fanned-out sub-questions, every cited URL with its source class, your current rank, and a verdict of reachable, contested or closed. A source-class summary showing who owns your category's citations and how concentrated that ownership is. For every reachable query, a passage specification written as final text, with the heading it sits under and its position on the page, so your writer copies rather than interprets. A working session with whoever owns those pages, because specifications only ship if that person agrees with them. Not included: We do not write articles, we do not build links, and we do not touch your CMS. We do not implement schema, though we will tell you where yours is wrong. Tracking what the Overviews do next is a separate engagement, and we will say so rather than fold it in quietly. Who this is not for: If your non-brand organic sits under about 20,000 clicks a month, the query set is too thin for five-check sampling to mean anything, and the money is better spent on pages. If a decision has already been made internally and you need a document that agrees with it, we are the wrong firm. About one engagement in five ends with us contradicting the brief that hired us, and we would rather tell you that now than in week three.