Why Do AI Assistants Recommend Only One or a Few Businesses Instead of Showing a Ranked List?

How many brands does an AI answer usually recommend?

There is no fixed number. In SparkToro's study, 600 volunteers ran 12 recommendation prompts through ChatGPT, Claude and Google's AI a combined 2,961 times in November and December 2025. The number of items in each answer varied from two or three to ten or more. The lists themselves also changed constantly. There was less than a 1 in 100 chance that ChatGPT or Google's AI would give the same list of brands in any two responses to the same prompt. The same order was rarer still.

Three things drive the count:

• How many real candidates exist. SparkToro noted that the range tracked how many entities appear in sources for the topic. There are only a few Volvo dealerships in one city, but thousands of recent science fiction novels.
• The question's wording. "What's the best..." invites one to three names. "What are some options for..." invites five or more. "List all..." invites a long list.
• How dominant the leaders are. In an analysis of the same prompts by the tracking company Gumshoe, top headphone brands appeared in 90–100% of answers. The top brand design agencies appeared in only 30–40%.

How many websites do AI assistants show compared with traditional search results?

Fewer businesses, but not necessarily fewer websites, and not the same ones:

So the short part of an AI answer is the list of names a reader sees, not the list of sources behind it.

The sources an AI answer cites are also different from the pages that rank. Ahrefs' analysis of 863,000 search results found that only 38% of pages cited in AI Overviews also ranked in the top 10 for the same query. The rest came from lower positions or from beyond the top 100. So being on page one no longer guarantees a place in the short list, and being on page three does not rule it out.

Local recommendations are narrower still. Among brands with 50 or more locations, SOCi counted just 1.2% of locations being recommended by ChatGPT in its 2026 index, while the same locations showed up in Google's local results 35.9% of the time on average.

Why do assistants narrow the field?

• The task is an answer. People ask assistants what to choose, not where to look. A useful answer picks a few options and explains why.
• Readability. A conversational reply with 30 names would be unusable. Products are designed and models are trained to give concise, helpful responses.
• Retrieval limits. Assistants pass a limited number of passages to the model when writing the answer. Candidates that did not make it into those passages cannot be named.
• Evidence strength. Models tend to name options with the most consistent, specific support across sources. An option mentioned once, vaguely, is easy to leave out.
• Constraints in the question. Budget, location, size or use case removes most candidates, sometimes leaving one or two.
• Caution on high-stakes topics. For medical, legal or financial questions, assistants may name fewer options or suggest checking with a professional or regulator.

Randomness sits on top of all this. Answers are generated fresh each time, so which of several similar candidates gets named varies from run to run.

What does this mean for businesses?

• Presence matters more than position. The order of names in an AI list changes from run to run, so being named is the win. Rank tracking in the traditional sense does not apply.
• Being a clear fit for specific constraints wins places. The more specific the question, the smaller and more targeted the list. A business that is clearly the best fit for "for small clinics" or "under £50 a month" can win those answers even if it rarely appears for the broad question.
• Evidence has to be consistent. Since assistants favour options with consistent support, conflicting facts across your site and profiles cost places.
• Third parties decide much of the list. Review platforms, directories and comparison articles supply the candidates, especially for broad questions.

Why the same few brands keep appearing is covered in why AI assistants keep recommending the same brands. How assistants choose among candidates is covered in how AI systems decide which brands to recommend.

How should you measure visibility when lists are short and random?

1. Run each question many times, at least 10 runs per question per assistant, and more for important questions.
2. Record appearance rate: the share of runs in which you are named. This is the core measure.
3. Group questions by topic and buying stage, and report the average appearance rate per group.
4. Track competitors the same way, so you can see your share of the available places.
5. Note how you are described, since a mention with the wrong price or category can do harm.
6. Ignore position within lists unless you have many runs, since order is close to random.

How can you appear in more short lists?

• Build pages for specific use cases, segments and locations, so you are the clearest answer to narrower questions. How focus helps is covered in whether focusing a page on one use case improves AI recommendations.
• Make key facts easy to find and consistent: what you do, for whom, prices, locations and differences.
• Earn places in the sources that supply candidates: review platforms, directories and frequently cited comparison articles.
• Ask every customer for an honest review, since assistants often explain their picks with review themes that customers supply in their own words.

What might an illustrative case look like?

Consider an illustrative independent bike shop in a large city. Asked "best bike shops in [city]", assistants name three to five shops, and it appears in about one answer in ten. Asked "bike shop in [city] that services e-bikes and cargo bikes", it appears in most answers, because its site has a detailed page on cargo bike servicing and its reviews mention it. The shop stops worrying about the broad question. It adds pages for other specialisms it has, such as wheel building and bike fitting, with prices and examples. It keeps asking every customer for an honest review in their own words. Its appearance rate across the specific questions it tracks rises steadily over the next quarter, while the broad question stays roughly where it was.

What do people ask about short AI recommendation lists?

Can I get an assistant to show more options?

Users can ask for more, and assistants will usually oblige. Most people accept the first answer, so the first few names matter most.

Why does my business appear sometimes and not other times?

Answers are generated fresh each time, and when several candidates are similarly strong, the selection varies between runs.

Is being named first better than being named third?

Slightly, perhaps, but order varies so much between runs that consistent appearance matters far more.

Do paid ads change the list?

It depends on the assistant. ChatGPT and Microsoft Copilot show ads separately from the answer, label them and say ads do not influence it. On Google, labelled ads can appear within AI Overviews and, since May 2026, on AI Mode recommendation lists as Highlighted Answers. The unpaid names still come from the sources the assistant retrieves.