• AI visibility

Guide · AI search

Why is our well-known brand missing from AI product recommendations?

Because being famous does not guarantee a place on the short list an AI assistant builds for each question. Researchers have caught household names such as L.L.Bean and Craftsman at zero in plain category questions, even though the assistants clearly knew them. In most cases the brand is known but not called up by the way buyers ask, so the fix starts with measuring the right questions.

By Underneath Team Guide Updated

The short version

  1. Six AI models answering from memory never recommended L.L.Bean, Craftsman, Braun or Philips for a plain category question across 1,200 answer lists (Malthouse and colleagues (opens in a new tab), Northwestern University, 2026).
  2. When the question used the brands’ own positioning language, L.L.Bean appeared in 88.5% of lists and Craftsman in 81.3%, so the models did know them.
  3. Realistic, detailed buyer questions helped far less: Craftsman reached 35.4% of lists and L.L.Bean 5.4%.
  4. In a study of 112 startups, ChatGPT recognized products by name 99.4% of the time but surfaced them for discovery questions 3.32% of the time (Sharma (opens in a new tab), 2026).
  5. In our consistency study, one ChatGPT answer showed only 57.8% of the brands its five answers to the same question named.

Can a market leader really be absent from AI answers?

Yes, and researchers have recorded well-known brands receiving no recommendations at all. Malthouse and colleagues (opens in a new tab) asked six AI models, including GPT-5.5, Gemini 3.1 Pro Preview and Claude Opus 4.7, a plain request: “I am looking for a [category].” Each model answered 40 times per category, giving 1,200 lists of up to five brands across cordless drills, hiking jackets, coffee makers, cat food and cruises.

Many large, established brands never appeared. The authors list L.L.Bean, Eddie Bauer and REI Co-op in hiking jackets, Craftsman and Black+Decker in drills, Braun and Philips in coffee makers, Freshpet in cat food, and Virgin Voyages and Regent Seven Seas in cruises. Their conclusion for brand managers is blunt: check whether you are recommended at all, not only where you rank.

One caveat matters for reading this. The models were called through their developer interfaces with web search switched off, so they answered from what they learned in training. The consumer apps your buyers use may search the web and give different results.

Why would an AI assistant know a brand but not recommend it?

Because recognizing a name and calling it up for a category are two different tasks. In Sharma’s study (opens in a new tab) of 112 startups from the 2025 Product Hunt leaderboard, ChatGPT recognized products when asked about them by name 99.4% of the time. When asked discovery questions such as “What are the best AI tools launched this year?”, it surfaced them in 3.32% of answers. The study counted a hit whenever the product name appeared in the answer, so the by-name figure may flatter true recognition. Younger companies face the same gap, covered in winning customers as an AI startup.

Malthouse’s team tested the same gap for big brands. They wrote deliberately unrealistic “probes” that borrowed language straight from each brand’s own marketing. With those cues, Craftsman appeared in 81.3% of lists and L.L.Bean in 88.5%.

So the models knew both brands and what they stood for. The authors conclude that low recommendation rates “cannot be attributed simply to an absence of brand knowledge”. The brand was in the model; the ordinary question did not bring it out.

Does the way buyers phrase a question change who appears?

Yes, and often by more than any other factor you can see. Malthouse’s team wrote 20 realistic buyer questions per category, such as a teacher asking for a tough, affordable waterproof jacket for school field trips. Across 240 lists per brand, Craftsman rose from zero to 35.4% of lists, and L.L.Bean only to 5.4%.

Our prompt phrasing study found the same sensitivity in live assistants. Asking the identical question again kept the same first brand 68.0% of the time. Adding “on a tight budget” kept it only 15.3% of the time.

A vendor data set adds a warning about specific questions. Kumar (opens in a new tab), a co-founder of the AI visibility company Ranqo, found broad “best X” questions surfaced a tracked brand about 23% of the time, while specific problem and use-case questions did so about 11% of the time. A brand can be findable for the broad ask and nearly invisible for the specific one. Smaller players face a related question, covered in whether smaller websites get cited.

Is it about your price tier or positioning?

Sometimes, in some categories, but no single rule explains it yet. In drills and hiking jackets, the models in the Northwestern study favored premium brands such as DeWalt, Milwaukee, Patagonia and Arc’teryx, while giving little room to mass-market names such as Black+Decker, Craftsman, Eddie Bauer and L.L.Bean.

Conventional popularity explained the results only in places. In cruises, Carnival and Royal Caribbean, the two brands consumers think of most readily (scored 177 and 162 on Kantar BrandZ salience), were among the most prominent picks. In the other categories, that measure and the models’ picks showed little systematic link, and the premium pattern held in only two of the five categories.

The authors’ practical reading is that what a brand is associated with matters, not only how visible it is. They advise stating a few distinct, relevant points of difference consistently across your own channels and the reviews, news and conversations you can influence. For device makers, whose buyers weigh compatibility and ecosystem, our guide to consumer tech brands on AI shortlists shows how this applies.

Could you appear in some answers but not the one you checked?

Yes, because AI answers change from run to run and from assistant to assistant. In our consistency study, we asked ChatGPT, Gemini and Perplexity the same 20 buyer questions five times each. Of the brands ChatGPT named for a question, 25.2% appeared in all five runs and 36.6% in only one.

Different assistants also disagree. In our four-assistant study of 80 buyer questions, 66.3% of recommended options came from only one assistant, and all four agreed on the first pick for 10.0% of questions.

The flip side matters too. In Kumar’s vendor data, 63.2% of brand, question and assistant combinations were never mentioned in any run. If your brand is missing across many runs and assistants, that is probably a stable gap, not bad luck.

What should you do about it?

Measure whether you appear at all, then find out why the ordinary question leaves you out.

  1. Define your competitors first. List the brands you compete with before you look at AI answers, so you can see who is missing, including you.
  2. Ask plain category questions repeatedly. Use several runs on each assistant your buyers use. One answer is a sample, not the result.
  3. Add the needs your brand is built for. Write questions as your target buyers would, with budget, use and experience. This shows whether your positioning connects.
  4. Test whether the assistant knows you. Ask with your own positioning language. If you appear then but not otherwise, the gap is activation, not awareness.
  5. Make your positioning consistent everywhere. Repeat the same few points of difference on your site and in the reviews, coverage and discussions that describe you.

If you want help running this kind of audit, see our approach to generative engine optimization.

What does the research not tell us yet?

The evidence explains how brands go missing better than it explains how to bring them back.

  • Search was switched off. The Northwestern results come from models answering from training alone. How far web search changes the picture for big brands is untested in that study.
  • The needs-based tests are small. They cover two brands with hand-written questions, and the authors say the approach needs validation across many brands.
  • The price-tier pattern is unexplained. It held in two of five categories, and the authors do not yet know why.
  • No fix has been proven. Whether supplying clearer positioning information raises recommendations is listed by the authors as future research.

Frequently asked questions

Why doesn’t ChatGPT recommend my brand even though it knows it?

Recognition and recommendation are separate. In a study of 112 startups, ChatGPT recognized products by name 99.4% of the time but surfaced them for discovery questions 3.32% of the time.

Does being a market leader guarantee AI visibility?

No. A Northwestern study found brands such as L.L.Bean, Craftsman, Braun and Philips received no recommendations across six AI models for plain category questions.

How do I check whether AI assistants recommend my brand?

Ask unbranded buyer questions several times on each assistant and count how often you appear. In our tests, a single ChatGPT answer showed only 57.8% of the brands that five answers named.

Will more detailed buyer questions bring my brand back?

Sometimes, partly. Detailed questions lifted Craftsman to 35.4% of lists but L.L.Bean only to 5.4%, so detail helps most when your positioning clearly fits the need.

Sources

Free strategy call

Some questions are easier to answer about your own business.

Bring the one that matters most. On a free 30-minute call we’ll take a first look at it and send you a short written read afterward.

Ask about my businessHow we work