Google Now Sells the Ad and Writes the Recommendation. Who Checks Either One?

On Alphabet's October 28 earnings call, Google's Philipp Schindler said that AI Mode queries monetize less well than traditional Search, and that retailers are adopting Google's checkout protocol quickly (paraphrased from Investing.com's transcript). Matthew Bertram's view of what the call means for the company is on matthewbertram.com. This page is for the people on the other side of the answer: the advertiser, the SEO, the merchant. In AI Mode the sponsored result, the organic recommendation and, for some merchants, the checkout now come from one company in one answer. What can you see of each?

Advertisers get an aggregate and an explainer they cannot edit

Ads in AI Overviews and AI Mode are not a separate buy. On its help page Google says “Existing Search, Shopping, and Performance Max campaigns are automatically eligible to serve ads within AI Overviews and AI Mode”, and that “you can't target ads exclusively to AI Overviews or AI Mode, or opt out of showing ads in them separately from Search”. On reporting it says “Performance metrics for ads in AI Overviews and AI Mode are included in aggregate in your standard Google Ads campaign reports”, and the search terms report works the same way: “Queries that trigger your ads in AI Overviews and AI Mode are included in aggregate in your search terms report”. The page describes no segment that splits those surfaces out of Search.

Who gets matched is not left to the keyword list either. Google says queries in these surfaces are “often longer, exploratory, and conversational compared to traditional searches”, and that exact and phrase match keywords “aren't eligible for contextual matching on broader exploratory questions”. Its advice is AI-powered targeting: AI Max for Search, broad match, Shopping campaigns or Performance Max. An account built on exact and phrase match alone is, by Google's own description, matched only when the query itself shows direct commercial intent.

Some placements carry a second voice. Google calls it an AI explainer, “concise, independent text generated by Google's Gemini models”, and says: “Unlike creative assets, advertisers can't manually write, edit, approve, or opt out of them”. In some layouts it draws on the query, the AI response and public sources including your website, and the lever the page names for it is keeping your own site content clear and current. In its May 20 post Google said that “both of these new formats will feature an independent AI explainer as part of the ad”, naming Conversational Discovery ads and Highlighted Answers. A Highlighted Answer is an ad that can sit on the recommendation list itself: when AI Mode lists recommendations, “highly relevant, high-quality ads are eligible to appear on that list as a Highlighted Answer”.

Google also states the wall between the two halves: “Buying ads doesn't influence organic AI answers, and whether your site is cited organically doesn't affect your ad auction eligibility”. That is a claim about Google's systems, and nothing in either report described on this page lets an advertiser or a site owner check it.

The May 20 post says nothing about how its formats will be reported. Its advice for taking part is to “build a strong foundation with AI Max for Search, AI Max for Shopping campaigns and Performance Max”, all of them campaign types the help page files under AI-powered targeting.

Organic gets impressions, and a study that cannot say why

The organic tool is the Generative AI performance report in Search Console. Google's announcement describes “dedicated views of your impressions within generative AI features on Search, such as AI Overviews and AI Mode”, and carries the note “As of August 31, 2026, we've rolled out these insights to all websites worldwide.” The help page lists what the report has: impressions, pages, countries, devices and dates, with AI Overviews and AI Mode counted together. Experiments are left out: “Search Console doesn't include data from experiments in Search Labs”. A site with too little volume sees no report at all: “This usually happens when your site hasn't received enough impressions in generative AI features on Google Search.” Google says it is working with site owners on what else the report should show, including “adding additional metrics over time”.

So a site owner can see that a page appeared. There is no click count in the list Google gives, no record of which brand the answer recommended instead, and no copy of what the answer said.

One public attempt to fill that gap is Seer Interactive's April 24, 2026 update. On informational queries, Seer found that a brand cited in an AI Overview earned a higher click rate than a brand that was not, on both organic and paid results.

Seer's own summary: “Being cited in the AIO delivers +120% more organic clicks per impression versus when you are not cited. But it still underperforms No AIO by -38% in informational queries”. Two cautions come from the same report. The first: “We cannot claim causation. Higher-authority brands are also more likely to be cited.” The second: “AIO status is a static label from the most recent SERP snapshot, applied retrospectively across all months.”

The paid result needs a closer look. Seer describes “a 4+ percentage point advantage that persisted every single month in 2025”. Its monthly table does not quite agree: in April 2025 the paid click rate was 16.07% where the brand was cited and 14.35% where it was not, and in May 2025 it was 14.41% against 15.62%, lower for the cited brands. Either the sentence or those two rows is off, and we cannot tell which. The chart uses the full-year averages, and the monthly claim stays unproven.

For an SEO that leaves two questions Search Console cannot answer: was this page cited, and did being cited pay? Seer shows an association. Nothing published shows the cause.

The checkout is moving onto Google's page

A recommendation can now end in a sale on Google's own surface. In February (Google, February 11), “AI Mode already surfaces organic shopping recommendations based on what's most relevant to a query. Now, we're testing a new ad format to showcase retailers that offer those products, clearly marked as sponsored.” The same post said “UCP-powered checkout is rolling out now, letting US shoppers buy items from Etsy and Wayfair, right in AI Mode in Search and the Gemini app (with Shopify, Target, and Walmart coming soon)”. On May 20 Google added that “We've added native checkout integration for Universal Commerce Protocol (UCP) merchants” to Direct Offers, “a new monetization experience in AI Mode, Direct Offers”, which it began piloting in January.

The Merchant Center help page is plain about the change. The feature is covered for products eligible in the United States, Canada and Australia, and “The checkout feature enabled by UCP is available for select merchants at this time.” With the existing checkout button the transaction occurs on the merchant's site; with this one, “the checkout happens directly on Google's surfaces while keeping you the merchant of record”. The Buy button appears only on listings that carry the native_commerce(checkout_eligibility) product attribute.

A sale that completes on Google's surface does not have to pass through the merchant's site, so site analytics is not where it will be counted, and the page does not say where the merchant will see it.

What ChatGPT and Perplexity recommend instead

Google is not the only engine making product recommendations. BSP measured the other two. In the AI Answer Corpus, 230 of the 331 prompts ask an engine to recommend something.

Where ChatGPT named a brand at least once in 12 identical runs, a single run named it 45.9% of the time, and all 12 agreed in 4.7% of those prompt-months. Perplexity agreed across all 12 runs in 75.1%, but its repeats may be close to one observation. The full set is on one check against twelve. Google AI Mode is not one of the engines in the corpus, so we do not know whether one AI Mode answer is any steadier. A screenshot of one is one draw.

Want a product prompt checked the same way? Tell us the prompt and we will say whether it is in the November panel.

Limits

• Google's pages. Everything attributed to Google is from pages read October 28, 2026: the Google Ads Help page, the February 11 and May 20 posts on the Ads and Commerce blog, the Search Console announcement and help page, and the Merchant Center help page. Google edits these. The formats table is a snapshot, and a format labelled testing can change within a week.
• The earnings call. Philipp Schindler's remarks are paraphrased from one transcript, Investing.com's. We could not confirm the wording against a second copy, so treat them as a single-source claim until you have heard the call.
• Seer. Informational queries only, 53 brands, an association and not a cause, and a study of AI Overviews, not AI Mode. The AI Overview label is a snapshot applied back to every month.
• The corpus. Two engines, ChatGPT through the API version, gpt-5-mini, with web search on, and Perplexity (sonar). 12,408 answers to 331 prompts across 7 industries. The API version is not what a person sees in the app. We have not measured whether an ad, or a citation, changes what an engine recommends.
• Ad performance. This page measures nothing about how AI Mode ads perform.
• Disclosure. Matthew Bertram owns shares of Alphabet (GOOGL).

Quoting a figure? Link to this page as the source: https://www.bestseopodcast.com/research/google-ads-seo-ai-product-recommendations. Google's wording is quoted from pages read October 28, 2026; the Seer figures are Seer's, published April 24, 2026; the corpus figures are dated 2026-10-25. The page is the citation: there is no download.

Who did this

Matthew Bertram

Author · Host of The Best SEO Podcast since 2009 · Owner and CEO, EWR Digital

Matthew Bertram runs EWR Digital, an SEO and AI-visibility agency, and wrote the article on matthewbertram.com that this page accompanies. Every quotation here was checked against the saved text of the Google or Seer page it came from before publishing; the corpus figures come from the study pages linked above. Questions go to him through the contact page. More about Matthew Bertram.