AI Overviews now sit on 88% of shopping-style searches. Here are the ecommerce numbers.

For most of 2025 there was a comfortable rule of thumb in search: Google's AI Overviews showed up on informational questions ("how long does cold brew last") and mostly stayed away from searches where people were about to spend money. If you sold products, the shopping results and the ten blue links were still where the action was.

A study Semrush published on October 21 says that rule is gone. We went through it, plus the cell-level numbers PPC Land pulled from its charts, to see what it means for brands that sell food, drink, beauty and household goods.

What Semrush measured

Semrush tracked 400 US desktop keywords every month from September 2025 to August 2026. They were split across three industries (banking, ecommerce and software) and five kinds of commercial search, which Semrush calls modifiers:

• Alternatives: "alternatives to [product]"
• Best X for Y: "best moisturizer for dry skin"
• Brand vs brand: "keurig vs nespresso"
• Price or cost: "how much does [product] cost"
• Reviews: "[product] reviews"

Those are the searches a shopper makes when they've stopped browsing and started choosing. The question was simple: how often does an AI Overview appear on them?

The headline: from under half to almost all

In September 2025, 45.75% of the tracked keywords showed an AI Overview. By August 2026 it was 88%. Weighted by search volume, coverage was 97.08%, which means the 12% of keywords still without an Overview are mostly the small ones. Semrush says they account for less than 3% of the panel's search volume.

Every modifier rose. Reviews moved most.

Sep 2025Aug 2026

Show the data

Search type	Sep 2025	Aug 2026
Alternatives	48.44%	82.81%
Best X for Y	42.72%	96.12%
Price/cost	43.59%	82.05%
Reviews	30.14%	89.04%
Brand vs brand	63.41%	86.59%

Source: Semrush, AI Overviews commercial modifiers study, October 2026

The "best X for Y" line is the one we'd circle. Those searches went from fewer than half showing an Overview to 96.12%. That's the most common way a shopper asks for a recommendation, and it's now answered by a summary before any link.

Ecommerce went from the lowest to nearly caught up

The three industries started in very different places. Software already had Overviews on 68% of its commercial keywords a year ago and reached about 91%. Banking ended around 89%. Ecommerce started lowest, at 32.62%, and finished at 85.11%. That's the biggest jump of the three.

Inside ecommerce, the mix by search type looked like this in August, based on PPC Land's reading of Semrush's heat map:

Show the data

Search type	Share with an AI Overview
Reviews	96.4%
Best X for Y	88.9%
Alternatives	84.6%
Brand vs brand	80%
Price/cost	71.43%

Source: PPC Land, from Semrush's study data. Each cell covers roughly 21 to 29 keywords, so treat single cells as directional.

The price line is the one that surprised us. A year ago, only 4.76% of ecommerce price and cost searches showed an Overview. In August, 71.43% did. Google was apparently reluctant to summarize prices, which change constantly. It isn't anymore. If an Overview tells a shopper what your product costs, it's pulling that number from somewhere, and it may be a retailer page, a stale review or a reseller.

Reviews are the other one to watch. A review search used to send people to review sites and retailer pages to read for themselves. At 96.4% coverage in ecommerce, the shopper now usually gets a summary of what reviewers think first.

Volume tells a harsher story than keyword counts

Counting keywords treats a search done 50 times a month the same as one done 50,000 times. PPC Land pulled the volume-weighted numbers for August. They are higher on every line.

Search type	Share of keywords	Share of search volume
Alternatives	83%	86%
Best X for Y	96%	98%
Brand vs brand	87%	99%
Price/cost	82%	96%
Reviews	89%	94%

Brand vs brand searches with real volume behind them are almost all covered: 99%. If people search "your brand vs the market leader," assume a Google summary is answering it.

On the searches where people choose, the first thing they read is now Google's summary of other people's pages.

One example worth stealing

Semrush included a small example that shows why this matters beyond coverage. For "should I choose Klaviyo," the AI Overview cited Klaviyo's own guide. For "should I choose Mailchimp," it cited Zapier, a third party. Same kind of question, different sources, and one brand got to explain itself while the other was explained by someone else.

That's software, not snacks, but the mechanism is the same for a CPG brand. When the Overview for "is [your brand] worth it" cites a review roundup instead of your page, the roundup's framing becomes the answer. We think the most practical thing a brand can do with this study is search its own name with each of the five modifiers and write down which sources the Overview cites. That list is your real citation footprint.

Why this lands harder on CPG than on software

Software buyers often go straight to a vendor's site, sign up for a trial and decide there. A shopper choosing a sunscreen, a protein powder or a dish soap usually doesn't. They search, skim, and buy wherever is convenient, often at a retailer they already use. For that shopper, the summary at the top of the page may be most of the research they do.

That makes three kinds of CPG search especially exposed:

• Need-based searches. "Best shampoo for curly hair" or "best snacks for kids with nut allergies" are textbook "best X for Y" queries, the type now covered 96% of the time. The Overview decides which brands make the shortlist.
• Head-to-head searches. Challenger brands live on "us vs the big brand" comparisons. If that search returns a summary sourced from a third party, the challenger's own comparison page needs to be good enough to be the cited source.
• Price checks. A shopper checking whether a premium product is worth it sees a summarized price and a summarized verdict together. If either is out of date, there's no product page visit to correct it.

None of this means the shopping ads and organic listings below the Overview stop mattering. It means the brand's story is now told twice on the same page: once in Google's words at the top, then in yours further down, and fewer people scroll.

What the study doesn't tell you

We'd be doing the data a disservice if we stopped at the big numbers. A few limits:

• No clicks. Semrush measured whether an Overview appeared, not what people did after. The best public evidence on clicks is still Pew Research Center's 2025 study, which found people clicked a regular result on 8% of visits with an AI summary and 15% without. Treat that as a direction, not a forecast for your category.
• Small cells. 400 keywords split across three industries and five types leaves roughly 21 to 29 keywords per cell, as PPC Land points out. The overall trend is solid. A single cell can swing on a handful of keywords.
• Desktop only, US only. Mobile and other markets may differ.
• Three industries, none of them food or beauty. Ecommerce here is general retail. We'd expect grocery and personal care to look similar, but that's our inference, not data.
• The publisher has an interest. Semrush, now part of Adobe, sells tools that track AI Overviews. That doesn't make the numbers wrong. It's worth knowing.

What we'd do with it

Semrush's own recommendations are sensible: map which sources get cited for your commercial searches, keep comparison and pricing pages current, work with creators (YouTube shows up often in Overviews), and earn reviews on the sites that get cited. For a CPG brand, we'd turn that into four concrete jobs.

1. Run the five modifiers on yourself. "Best [category] for [use]," "[brand] reviews," "[brand] vs [competitor]," "[brand] price," "alternatives to [brand]." Screenshot the Overview and list its sources. Do it again in a month.
2. Fix the price source. Price searches went from almost never summarized to mostly summarized. Make sure your own product page states price and size clearly in text, and that your [structured data](term:Structured data) matches what retailers show. Tip 2 covers the basics.
3. Give the review summary something to work with. If Google is summarizing what reviewers think, the quality and specificity of your reviews on retailer sites matters more than the star average. Tip 10 has more on how reviews feed assistants.
4. Don't opt out. Google now lets any site exclude itself from AI Overviews in Search Console. Google's help page is clear that other sites' content will still appear and may look similar to yours. For a brand, that means losing your voice in the summary, not removing the summary. More in today's tip 34.

There's also an obvious tension here with the other study everyone is passing around this week, the one showing AI chatbots send very little traffic. Both are true. The AI answer that most shoppers see is Google's, and it's counted as organic search. We wrote about why that matters in a separate think piece.

Checking these answers by hand works for a dozen searches. Teams with hundreds of SKUs tend to automate it; XENA Intelligence built SearchPanel for that kind of per-product check, and there are others. Either way, the first step is the same: look at what Google says when someone asks about you, because on almost every buying search, it's now saying something.