Google AI Overviews Cite Your Listicle, Recommend Rivals: Why?

Why do Google AI Overviews cite your listicle but recommend competitors? Because Google treats your "best tools" page as research material, not as a verdict. It pulls the list of options from your page, checks each one against what the rest of the web says, and recommends the brands with the strongest independent support, which are often your rivals.

That gap matters more than any ranking report. As of October 2026, the clearest public data shows that a self-promotional listicle can earn the citation link and still hand the recommendation, the part buyers act on, to the brands it compares itself with. This guide explains why that happens, what the research found and how to change it.

What Does It Mean When AI Overviews Cite You but Recommend a Competitor?

It means Google used your page as a source for the answer but told the reader to choose a different brand. Your link appears in the AI Overview, yet the product names in the answer belong to someone else.

The pattern shows up most on "best [category]" queries. These are the high-intent searches where a buyer is close to choosing, so they are also the queries where vendors publish listicles ranking themselves first.

Here is the example that made the pattern famous. For "best LMS for selling courses", Google's AI Overview cited Oasis LMS's own listicle, which ranks Oasis first. The answer then recommended Kajabi, Thinkific, LearnWorlds and Teachable, all of which appear in the Oasis article.

If you're still getting to grips with how the different Google AI surfaces work, our explainer on Gemini vs AI Overviews vs AI Mode covers the basics.

Why Google AI Overviews Cite Your Listicle but Recommend Competitors

Google AI Overviews cite your listicle but recommend competitors because retrieval and recommendation are separate decisions. Your page wins at retrieval because it is relevant and well structured. It loses at recommendation because the model weighs what many sources say, not what one vendor says about itself.

Google's own documentation describes part of this process. It says AI Overviews and AI Mode may use "query fan-out", issuing multiple related searches across subtopics and data sources, and that the models identify more supporting web pages while the answer is generated. We go deeper on fan-out in our post on Google's R4T-Diffusion and AI visibility.

In practice, the journey from your listicle to a rival's recommendation tends to follow four steps:

1. Your listicle is retrieved

A page titled "10 Best CRM Tools for Startups" matches the query almost word for word. It has headings, a ranked list and short descriptions, so it is an easy candidate for retrieval.

2. The model extracts the entities

The model reads your page as a list of options in the category. Your brand is one entity on that list, and so is every competitor you named.

3. Each entity is checked against other sources

Fan-out searches bring in reviews, forums, publishers and other listicles. Brands that appear across many of those sources look like the consensus answer. A brand that ranks first mainly on its own domain looks like a single, interested opinion.

4. The answer recommends the consensus

The AI Overview names the brands with the broadest support and keeps your page as a citation, because it supplied useful facts. You did the research for the answer, and your rivals got the credit.

What Did Lily Ray's Research Find?

Lily Ray's research found that Google usually recommends someone else when it cites a brand's self-ranking listicle. The share rose from 69% in June 2026 to 83% in her September 2026 update.

For the original study, Ray, VP of SEO Strategy and Research at Amsive, ran 100 B2B "best [category] software" queries in Google, checking each one three times between April and June 2026. Of those, 80 triggered an AI Overview. Self-ranking listicles were cited 323 times, and in 224 of those cases the publishing brand was not recommended, according to Search Engine Land's coverage.

The same split appeared in CRM, help desk and SEO software, not just LMS. Ray also found that the brands Google did recommend had far more referring domains and more mentions across AI Overviews and ChatGPT than the brands that were cited but passed over.

Her September update, reported by PPC Land, added two changes:

So Google now cites these pages less often, and when it does, it is even less likely to recommend the brand that wrote them.

Citation vs Mention vs Recommendation: What's the Difference?

A citation links to your page, a mention names your brand in the answer text, and a recommendation tells the reader to choose you. Each is a different level of visibility, and each can happen without the others.

Citation-only appearances are more common than most teams expect. A June 2026 Semrush study with Kevin Indig logged 3,981 domain appearances across 115 prompts and four engines, including Google AI Overviews, and found that 61.7% were ghost citations: the site was cited, but the brand was never named.

There is an Indian angle in that data too. India was among the countries with the highest brand mention rate in the study, at 50%. Comparative prompts such as "best", "vs" and "recommend" also produced a 43.3% mention rate, about 2.4 times the rate for informational prompts. The listicle query is exactly where brand naming happens, which is why losing the recommendation there hurts.

Which Signals Make Google Recommend a Brand?

Google appears to recommend the brands with the strongest independent consensus. In Ray's data, recommended brands tended to lead their categories, to be widely mentioned on third-party sites and to have stronger link profiles than the brands that were only cited.

The common thread is corroboration. A single vendor page saying "we are the best" is one claim from an interested party. Ten independent pages that name a rival as a top pick look like a fact.

Google's guidance on reviews points the same way. Its documentation on writing high quality reviews asks that when you recommend something as the best, you "include why you consider it the best", and that you "provide evidence such as visuals, audio, or other links of your own experience." Most self-ranking listicles skip both.

This is the same logic we cover in how AI agents choose which brands to recommend, applied to Google's AI Overviews.

Do Self-Promotional Listicles Still Work for AI Search?

Self-promotional listicles still earn citations and a modest lift in brand mentions, but they are a weak route to recommendations. Two 2026 vendor studies support both halves of that answer.

Scrunch analysed about 10,000 AI-cited URLs collected between May and June 2026, including 818 self-ranking pages, and found that the author brand's recommendation rate rose from about 4% to about 7% when its listicle was cited. In 63.9% of responses, the AI used the listicle as raw material for a balanced overview rather than adopting its number one ranking.

Several other findings from that study are useful:

• Google AI Overviews recommend less often. When a self-ranking page was cited, the author brand was recommended in 8.9% of AI Overview answers, compared with 15.6% on Claude and 14.8% on Gemini.
• Mentions rise more than recommendations. Brand mentions appeared in about 39% of answers that cited the listicle, against about 19% when it was not cited.
• List size matters. Pages listing 4 to 10 brands performed best for the author brand.
• Rivals are often already winning. Competitors were recommended instead of the author in 24.3% of citing answers, though a placebo test suggested those rivals were already gaining for other reasons.

“Crowning yourself No. 1 may give you a small boost, but don’t expect the crown to do all of the work.”

Niharika Sharma, Scrunch, September 2026

Peec's earlier analysis of 232,000 citations in software review prompts, from December 2025 to February 2026, found that about 11% of AI search citations came from self-promotional listicles. ChatGPT cited them least, at about 3.6%, against roughly 10% for Google AI Mode and Perplexity. Even so, Peec's researchers advised against the tactic because of reputational and algorithmic risk.

Put together, the data says a self-ranking listicle is a citation asset, not a recommendation engine. Ray's September figures suggest Google is also citing these pages less, so the citation benefit may be shrinking.

How to Measure Whether AI Overviews Recommend Your Brand

You measure it by running a fixed set of buyer queries repeatedly and recording citations, mentions and recommendations as separate fields. A report that counts only cited URLs will hide exactly the problem this post describes.

1. Build a "best of" query set

List 30 to 100 buyer queries for your category, such as "best courier aggregator for small business in India" or "best Shopify app for COD". Include "best for" variants by use case, city and budget.

2. Run each query several times

AI Overviews change from run to run, as Ray's three checks per query reflected. Run every query at least three times per cycle, logged out, from a consistent location.

3. Log three outcomes per answer

For each answer, record whether your page is cited, whether your brand is mentioned and whether it is recommended. Note which competitors are recommended and which of your URLs are cited.

4. Flag "cited but not recommended" answers

These are your highest-priority rows. They show that Google trusts your page for facts but not your brand for the verdict.

5. Calculate recommendation share

Divide the answers that recommend you by the total answers for the query set. Track this alongside citation rate, by topic, every week.

A spreadsheet is enough to start. Our guide to AI visibility tracking for small marketing teams shows a lean weekly setup.

How to Turn AI Overview Citations Into Recommendations

You turn citations into recommendations by giving Google independent reasons to agree with you. That means more honest comparison content on your own site and, above all, more coverage of your brand on sites you don't control.

You don't have to delete your existing listicles. They still earn citations and mentions, and a fair, well-evidenced comparison helps readers. Just stop counting them as your recommendation strategy. For more on building pages that engines want to quote, see our guide on how to get cited by LLMs.

Give each change time. AI answers vary between runs, so judge results on several weeks of repeated checks, not on the first AI Overview you see after publishing.

Conclusion

The answer to why Google AI Overviews cite your listicle but recommend competitors is that a citation and a recommendation answer different questions. The citation asks whether your page is useful evidence. The recommendation asks which brand the wider web agrees is best, and a vendor ranking itself first is rarely enough to settle that.

As of October 2026, the data points one way. Ray's studies show the author brand is usually left out, Scrunch finds only a modest lift from self-ranking pages, and Google is citing fewer of them than it did in June.

Start by measuring recommendation share on your "best of" queries. Then put your effort where recommendations come from: honest comparisons with real evidence, and a steady flow of independent mentions on the sites Google already trusts.