Google AI Overviews cite brands, but still recommend competitors

Google’s AI Overviews are handing brands a dangerous half-win. A page can be visible inside the answer, rack up the citation, and still lose the recommendation to a competitor with stronger proof, broader authority, and more credible third-party support.

That is the real lesson from Lily Ray’s latest analysis: in AI search, being mentioned is no longer the same thing as being chosen. For agencies, that shifts the job from chasing visibility alone to shaping the content that actually survives recommendation pressure.

Citation is not the same as recommendation

Ray analyzed 100 B2B “best [category] software” queries in Google AI Overviews across April 15, May 15, and June 8, 2026. Of the 80 prompts that triggered an AI Overview, self-promotional listicles were cited 323 times. Yet in 224 of those cases, Google cited a brand’s own page without recommending that brand.

That split matters because it exposes a new fault line in AI search. A brand can own the citation and still lose the sale. The user sees the brand in the answer, but the recommendation layer points somewhere else, which means the traffic and the revenue can follow the competitor instead.

The mismatch showed up across help desk, task management, survey, CRM, and SEO software queries, so this is not a one-category oddity. In one example, Google surfaced Oasis LMS as a source but recommended Kajabi, Thinkific, LearnWorlds, and Teachable instead. That is exactly the kind of outcome agencies need to start measuring, because the client may be present in the AI Overview and still be absent from the actual decision.

Why this should change agency strategy

The old SEO reflex was simple: get the page into the result, then celebrate the impression. AI Overviews punish that mindset. If Google is synthesizing a single answer, then the brand that appears inside the answer is only part of the story, and the recommendation itself becomes the real battleground.

Ray’s reporting suggests that brands already leading their categories, those with stronger link profiles, and those mentioned more often by third-party sources were more likely to be recommended. That is a useful clue for agencies because it shows why thin, self-referential “best” pages are such a weak foundation. A page that merely repeats why a company thinks it is No. 1 does not carry much weight when Google is deciding which brand to put forward.

This is where the agency conversation changes. Clients do not need to hear that they “ranked” in an AI Overview and leave it at that. They need to know whether the content is persuasive enough to survive a comparison against Kajabi, Thinkific, LearnWorlds, Teachable, or whichever competitor shows up beside them in the answer box.

The warning from earlier SaaS losses

There is already a precedent for what happens when brands lean too hard on self-ranked “best” pages. A February 4, 2026 Search Engine Land report said some SaaS brands saw 30% to 50% visibility drops after depending heavily on those pages. Many of the affected pages were lightly refreshed with “2026” in the title, and a lot of them lived in blog, guide, and tutorial subfolders.

That pattern should make every agency uneasy. Refreshing the date in a title is not the same thing as improving the substance, and Google’s systems are built to notice the difference. Google Search Central says its ranking systems are designed to prioritize helpful, reliable, people-first information, not content made to manipulate search rankings. Its reviews system is meant to evaluate first-party content that offers recommendations, opinions, or analysis, which is exactly why the quality of the comparison matters so much.

In plain English: if a page reads like a self-promotional shopping list with a new year pasted on top, it is fragile. It may still get cited, but that does not mean it will be trusted enough to win the recommendation.

What to build instead

Agencies should stop treating listicles as the finish line and start treating them as one layer in a broader recommendation system. The content that performs best in this environment will usually do four things well:

• Show real comparison criteria, not just feature dumping.
• Prove authority with examples, links, mentions, and evidence that are not limited to the brand’s own site.
• Differentiate the product in language a buyer would actually use during evaluation.
• Back up claims with enough specificity that a competitor cannot easily outclass the page in an AI synthesis.

That means listicles need better job descriptions. A “best CRM software” page should not just rank products and call it a day. It should explain who each product is for, what trade-offs matter, where each option breaks down, and why one tool is a better fit for a particular buyer profile.

Comparison pages need the same treatment. The useful ones are not mushy brand brochures with alternating bullets. They are decision assets, built around hard distinctions, pricing realities, implementation pain, integrations, support quality, and the specific buyer scenario that makes one product more credible than another.

Bottom-funnel content needs to be even sharper. If AI Overviews are synthesizing a single answer, then case studies, proof pages, alternative pages, and feature-depth pages have to give Google enough confidence to recommend the brand, not just mention it. Third-party mentions matter here, and so does category leadership, because Ray’s analysis suggests those signals help separate the cited brands from the recommended ones.

How to explain this to clients

The cleanest client message is also the hardest one: citation is visibility, recommendation is influence. Agencies should show both, because one without the other can make a campaign look healthier than it is.

That framing also helps with reporting. Instead of a single “AI visibility” metric, agencies should break performance into layers: citation share, recommendation share, authority signals, and conversion influence. A brand can win one layer and lose the others, and that is exactly why old-school rank tracking is not enough in an AI-mediated search result.

Amsive, where Lily Ray is vice president of SEO and AI Search, has been pushing that broader view for a while. Its June 2025 AEO guidance said AI Overviews appeared in 16% of all Google desktop searches in the United States at that time, and it argued that AI engines now synthesize single responses rather than serving a list of blue links. That changes the game. Visibility still matters, but visibility alone does not close the loop.

The agencies that adapt fastest will be the ones that stop selling “we got you cited” as the headline win. In the AI era, the better question is whether the brand is the one Google trusts enough to recommend when the buyer is ready to choose.