Research · The AI shortlist · October 2026
How many brands does AI recommend?
To find out, we analysed 197,314 AI answers to more than 1,600 real buying and research questions, tracked daily across five leading AI engines (ChatGPT, Google AI Overviews, Gemini, Perplexity and Copilot) over eight weeks in the UK and US. For each answer we counted how many of the competing brands tracked in that category were named, and in what order. When AI names brands, it usually names two or three. And the brand it names first is the one most likely to be there tomorrow.
brands named per answer, when names any. of those answers name three or fewer.
of the time, the brand named first that day was also the brand named most often for that question over eight weeks.
of brands named first were named again the next day. For brands named fourth or later it was .
Not enough daily data on to measure whether brands are named again the next day. Across all engines, first place came back 72% of the time.
of questions had one brand named first on at least half the days. Only were a genuine contest.
01 · How long is the shortlist?
Typically, AI only names two or three brands
A Google results page gives you ten links. An AI answer gives you a handful of names. Across all five engines, when AI named any of the brands we track, it named 2.5 on average. More than a third of those answers named just one. Fewer than one in four named four or more. The room on the list is small.
Of the answers that named at least one tracked brand, how many did they name?
We counted the competing brands each customer tracks in reconnAI, typically the 10 to 30 that matter in their market. Brands outside that list were not counted, so the real list can be a little longer. The shape of the shortlist is the finding, not the exact number.
02 · By engine and by question
Every engine keeps the list short. The question decides how short.
Google AI Overviews, ChatGPT and Gemini all name between two and three tracked brands when they name any. Copilot names more, but we only track it in two categories. What changes the list most is the question itself. Ask AI to choose, with words like "best", "top" or "recommend", and it names more brands.
| Engine | Brands per answer | Named 4 or more |
|---|---|---|
| Google AI Overviews | 2.4 | 22% |
| ChatGPT | 2.5 | 24% |
| Gemini | 2.9 | 32% |
| Perplexity* | 2.6 | 25% |
| Copilot* | 3.9 | 55% |
Average number of tracked brands named, among answers that named at least one. The gap held in .
03 · Who gets named first?
Most questions have a usual leader
AI shortlists change from day to day, but the top spot is steadier than the rest. Across all engines, one brand was named first on 73% of the days AI named anyone for the typical question. In 44% of questions a single brand took first place at least four days in five. Only 16% of questions were a genuine contest, with no brand first even half the time.
How firmly is first place held? Share of questions
of the time, the brand named first on a given day was also the brand it named most often for that question over the eight weeks. Being first and being a regular go together.
04 · Does the order matter?
The brand named first is the one that sticks
We took every brand AI named and checked whether it was named again when the same question was asked the next day. The further down the list a brand sat, the less likely it was to come back. Across all engines, brands named first came back 72% of the time. Brands named fourth or later came back 56% of the time. First beat fourth in every one of the eight categories with enough data to compare.
Share of brands named again the next day, by the order they were named today
Same question, same engine, same country, one day apart. Based on the tracked daily , each weighted equally. First beat fourth in . Brands named first may simply be the stronger brands, so this shows a pattern, not a cause.
We track daily in only one category, which is not enough to report this reliably on its own. Choose All engines to see the combined figure.
The brand named first is points more likely to be named again than brands named fourth or later . On an AI shortlist, the gap between first and the rest is wider than the gap between second and third.
Around 1 in brands named first today were gone from the answer tomorrow. The top spot is the best place to be, and it still needs watching.
Brands named first came back the next day 72% of the time; brands named fourth or later, 56%.
05 · What it means
Getting on the list is the job. Getting to the top is the prize.
There are two or three seats
Page one of Google has ten results. An AI answer has two or three names. Visibility in AI is a tighter race, and the share of answers that name you is the score.
Know your questions
Choosing questions ("best", "top", "which") name more brands than explaining questions. Find the questions where your buyers ask AI to choose, and measure yourself on those first.
Find out who leads
Most questions have a usual leader. If it is you, protect the position. If it is a rival, you know exactly whose spot you are competing for, question by question.
Track the order, not just the mention
A brand named first is far more likely to be named again. Being on the list matters, but where you sit on it tells you how secure that place is.
How reconnAI measures this
Your place on the list, every day
This study is built on the same daily data reconnAI collects for every customer, including which brands each answer names and in what order.
The shortlist has two or three seats
Your share of answers. Your buyers’ questions run daily on the AI engines that matter to you. You see how often you are named, for every question and every engine.
Most questions have a usual leader
Who is named first. See which brand leads each question, how often you take first place, and which competitors are taking it instead.
The question changes the list
Question by question. Group your questions by topic and type, and see where the shortlist is long, where it is short, and where you are missing from it.
The order tells you how secure you are
Trends over time. Watch your share of answers and your share of first places week by week, so you can see a position slipping before it is gone.
Questions
Common questions
How many brands does ChatGPT recommend in an answer?
In our study, when ChatGPT named any of the competing brands tracked in a category, it named 2.5 on average. Google AI Overviews named 2.4 and Gemini 2.9. Three quarters of answers that named a brand named three or fewer. We only counted the brands tracked in each category, so the full list can be slightly longer.
Does the order AI names brands in matter?
Yes. Brands named first were named again the next day 72% of the time. Brands named fourth or later came back 56% of the time. First beat fourth in all eight categories with enough data to compare. Brands named first may simply be stronger brands, so we treat this as a pattern rather than a cause.
Does AI recommend the same brand first every time?
Not every time, but usually. For the typical question, one brand was named first on 73% of the days AI named anyone, and 84% of questions had one brand first on at least half the days. Only 21% of questions had the same brand first on every day we checked.
How does AI decide which brands to recommend?
The engines don’t publish their rules, and our study measured what they do rather than why. What we can see is that the list is short, that choosing questions produce longer lists than explaining questions, and that the brand named first is the most likely to be named again. Many engines also search the web before answering, so the pages, forums and videos they cite shape which brands appear.
How do I get my brand recommended by ChatGPT?
Start by measuring. Find the questions your buyers ask, see how often each engine names you and in what position, and find out which competitor usually leads. Then look at the sources AI cites for those questions, because that is where the names come from. We measured what the shortlist looks like, not what gets a brand onto it, so track any changes you make over weeks rather than judging them from a single answer.
Method
How we measured it
reconnAI puts the same real buying and research questions to AI engines every day. For this study we used more than 1,600 questions from ten categories, asked in the UK and US on the free version of each engine, between 31 August and 25 October 2026. We left out questions that name the company being tracked, and we never counted a brand that was named in the question itself.
In each category we counted the competing brands that reconnAI tracks for that customer, typically 10 to 30, and the order each answer named them in. Brands outside those lists were not counted. "First place" analysis covers questions with at least ten days of answers naming a tracked brand (five for the four categories tracked for less of the period). "Named again" compares each answer with the answer to the same question, on the same engine, in the same country, one day later. Choosing questions are those containing words such as best, top, recommend, which, versus or alternative.
Every figure is worked out for each category first and then averaged, so each category counts equally and no single one dominates. The engine selector shows the same calculation restricted to one engine. Google AI Overviews and ChatGPT are tracked in all ten categories, Gemini in five, Perplexity and Copilot in two, so figures for the last two are a guide rather than a benchmark. For the headline figure of 2.5 brands per answer, we are 95% confident the true average is between 2.1 and 3.1. All data is anonymised.
Are you on the shortlist, and where?
We run your buyers’ questions every day across the major AI engines and show you how often you are named, in what position, and who leads each question.