Branded vs category AI visibility: the naming gap

Two different questions, and a gap between them

Ask an AI assistant about a brand by name and most assistants with any material on that brand can say something about it. Ask the same assistant an unprompted category question, “which sunscreen should I buy”, with no brand named, and the set of brands it offers is usually far smaller. That second, harder test is the one that matters for a shopper who has not yet decided what to search for.

An India-focused audit of D2C brands by NeuGenM put a figure on that gap: across the brands it checked, assistants named the brand correctly in about 96.8% of prompts that asked about it by name, against about 10.6% of prompts that asked an unprompted category question with no brand named.¹

The gap, as NeuGenM reported it

Why the category question is the one that matters

A shopper who already knows your brand and types your name into an assistant has already found you; the assistant is confirming a decision made elsewhere. The shopper who types a category question with no brand in mind is the one an assistant can send towards a rival instead, and that is the moment a brand with a name-only presence loses out silently, with no missed search to notice.

The gap NeuGenM describes says the industry pattern, not your brand’s own figure. A brand with a strong branded result can still have a weak category result, and the only way to know where your brand sits is to ask the category question yourself, without your own name in it. A brand new to a category and a well-established one can both show a wide gap, for different underlying reasons, so the figure alone does not tell you which kind of problem you are looking at.

Checking your own branded and category presence

You can run a first version of this check by hand, the same way as any AI visibility check.

1. Ask two or three assistants a prompt that names your brand directly, and record what each says.
2. Ask the same assistants the unprompted category question a shopper would actually type, with no brand named, several times.
3. Record whether your brand appears in the category answers at all, where in the list, and what reason, if any, the answer gives.
4. Repeat on a schedule. Assistant answers change without notice, so one run is a snapshot, not a trend.

Write down your category questions the way an actual shopper would type them, not the way your own team talks about the category internally. A shopper rarely uses the precise category label a brand uses in its own marketing, and a question phrased in marketing language mainly tests whether the assistant has read that marketing, which is a narrower and less useful question than the one you actually want answered.

It is also worth asking the category question with small variations, a stated budget, a stated need such as sensitive skin or a stated occasion, since these can change which brands an assistant offers even though the underlying category has not changed. A brand that appears for a broad category question but never for the specific variations its actual customers care about has a narrower gap than the headline result suggests, and a wider one hiding underneath it.

For the fuller method, including how to word questions so they read like a shopper rather than a search for your own name, see how to check whether AI assistants name your brand and share of AI vs share of search.

What one attributed figure can and cannot tell you

A single industry figure such as NeuGenM’s is useful for framing the size of a gap, and it is not a promise about any one brand, any one category or any one assistant. The brands an audit like this covers, the assistants it queried, and the exact wording of its prompts all shape the number it reports, and a different method could reasonably report a different one. Treat the branded-versus-category gap as a pattern worth checking for, not a target to hit.

A category result can mean two different things. An assistant can fail to name your brand because it has little material on your brand to draw from, or because the material exists but is thin, inconsistent or hard for the assistant to read. The first is a visibility problem in the broadest sense; the second is closer to the kind of gap that a check of your own pages and of which sources AI assistants cite can find. A category check on its own does not tell you which of the two you are looking at, which is another reason to treat it as a starting point rather than a verdict.

A worked example, with an invented brand

This example is an illustration. The brand and the results are invented, and nothing in it was measured.

Suppose a fictional brand, “Vantra”, sells a range of protein snacks. Asked directly, “tell me about Vantra protein bars”, three assistants each give a reasonably accurate answer: ingredients, pack sizes and where the range is sold. Asked instead, “which protein bar should I buy in India”, with Vantra’s name left out, none of the three assistants mentions Vantra in their first answer, and one mentions it only after a follow-up question naming the price range Vantra sells in.

Read on its own, the branded result would suggest Vantra’s AI presence is fine. The category result says otherwise: shoppers who have not already decided to look for Vantra by name are unlikely to be offered it. That is the gap this guide describes, made concrete for one brand, and it is exactly the comparison a real check, run on your own brand and your own category questions, is meant to surface.

Why the gap tends to be this shape

It is worth understanding, at least in outline, why a branded question is so much easier for an assistant than a category one. When you name a brand directly, the assistant has a specific target to look up or recall: whatever it absorbed during training about that name, and whatever it can find by searching for it if it can search the web. The task is closer to lookup than to judgement.

An unprompted category question asks something harder: out of everything the assistant knows or can find about a category, which few names does it choose to surface first, and why those and not others. That is closer to a ranking or a recommendation task, and it rewards brands that are widely and consistently discussed across many independent pages, not just brands with a page of their own that states the facts correctly. A brand can have an accurate, complete website and still lose the category question if almost nothing beyond its own site discusses it.

This is also why the gap NeuGenM describes is closer to an industry pattern than an individual brand failing: most D2C brands invest heavily in their own product pages, which helps the branded question, and far less in being discussed across the wider set of reviews, comparisons and forum threads an assistant draws on for a category answer.

Where Rewalto fits

Run the branded and category questions separately, and save every answer. That gives your team a record of its own question set rather than a claim based on an industry average.

A website readiness check answers a different question from an assistant-answer review. A readiness check reads your own site; an answer review reads what assistants say. For Rewalto’s current products, see Data.

Common questions

Notes

1. NeuGenM, “D2C AI/GEO/AEO Audit: India, 2026”. Cited here for problem framing, attributed by name, not a Rewalto measurement and not independently verified by us. Back to text