Insights
How AI Decides Which Brands to Recommend
Key takeaways
- When someone asks an AI assistant for a recommendation, it usually names only a few brands, not a page of ten options. The list is short, and the brands left off it are effectively invisible for that question.
- Traffic that arrives from an AI answer converts far better than ordinary search traffic. Adobe measured AI-referred visitors converting about 42 percent better, and several cross-industry reports put the gap even higher.
- Being recommended is not random. AI systems lean on the sources they can read and trust, which means the brands they name tend to be the ones with clear, consistent, well-structured information across the web.
- This is winnable, but not by accident. Becoming one of the named brands takes a deliberate answer-engine strategy, which is exactly the work most companies have not started.
A few weeks ago I sat with a founder who had spent years and real money getting his company to the top of Google for his category. He was proud of it, and he had earned it. Then I opened ChatGPT in front of him and asked it to recommend the best firms for exactly what he sold. It named three companies. He was not one of them.
That is the moment a lot of executives are about to have, if they have not had it already. The question is no longer only where you rank on a page of blue links. It is whether you are one of the two or three names an AI assistant says out loud when a buyer asks it who to hire or what to buy. Those are very different games, and most companies are still only playing the first one.
The list is short, and that is the whole story
Traditional search gave you room to breathe. Even at position six or seven, you were on the page, and a motivated buyer might still find you. The AI answer does not work that way. When someone asks an assistant for a recommendation, it does not hand back a page of ten options. It gives a short list, often just a few names, sometimes only one.
That list has been getting shorter, not longer. After an update late in 2025, ChatGPT noticeably cut how many brands it tends to name in a single answer, dropping from roughly six or seven down to three or four (Position Digital). It is worth being honest that this varies a lot. Research from SparkToro found AI recommendations are genuinely inconsistent, sometimes naming two or three options and sometimes ten or more depending on how the question is asked (SparkToro). But the direction is clear. The default is a curated short list, and a short list is a brutal thing to be left off of.
Here is the part that should get a CEO’s attention. When you are not on the list, you do not get a worse position. You get no position at all for that question. The buyer never sees you, never bounces off you, never even knows you exist. There is no second page to climb to.
Why the named brands win more than their share
It would be one thing if AI referrals were a small trickle of curious traffic. They are not, and the people who arrive through them behave like serious buyers.
Adobe’s analysis of AI-driven traffic found that visitors arriving from AI assistants converted about 42 percent better than other traffic, spent roughly 48 percent more time engaging with the page, and generated meaningfully higher revenue per visit (Adobe Digital Insights). Other cross-industry reports put the conversion gap even wider, with some measuring AI-referred visitors converting several times better than standard organic search (Pixis). And the volume is climbing fast, with AI-sourced traffic to retail sites up several hundred percent year over year.
The reason is intuitive once you say it plainly. A person who clicks a search result is still shopping. A person who acts on an AI recommendation has effectively been pre-sold. The assistant did the comparing and handed them a vetted answer. They arrive closer to a decision, with more trust already in place. That is why the brands on the shortlist are not just getting more visibility, they are getting better buyers.
So how does the AI actually choose?
This is where the mystique tends to take over, so let me keep it grounded. AI assistants are not making editorial judgments about who is the best company in some deep sense. They are assembling an answer out of the information they can find, read, and trust about your category. The brands that get named are, again and again, the ones that are easiest for the machine to understand and most consistently represented across the sources it draws from.
A few things separate the named from the unnamed:
| Brands the AI names | Brands it skips | |
|---|---|---|
| Information about them | Clear, specific, and current across their own site and the wider web | Thin, generic, or out of date |
| Structure | Clean technical foundations the models can parse and trust | Hard for machines to read or interpret |
| Outside validation | Mentioned accurately in the publications, reviews, and discussions AI leans on | Largely absent from those third-party sources |
| Consistency | The same story everywhere, so the model has one coherent answer | Conflicting details that make the model hedge or move on |
| Topical depth | Recognized as genuinely about their category | Spread thin, with no clear authority signal |
None of that is about being the loudest or the biggest. It is about being legible and trusted to a system that is reading the whole internet and deciding, in a fraction of a second, which few names to repeat.
What it takes to get on the list
The good news is that this is influenceable. The work has a name now, answer engine optimization, and it is a discipline rather than a trick. The path looks like this.
The diagram below is the simple version of how a brand moves from invisible to recommended.
Step 1
Audit the answer
See what AI says about you now
→
Step 2
Fix the source
Clear, structured, current content
→
Step 3
Earn the signals
Presence in sources AI trusts
→
Result
On the shortlist
Named when buyers ask
It starts with finding out what the assistants already say when someone asks the real questions your buyers ask, because you cannot fix a problem you have not seen. From there it is making your own information clear, current, and structured so the models can actually understand you, then building accurate presence in the third-party sources the models trust, then keeping every fact about you consistent so the AI has one clean story to repeat. And because the models keep changing, it is work you hold rather than finish.
If you want the deeper background on the mechanics, our explainer on what answer engine optimization is and the piece on how GEO differs from SEO both go further. The companion question, what the AI is saying about you specifically, is covered in what ChatGPT says about you.
The window is the opportunity
The reason I am pushing clients on this now is that most of their competitors have not moved. A striking share of brands still have effectively zero presence in AI answers even after years of investing in conventional search. That gap is not a permanent state. It is a head start available to whoever takes the shortlist seriously first.
A few years from now, being named by the AI will be table stakes, fought over the way page-one rankings are fought over today. Right now it is open ground. The brands that claim their place on the shortlist this year are the ones the assistants will keep recommending as the behavior hardens. That is the work we do, and it is the rare kind that rewards moving early.
Sources
Frequently asked questions
How do AI assistants decide which brands to recommend?
+
They assemble an answer from the information they can find, read, and trust about a category. The brands that get named tend to be the ones with clear, current, well-structured information on their own site and accurate, consistent mentions across the third-party sources the models rely on, such as reputable publications, reviews, and active discussions. It is less about being the biggest brand and more about being the most legible and consistently represented one.
How many brands does ChatGPT usually recommend in one answer?
+
Usually only a few. After an update in late 2025, ChatGPT trimmed the number of brands it names to roughly three or four per answer, down from six or seven. The count varies with how the question is phrased, but the default is a short, curated list rather than a page of ten options, which means brands left off it are effectively invisible for that query.
Is traffic from AI recommendations actually valuable?
+
Yes, and it tends to be unusually high quality. Adobe found AI-referred visitors converted about 42 percent better than other traffic, spent more time on the page, and produced higher revenue per visit. The likely reason is that someone acting on an AI recommendation has effectively been pre-sold, since the assistant already did the comparing, so they arrive closer to a decision.
What is answer engine optimization?
+
Answer engine optimization, or AEO, is the practice of structuring your brand's information and presence so AI assistants like ChatGPT, Gemini, Google's AI Overviews, and Perplexity can find, understand, and accurately recommend you. It overlaps with traditional SEO but focuses on being cited and named in AI answers rather than only ranking in a list of links.
How long does it take to get recommended by AI tools?
+
It is not instant, because it depends on the models picking up clearer, more consistent, and better-validated information about you across the web. The first step is auditing what the assistants say today, then improving your own content and your presence in the sources AI trusts. Because models update continuously, it is best treated as an ongoing position to hold rather than a one-time fix.