Google has spent more than a year putting AI-generated answers directly into the results page, but until recently there wasn't much reliable data on their real reach. In January 2026, 25.8% of searches in the United States already showed an AI Overview, and for informational searches that figure rises to 39.4%. For any business that depends on people finding it through Google, that changes the rules of the game quite a bit.
What the data says
A study by Ahrefs covering 300,000 keywords found that pages ranking first but not cited inside the AI Overview lose 34.5% of the clicks they had before it existed. Pages that do get cited, on the other hand, gain 35% more clicks than before. Pew Research Center reached a similar conclusion with its own study of 68,000 searches: the share of people who end up clicking on any result drops from 15% to 8% when an AI Overview appears in between.
Why some websites make it into the answer and others don't
Here's the nuance that interests me most as someone who builds websites: ranking well helps, but it isn't the only thing that counts. Almost all the pages cited in an AI Overview (97%) were already among the top 20 organic results, so traditional SEO still matters as much as before. What's changed is that, within that group, Google chooses what to cite by looking at something else: how clearly the content is structured and the authority signals of whoever publishes it, not just the ranking.
- Well-implemented structured data: FAQ schema on question-and-answer content, HowTo on step-by-step content, Article with clear authorship on articles and news.
- Real authorship behind the content, showing who signs it and why they know the subject, not an unsigned, unattributed text.
- Content that can't simply be copied: proprietary data, real case studies, tools or calculations that only exist on that website.
What this changes for a small business
When I review the code of a website a client built themselves with a generic template, I almost always find the same thing: zero structured data, titles that say nothing specific, and content that looks too much like the business next door's. That website might load fast and look decent, but to Google it's indistinguishable from a thousand others, and now that indistinguishability costs more than before: it simply gets left out of the answer.
In the projects I run now I add the correct schema from the start (organization, services, frequently asked questions, reviews) and make sure every page clearly states what it's about and who signs it. This used to be just one good practice among others. With AI Overviews occupying the whole top of the search results, it's the difference between Google using you as a source or passing you by.
