Your Google Ads search terms report can't show you AI Mode. A 400-account study shows why

If you buy Google Search or Shopping ads for a food, beauty or household brand, you have probably been told two things this year. First, people ask Google's AI Mode much longer questions than they type into ordinary search. Second, your ads can now appear inside AI Overviews and AI Mode. Put those together and you would expect your search terms report to fill up with long, chatty questions.

A study published by Search Engine Land on November 9, 2026 checked. It found that the long questions are not there, and that the reason is not that nobody asks them. It is that Google's reporting is not built to show them.

What the study looked at

Rémi Kerhoas, paid media and data director at the agency Eskimoz, analyzed impression and conversion data from more than 400 Google Ads accounts, pulled with an account-level script, covering Search and Performance Max campaigns from the first quarter of 2024 to the third quarter of 2026. He grouped search terms by length (one word, two to three, four to seven, eight or more) and estimated unreported impressions and conversions by subtracting reported terms from ad group and keyword totals.

The design used three markets on different schedules, so that any AI effect should show up first and most clearly in the US.

Table: Google AI search rollout by market, as listed in the study

Market	AI Overviews launch	AI Mode launch	Ads in AI Overviews	Ads in AI Mode
US	May 2024	May 2025	Mobile from late 2024, desktop from May 2025	Test from May 2025
UK	Aug 2024	July 28, 2025	Not available yet	Planned for Q4 2026
France	July 22, 2026	July 22, 2026	Not available yet	Not available yet

Source: Search Engine Land, Nov 9, 2026 (Eskimoz analysis).

What it found

Nothing, which is the finding. In US accounts there was no meaningful increase in longer search terms over the period, in impressions or in conversions. Broad match showed "an extremely thin slice of long-tail queries" next to a large, flat share of unreported terms. Most AI Max search terms were one to three words long. The UK and France looked the same as the US, even though one has no AI ads at all.

The author weighs three explanations and gives each a confidence level.

• Commercial prompts may simply be shorter (low confidence). Google's "three times longer" is an average across all AI Mode use, and a performance-focused agency's accounts may underrepresent long informational questions.
• AI ads may be a small share of delivery (moderate confidence). Google itself has said AI Mode monetizes less than traditional search, and serving there requires broad match, AI Max or Performance Max.
• The report doesn't show prompts as typed (high confidence). This is the one that matters.

The line in Google's help center

The third explanation comes straight from Google. Its help page on ad group prioritization says that when a search involves "more complex search journeys, for example, searches on Lens, AI Mode, AI Overviews, or auto-complete searches," then "the search term shown in your reporting represents the best approximation of the user's intent." The same page says these searches are not considered technically identical to a keyword, so AI-based prioritization picks the ad group that best matches the overall intent.

In plain terms: if someone asks AI Mode for "a breakfast cereal under 6 grams of sugar my kids will actually eat that I can get at Target," your report may show something like "low sugar cereal for kids." The long question was real. The report shows Google's summary of it.

On top of that, Google omits terms without enough query activity, and the study estimates unreported terms already hide about 40% of search terms. A long, one-off AI question is the least likely kind to clear that bar.

The exposure is growing even if the report can't see it

None of this means AI surfaces are a rounding error for ads. Other data points the other way. AccuRanker, tracking desktop searches where a tracked domain ranks first, found AI Overviews on a rapidly growing share of commercial searches during September.

Show the data

Date	Share with an AI Overview
Sept 1, 2026	26%
Sept 30, 2026	51%

Source: AccuRanker, Nov 5, 2026. Desktop searches where a tracked domain ranks first. Transactional searches went from 27% to 51% over the same month.

The study itself cites Similarweb's figure that AI Overviews appear in more than 40% of US searches, and notes that Alphabet's chief business officer said on the third-quarter earnings call that "lower monetization" in AI Mode had weighed on Search revenue growth. So more of the searches your ads compete in carry an AI answer, Google says AI Mode earns less per search than the classic page, and your reporting gives you an approximation of what was asked. All three can be true at once.

Why a CPG brand should care

Packaged goods brands are heavy users of exactly the campaign types that would serve in AI answers: Shopping, Performance Max, and broad match Search on category terms. They also sell products where the long question carries the decision. "Gluten free," "for sensitive skin," "no added sugar," "safe for septic systems" are the details a shopper puts into an AI Mode question and leaves out of a two-word search.

If the report flattens those questions into short terms, three habits stop working:

1. Mining search terms for new product language. The long-tail phrasing your shoppers use with AI will not show up there. Get it from customer service, reviews and retailer Q&A instead.
2. Negative keyword hygiene on long queries. You cannot exclude a phrasing you cannot see. Google's prioritization picks the ad group by overall intent, so ad group themes and landing pages carry more of the load.
3. Judging AI placement from term length. A flat share of long terms tells you nothing about how often you appeared in AI answers.

What we would watch instead

Campaign-level changes around rollouts. Compare conversion rate, cost per conversion and impression share for your main Shopping and Performance Max campaigns in the weeks around big changes, such as the expansion of ads in AI Mode. It is crude, but it measures outcomes rather than labels. Tip 13 covers the analytics side.

What actually shows in AI Mode for your category. Ask AI Mode the ten long questions your best customers would ask, signed out, from a US location, and note which products appear, sponsored and organic, and from which retailers. We covered the layout in our AI Mode ads checklist. Prompt checks, by hand or with a tool such as SearchPanel from XENA Intelligence, which asks assistants shopper-style questions, are the closest thing to a direct reading.

The product data the ads are built from. Shopping ads and AI Mode's product results both draw on Merchant Center data. If the long question is "under 6 grams of sugar," the attribute that answers it has to be in your feed and on your page, as tip 16 explains. That work pays off whether the report shows the question or not.

Google's reporting changes. The study notes that recent Google Ads updates have added more control and transparency, and leaves open the possibility of reporting broken down by AI feature. If that arrives, it will be the first real view of this traffic.

How to explain it to your finance team

Someone will eventually ask why the ad report doesn't show "the AI stuff." The honest answer fits in three sentences. Google decides which ad group serves an AI Mode or AI Overview search by its overall intent, and reports a short summary of the question rather than the question itself. Long, one-off questions are also the most likely to fall into the unreported bucket, which already hides a large share of terms. So the search terms report can tell us what themes we are buying, but not how often we appeared inside an AI answer.

Then show what you are doing instead: the campaign-level comparison, the monthly AI Mode check on your ten category questions, and the list of feed attributes you fixed. That is a defensible plan for a budget line that will keep growing whether or not the report changes.

The limits of the study

It is one agency's client base, mostly performance-focused, with fewer upper-funnel queries, which the author believes make up most AI Overviews and AI Mode inventory. Performance Max search terms only became available in the second quarter of 2025, which makes earlier data less reliable. Multi-country campaigns were excluded. And absence of evidence in a report is not evidence of absence in the auction; that is the point of the study.

It is also worth saying that this is not unique to Google. Citation and referral data from AI answers are getting harder to read across the board; a separate study published November 5 found Google had started routing 30% of AI Overview citation links through a redirect. The broader lesson for a CPG marketing team is to stop expecting AI search to show up as a clean new row in an old report. Measure the results you can see, check the answers directly, and fix the product facts the answers are built from.