Your Buyer No Longer Sees a List. They See an Answer.

A year ago, review searches in Semrush's 400 keyword panel rarely triggered an AI Overview. By August 2026, nearly 90% did. The panel covers banking, software, and ecommerce keywords in the United States, tracked from September 2025 to August 2026, and review searches made the largest jump of any commercial modifier in it. That shift changed which part of the buying journey SEO still controls.

Ranking a comparison page used to be the work. Now the work is winning a place inside the answer above the ranking, assembled from sources you do not own.

What Semrush measured

The study, published October 21, 2026, followed a fixed panel of 400 United States desktop keywords across five modifiers: alternatives, brand versus brand, best X for Y, pricing and cost, and reviews. Overall coverage rose from 45.75% to 88% when each keyword counted equally, and to 97.08% weighted by estimated search volume. The 12% of keywords with no overview accounted for less than 3% of the sample's estimated volume.

Search Engine Land, reporting the study on October 22, notes that Semrush did not measure clicks or conversions. Coverage shows how often an answer appears, not how often anyone clicks through it.

Every modifier finished above 80%. Best X for Y finished highest at 96.12%, and ecommerce saw the largest industry jump, from 32.62% to 85.11%. A year ago, review searches were the least likely commercial modifier to trigger an overview. By August 2026 they were the second most likely.

The answer is built from other people's pages

Semrush illustrates the point with a search for Mailchimp versus Klaviyo. The overview cites Klaviyo's own guide for the case to choose Klaviyo, then cites Zapier for the case to choose Mailchimp, alongside a YouTube video and another third party site. One brand speaks for itself in that answer. The other brand is described by someone else.

That split is the normal shape of a commercial answer. A comparison page, a review platform, a video, and a competitor's page can all feed the same paragraph. Your own page can rank well and still lose the deciding sentence. One way that happens is when the answer quotes a review site whose numbers are newer than yours.

This connects to my post on Google's official site box test, where I argued winning it was an identity problem before it was a content problem. Commercial answers add a second problem. Even a recognized brand still shares the answer with third parties, and those third parties can end up cited for claims your own page never states plainly. Separately, on a different channel, Marie Haynes' ChatGPT referral data showed a higher conversion rate than Google organic: 13.8% against 3.02%, from the 55 of her 100 sites that tracked conversions, with definitions varying across sites. That is a referral finding, not an AI Overview finding. One plausible reading is that a visitor who arrives through an answer has already done part of the comparison. I covered the study earlier this week.

The audit I run on commercial queries

I run commercial queries through the same fixed prompt panel I use for clients at First Rank, the panel behind the GEO campaign that grew LLM search traffic 98% in 30 days: the same money questions, four engines, logged in and out, run on a Monday cadence, scored on recurrence across three samples, as described in my official site box post. For commercial work I rebuild the question list from the five modifiers in the Semrush study. For each brand I track, I list its best, versus, alternatives, review, and pricing queries, then I record whether an overview appears, which sources it cites, and what each cited source says about the brand.

That log sorts each query into one of three states: missing entirely, described by a third party, or described in its own words. Each state gets a different fix. Applied to the Semrush example, Klaviyo sits in the third state on its own guide, while Mailchimp sits in the second, described by Zapier.

If the brand is missing entirely, the first check is whether a page exists that directly answers that modifier. A pricing page that hides prices behind a demo form does not answer a pricing query. A comparison page that only lists features does not answer a versus query. The page has to state the decision plainly: who the product fits, who it does not fit, what it costs, and how it differs from the alternative.

On a review platform you can influence, the fix may sit there. Check it for an outdated price or a discontinued plan, check its review guidelines, and, where requests are allowed, invite customers after purchase or onboarding. Where the citing source is an independent publisher or a video, your lever is usually your own page and your review footprint, not an edit to their page. Check whether the YouTube review that keeps getting cited tested an old version.

Once the brand's own words win the citation, the job is maintenance. In my experience, comparison pages and pricing guides go stale faster than blog posts, and a stale detail gives the overview room to replace your sentence with someone else's. I treat a changed price, a renamed plan, or a dropped feature as a reason to update the page that week, not at the next quarterly review.

Semrush lists five moves for this work: track mentions and citations, map the sources an overview uses, keep commercial pages updated, partner with YouTube creators, and earn reviews on the platforms that get cited. That framework is theirs; the layer I add is the three state sort, run on a Monday panel, before any page gets edited. A different dataset adds a caution. In Lily Ray's September 2026 re-run of 100 B2B software queries, brands that ranked themselves first were left out of the AI Overview recommendation 83% of the time when their listicle was cited. The pages that earn citations state facts a third party can repeat: prices, limits, comparisons with named alternatives, and review evidence from customers who are not on your payroll.

What to do Monday

Start with ten queries. Take your brand plus its two closest competitors and write out the best, versus, alternatives, review, and pricing queries a buyer would type. Run each one, screenshot the overview, and write down every cited source. Sort the results into the three states above: missing, described by others, or described in your own words.

Then fix in that order. Write the missing decision page first. Correct the third party page that misstates you second. Refresh your own commercial pages third.

For the ten commercial queries that matter most to you, who wrote the answer your buyer read?

I publish these breakdowns on dalindegraff.com.