Cited by AI Search Isn't the Same as Recommended

Quick read: A study by SEO researcher Lily Ray, covered by Search Engine Land, analyzed 100 B2B "best [category] software" queries in Google AI Overviews across three dates in spring 2026. Of the 80 queries that triggered an AI Overview, self-promotional "best-of" listicles were cited 323 times — and in 224 of those cases, roughly 69%, the AI Overview cited the brand's own page as a source but recommended a different, competing brand in the actual answer (Search Engine Land, 2026). Being cited and being recommended turn out to be two different outcomes, decided by two different steps of the same system.

Here's the example from that study that's stuck with me: a company called Oasis LMS wrote a "best LMS for selling courses" roundup and, unsurprisingly, ranked itself first. Google's AI Overview cited that page for the exact query it was built for. Then it recommended a different company. Oasis did everything the SEO playbook says to do, got pulled into the AI Overview as a source, and still didn't get named as the answer.

That's not a fluke. It happened in 224 out of 323 similar cases Ray tracked. And if you've been told that "getting cited by AI search" is the finish line, this is the part nobody's explained: it isn't.

What "cited" and "recommended" actually mean

AI Overviews and tools like ChatGPT and Perplexity run on two barely-related steps. First, retrieval: the system pulls a pool of pages that are relevant to the query — your website, a directory listing, a review page, a competitor's roundup. This step is mostly about topic match and technical readability. Second, and separately, generation: the system decides which two or three businesses it's actually going to name to the person who asked. That's a judgment call, and it runs on different signals than the retrieval step did.

A page can clear the first bar — it's about the right topic, it loads, it's readable — without clearing the second one, which is closer to "this is who I'd actually vouch for." That gap between the two steps is exactly what Ray's study measured. Getting cited depends on being verifiable and findable, and it's the entry ticket — but the study is evidence that the ticket doesn't come with a seat.

Why this matters even if you're not a software company

Ray's study was scoped to B2B software companies publishing "best CRM" or "best LMS" roundups and ranking themselves first — that's a specific SaaS marketing habit, not really something a local plumber or bakery does. I want to be straight about that: the 69% figure is a B2B software number, not a verified local-business number, and I'm not going to pretend otherwise.

What does carry over is the mechanism, not the number. Retrieval-then-generate is how these systems work across categories, not just B2B software. Any business whose page gets pulled into an AI answer as supporting evidence — a directory listing, a review page, your own website — goes through the same two-step split, whether the query is "best project management software" or "best HVAC company in [town]."

Translate it into a query an actual local business cares about: "top-rated bakery near me." The AI assembling that answer might pull your website, your Yelp page, and a local directory listing into its evidence pool — genuinely reading and using your content — and still name two competitors in the sentence the customer actually sees. There's no error message when that happens. Your site still loads, your listing's still there, and you just don't get named. It's the same silent gap we've written about before, where a business ranks fine in Google's traditional local pack and still goes missing from the AI version of that same search.

I ran a dog grooming salon for years before any of this, and the thing that would have driven me up a wall about this specific problem is exactly that: nothing breaks, nothing tells you it happened, and you find out only by noticing the phone didn't ring.

The mechanics: what actually separates a citation from a recommendation

Ray's research pointed at a pattern in which brands got named instead of just cited. Three things showed up together:

• Third-party corroboration, not just your own claims. The AI leaned toward brands other people vouched for, not brands vouching for themselves. A business's own "why we're the best" page reads as source material, not proof — these systems appear to discount a business's claims about itself more than claims made about it by somebody else. A write-up on a local blog, a "best of [town]" list someone else compiled, a detailed customer review — that's the corroboration that seems to matter.
• Being a known quantity going in. Existing category leaders had an edge. Showing up in the actual recommendation correlated with already being a recognizable name in that space, not just having the most technically optimized page.
• Authority spread across more than one place. The businesses that got named had reputation signals scattered across reviews, mentions, and outside citations — not concentrated in one polished, self-published page built to rank.

None of that is exotic. It's close to the same "prominence" factor Google has used to rank local search results for years — reviews, citations, and reputation — just applied to a newer decision. Prominence used to help decide who got listed. Now it's also helping decide who gets named out loud.

What won't help

• Writing your own "best in [category]" page and assuming the AI will agree with you. Ray's data is a direct warning here — that's precisely the move that got a brand cited and then skipped for the recommendation.
• Chasing citation count like it's the whole game. Getting pulled into more AI answers as a source is good and it's not nothing, but treating citation volume as the finish line is the exact mistake this study exposes. The recommendation is the outcome that actually sends someone your way.
• Assuming a fully filled-out Google Business Profile settles it. It helps with the traditional local pack. It's not what decides whose name gets said out loud in a generated answer — reputation spread across independent sources does more of that work.
• Waiting for a dashboard to flag this. There isn't one that tells you "cited, not recommended." The only reliable way to know is to ask the questions a customer would ask and see whose name actually comes back.

FAQ

What's the difference between being "cited" by AI search and being "recommended"?

Being cited means an AI system pulled a page — often your own website or a directory listing — into the pool of sources it used to build an answer. Being recommended means the AI actually named your business as one of the options it's suggesting to the person asking. A page can be used as source material without the business behind it ever getting named, which is what Lily Ray's research measured happening in roughly 69% of a specific set of B2B software search results.

Does the 69% figure from the study apply to local businesses too?

Not literally — Ray's study measured B2B software "best-of" queries specifically, not local service businesses, and there isn't a verified number for local search yet. What does carry over is the mechanism: these systems retrieve a pool of source pages first, then separately decide which names to actually say. That two-step process isn't industry-specific; it's how the systems are built.

How would I even know if this is happening to my business?

There's no dashboard for it. The most reliable way is to manually ask ChatGPT, Perplexity, or Google AI Mode the questions a customer would ask about your category and area, and see whether your business actually gets named — not just whether your website shows up as a source link somewhere in the response.

Does having my own "best [category] in [town]" page hurt me?

It's not automatically harmful, but Ray's research is a specific caution against expecting a self-published superlative to double as proof. These systems appear to weigh a business's claims about itself less heavily than claims other people make about that business.

What actually gets a business named instead of just cited?

Ray's findings point to third-party corroboration — reviews, mentions, and write-ups other people published, not just your own site — plus existing category recognition and a reputation spread across more than one source. That's closer to what local search has always called prominence than to a technical SEO fix.

Is there any point trying to get cited if it doesn't guarantee a recommendation?

Yes. Being cited is still the entry ticket — a business that never shows up as a source has no shot at the recommendation slot either. The point isn't to stop pursuing citations; it's to stop treating a citation as the finish line.

Can a small local business compete with bigger names on this?

The evidence points toward reputation and third-party corroboration mattering more than budget or site size, which lines up with how local search has generally worked. A well-run local business often has more specific, verifiable information available than a national brand's templated location page, and that specificity is part of what separates a recommendation from a citation.

---

Checking whether your business is actually getting recommended — not just quietly pulled in as a source nobody sees — is part of what a Get Found on Google and AI engagement tracks month to month.