AI Overviews and ads, what search advertisers can plan for now
Google is rolling out AI Overviews in the US and says it will test Search and Shopping ads in them. What is confirmed, what is not, and what to watch in your reports.
Two announcements in one week changed what a search results page can look like in the US. On May 14, at Google I/O, Google said AI Overviews would begin rolling out to everyone in the US, with more countries coming. On May 21, at Google Marketing Live, it said it would start testing Search and Shopping ads in AI Overviews for US users. Between them, these decide how much of your search budget will be exposed to a new layout.
The honest summary for advertisers is that very little has been specified yet, so most of the useful work is preparation and observation.
What is confirmed
- AI Overviews give a short summary of a topic, produced by a custom Gemini model, with links to learn more. Google said they would reach hundreds of millions of US users in the first week and that it expects more than a billion people to have access by the end of 2024.
- Ads will appear in dedicated slots and be labeled as sponsored. Coverage of Marketing Live describes a section labeled "Sponsored" above and below the overview, with ads shown when they are relevant to both the query and the information in the overview.
- The test is small and US-only at first. Reporting on the event says advertisers do not need new ads, and that existing Search, Performance Max and Shopping campaigns are eligible.
Not confirmed: when the test will start, how many queries will show an overview, what controls advertisers will get, and how impressions in these slots will be reported. Marketing Dive noted that it had asked about availability and had no date to report.
Where the exposure is
An AI Overview is most likely to appear on informational and comparison queries, the kind where a person wants a quick explanation before choosing. For an app advertiser, that points to category and problem-led queries rather than brand terms. Someone searching for an app by name is not looking for a summary, and your brand campaign is the least likely to be affected.
That gives you a way to prepare without guessing. Split your search spend into three groups: brand, high-intent category and long-tail informational. Put the third group on a watch list. If click-through rate on informational queries drops or impression share shifts in the coming weeks, you will see it there first.
What to do with campaign settings
Do not rebuild anything yet. There is no setting to change for ads in AI Overviews, and reacting to a test that has not begun invites mistakes. What does make sense now:
- Tighten your conversion tracking. Placement in an overview may be driven by relevance, so campaigns that already send clear conversion signals will have better data to bid with.
- Check your landing pages and extensions. If an ad shows next to a summary, the person arriving has already read a generic answer. The page must offer something the summary did not, such as a specific feature, a price, or a clear next step.
- Review broad match keywords and negatives. Because Google says eligibility depends on relevance to the query and the overview, a loose keyword set may enter auctions you did not intend. Look at your search terms report for irrelevant matches.
- Keep a baseline. Record click-through rate, cost per click and conversion rate by query group for the last 30 days.
What not to conclude
Do not read this as a reason to pull spend from search. Google has also said, in its launch post, that links inside AI Overviews get more clicks than a traditional listing for the same query. That is Google's claim about organic links and has not been independently checked. It tells you nothing about how paid slots will perform. Treat both sides, the optimism and the alarm, as unproven until the data is in your own account.
If you buy in markets outside the US, nothing here applies yet. Google has said only that more countries are coming.
Practical takeaway
Group your search queries by intent, record baselines for each group, and set a calendar reminder to compare after the test begins. Make decisions from your own reports, with a defined threshold, rather than from headlines.