Comscore just put a hard number on something service business owners have been feeling for a year. In June 2026, 39.4% of U.S. desktop searches on Google returned an AI Overview. In July 2025 that figure was 25.8%. That is a jump of 13.6 percentage points in eleven months, reported by Search Engine Land from Comscore’s Q2 2026 AI Intelligence Report.
Four in ten searches now open with a machine’s answer instead of a list of blue links. Search demand itself is not shrinking. Comscore counted 77 billion U.S. desktop searches in Q2 2026, up 8% from Q2 2024. People are still looking for what you sell. The route they take to find it is what changed.
The answer layer is the default now, not the exception
Think about what that means in practice. A homeowner types in a problem. Before they see a single business website, a model has already read the web, decided which companies look credible, and written a short answer that names two or three of them. Everyone else sits below the fold. Most people never get there.
That is the part worth sitting with. You are not competing for position ten anymore. You are competing to be one of the handful of names a machine is willing to say out loud when someone asks who to hire.
Be precise about what the data does and does not show. Comscore measured how often AI Overviews appear, not how many clicks they absorb. Anyone throwing around a specific click loss figure is guessing. But the structural change is not in dispute: the surface where buyers first encounter your category is now an answer, and answers have a much shorter guest list than a results page does.
“Just get me into ChatGPT” stopped being a strategy
The same Comscore report shows the assistant market splitting apart. ChatGPT’s share of AI assistant prompts fell from 70% in January 2026 to 50% in June. Gemini climbed from 17% to 30%. Claude went from 2% to 11%. On Bing, Copilot Search showed up in 17.3% of desktop searches in June 2026, up from 11.8% a year earlier.
Read that as a buying market, not a tech story. Your prospects are now spread across four or five answer engines, each pulling from a different blend of sources, each with its own idea of what makes a business trustworthy. Optimizing for one of them is a coin flip. Businesses that got recommended by one assistant last quarter are finding out that it bought them nothing everywhere else.
This is also why the old agency pitch of “we will rank you” has quietly stopped matching reality. Ranking is one input into a system that now decides recommendations, and it is no longer the most important one.
What actually decides whether a machine names your business
At a high level, four things move the needle, and they compound.
First, whether a machine can read your site at all. A lot of modern sites look beautiful to a human and return almost nothing usable to an automated reader. Second, entity clarity: whether your business is described so consistently across the web that a model can be confident who you are, what you do, and where you serve. Ambiguity gets you skipped, because models avoid naming businesses they cannot verify.
Third, corroboration from sources you do not own. Reviews, citations, listings, mentions on sites the model already trusts. A claim on your own homepage carries almost no weight. The same claim confirmed in five independent places carries a lot. Fourth, structured signals that spell out your services, service area, pricing posture and credentials in a format built for machines rather than for eyeballs.
None of that is a weekend project, and the shortcuts are getting more expensive by the month. Google began rolling out its September 2026 spam update on September 24, globally and across all languages, with a rollout window of up to two weeks. It is the fourth spam update Google has announced this year. The pattern is consistent: the cheap tactics get cheaper to detect, and the durable signals get more valuable.
The businesses winning this are not doing anything magical. They built a foundation that machines can read, verify and repeat, and they did it before their competitors thought to. You can see the shape of that work on our how it works page, and what it produces on results. Common questions are answered on our FAQ.
Andy Rodriguez is the founder of Agentic Growth Engine, where he helps service businesses across the country get found, trusted and recommended by AI search.
So here is the only question that matters this quarter. When someone asks an AI who to hire in your market, does it say your name, or your competitor’s? There is one way to find out. Run our free AEO and SEO visibility audit and see exactly how the answer engines describe your business right now, before you spend another dollar guessing.