Why Do AI Assistants Keep Recommending the Same Brands?
AI assistants keep recommending the same brands because, for broad questions, the sources they draw on agree. The brands named most often in reviews, roundups, comparison pages and everyday writing become the default answer. Search-based answers then retrieve the same well-linked pages that name them. The exact list changes from run to run, but the core usually stays. A SparkToro study found that almost no two answers gave the same list of brands. Tracking in concentrated categories, such as headphones, shows the leading brands appearing in nearly every answer. Smaller brands rarely displace the leaders on broad questions. They break in through specific questions, third-party lists and clear positioning.
What does the evidence show?
- Lists vary, leaders persist. SparkToro's study, led by Rand Fishkin, had 600 volunteers run 12 prompts in ChatGPT, Claude and Google's AI tools, nearly 3,000 times in all. There was less than a 1 in 100 chance of getting the same list of brands twice in ChatGPT or Google's AI. The chance of the same order was about 1 in 1,000. The length of lists tracked how many credible options the topic had.
- Concentration depends on the category. Tracking reported alongside that study found the top headphone brands appearing in 90% to 100% of answers. The top brand design agencies appeared in only 30% to 40%. Categories with a few dominant names repeat them. Fragmented categories spread answers across many providers.
- Third-party lists matter. A study of about 200,000 AI responses found that being included in frequently cited third-party lists was strongly associated with being mentioned.
Why do the same names come back?
| Reason | What happens |
|---|---|
| Source consensus | Most roundups and reviews for "best [category]" name the same leaders, so answers built from them do too |
| Frequency in training data | Brands written about most are the ones models associate most strongly with a category |
| Retrieval | Search-based answers pull the pages that rank and are cited most, which tend to feature leaders |
| Broad questions | "Best CRM" invites the most widely recognised options; there is no detail to favour anyone else |
| Entity clarity | Leaders have clear, consistent descriptions everywhere; smaller brands are described vaguely or inconsistently |
Researchers have raised a related concern about popularity bias, where popular items are recommended disproportionately. They have also found it is not inevitable. One study found a simple language-model recommender showed less popularity bias than traditional recommender systems on a film task. The repetition you see in brand answers has more to do with what the sources say than with any fixed preference in the models.
How does it differ by type of question?
- Broad category questions ("best project management software") return the leaders almost every time.
- Specific questions ("project management software for architecture firms that bills by the hour") open space for specialists, because fewer sources cover them and the leaders' pages may not.
- Constraint questions ("under $10 a user", "self-hosted", "UK data residency") filter out leaders that do not fit.
- Local questions favour whoever the local sources name.
How buyer phrasing changes recommendations is covered in whether buyer phrasing changes AI recommendations more than the platform you use.
What can a challenger brand do?
- Pick the questions you can win: specific use cases, segments, constraints and locations where you are genuinely the better fit.
- Publish pages for those situations with facts the leaders' pages lack: pricing for that segment, workflows, integrations, proof.
- Earn a place in the lists that are cited for those questions, which are often niche roundups rather than the big ones.
- Make your entity clear: one name, one descriptor naming your category and customer, used everywhere.
- Ask every customer for an honest review in their own words; reviews that describe real situations help, but never suggest what to write.
- Measure share on your chosen questions, not on the broad category, where leaders will stay dominant.
How assistants choose brands in general is covered in how AI systems decide which brands to recommend, and why they often name only a few businesses in why AI assistants recommend only one or a few businesses instead of a ranked list.
When do new brands break into broad answers?
Usually when the sources change, which happens in a few recognisable ways:
- A new category or format that the leaders do not yet cover, where early coverage names the newcomer.
- Sustained independent coverage: repeated inclusion in the main roundups, reviews from well-known testers, or awards from credible bodies.
- Visible momentum: funding news, notable customers or rapid growth that journalists write about.
- A leader stumbling: price rises, outages or product changes that reviewers criticise, which opens room in comparisons.
Each of these changes what many sources say at once. A single page or a single mention rarely does. Search-based answers pick up the change first, and answers from model knowledge follow later, as models are updated.
What should category leaders do?
Leaders are not immune. Their risks are different: outdated descriptions, old prices repeated from old reviews, and specific questions where a specialist fits better. A leader should keep its facts current across sources, and correct old claims in the reviews answers still cite. It should watch the specific questions where challengers are gaining. Being named in 95% of broad answers means little if the answers describe a product that has changed.
How should you measure it?
Because lists vary so much between runs, a single answer tells you almost nothing. Run each question many times, at least ten per assistant, and record how often each brand appears, its mention rate. Compare your rate on broad questions with your rate on your chosen specific questions. A challenger might sit at 5% on "best [category]" and 40% on "[category] for [niche]". The second number is the one to grow.
What might an illustrative case look like?
Consider an illustrative accounting software start-up that never appeared in answers to "best accounting software for small businesses", which named the same four market leaders in almost every run. It chose a narrower question it could win: "accounting software for UK construction subcontractors". There it had specific features for the industry's tax scheme and payment practices. It published a page for that situation with pricing and a customer example. It pitched two construction trade publications and asked every customer for an honest review. Four months later, it still rarely appeared for the broad question. It appeared in over half of answers to the construction question, and those leads converted well.
What do people ask about repeated brand recommendations?
Will the leaders always dominate broad questions?
Usually, while the sources agree. Their share can shift when new roundups, reviews or coverage change the consensus.
Is it worth trying to appear for "best [category]" at all?
Track it, but invest where you can win. A strong position on specific questions often brings better-fitting buyers.
Do assistants deliberately favour big brands?
There is no evidence of deliberate favouritism. They reflect what their sources say, and big brands are written about most.
Does paying for ads in AI products change which brands are recommended?
Where AI products show ads, they are labelled: OpenAI and Microsoft keep them apart from the answer, and Google shows them within AI Overviews and AI Mode as ad slots, not as part of the organic recommendation. They can put your brand in front of the user, but they do not change which brands the answer recommends.
Published by The Kovmere Team · updated Sep 29, 2026 · More resources