Does ChatGPT recommend your brand? Here's how to find out in 5 minutes

⏰ The 30-second version

• Your customers are asking AI "what should I buy?" — and most brands have never checked whether they show up in the answer.
• You can test this yourself in five minutes. No tool required. The method is below.
• The trick: never type your brand name. Ask the way a customer asks, then see who the AI names.
• Test more than one AI. When we ran 73 well-known beauty brands through ChatGPT, Claude and Gemini, 34% never appeared in any of them — and the ones that did appear often showed up in only one.
• A score of zero isn't a product problem. It's a content-structure problem, and it's fixable.

The question nobody at your company has asked

You know your search ranking. You probably know your ad ROAS to two decimal places. Here's one most teams have never checked:

When a customer asks ChatGPT for a recommendation in your category, does your brand get named?

It matters more than it sounds. Search results give you ten links and a second page. An AI answer gives three or four names and no second page. If you're not in that handful, you don't exist for that customer — and unlike a lost search click, this leaves no trace in your analytics. There's no impression to see, no bounce to explain. You simply never came up.

The 5-minute self-check

Open ChatGPT. Then do this:

Step 1 — Ask like a customer, not like a brand. Never include your brand name. The test is whether the AI produces you on its own.

Write the sentence a real buyer would type:

• "Best moisturizer for dry, sensitive skin?"
• "Recommend a Korean sunscreen for acne-prone skin"
• "What's a good vegan skincare brand?"

Fifteen questions is a solid sample. Ten works. Fewer than five and you're reading noise.

Step 2 — Count. For each answer, note whether your brand appears at all, and which competitors do. Simple tally: appeared / didn't.

Step 3 — Repeat in Claude and Gemini. This step is the one people skip, and it's the one that changes decisions. Models disagree far more than you'd expect. In our test, one brand was the single most-recommended name in ChatGPT (12 of 15 answers) and appeared zero times in the other two. Checking one model and relaxing is how brands convince themselves they're fine.

Step 4 — Read your score.

Appeared in	What it means
0 of 15	You're invisible in AI answers. Most brands land here, including large ones.
1–3	You exist, but only for narrow phrasings. Fragile.
4–8	You're a real candidate in your category. Worth defending.
9+	You're a default answer. Protect this position.

What the numbers usually look like

If you score zero, it's worth knowing you're in ordinary company. We ran this exact test on 73 Korean beauty brands — the kind you'd recognize by name — across all three models:

• 34% never appeared in any of the three, including brands spending heavily on advertising.
• The winner differed by model. Each AI had a different "number one" brand.
• Niches were already claimed. For vegan skincare, one brand came up again and again. For acne, a different one. If you're not the name attached to your niche, you're not in the running for it.

The pattern holds outside beauty. Recognition and AI citation are unrelated systems.

Why brands score zero (it isn't product quality)

An AI doesn't know your brand is famous. It looks for a sentence it can lift and use as an answer. That's the whole mechanism.

Which is why the best-branded companies often score worst. Their copy is atmospheric — polished, emotional, beautiful to a human reader, and useless to a model:

• "Radiance that lasts all day" answers no question anyone typed.
• "Formulated with three hyaluronic acids for dry and combination skin; fragrance-free, safe for sensitive skin" answers several.

You don't have to delete the first line. You add the second one underneath it. That's most of the work.

What to do with your score

If you scored 0: pick one question you failed and write a page that answers it directly — conclusion in the first paragraph, specifics in the body. One page. Not thirty.

If you scored 1–3: find the questions where a competitor appears and you don't. Those are the pages to write next, in order.

If you scored 4+: start tracking. Model updates move these results, and positions you never verified are positions you can lose without noticing.

Do it now, not later

Every week you don't check, customers are asking and getting an answer with someone else's name in it. The test costs five minutes and tells you which of the four situations above you're actually in.

If you'd rather not run fifteen prompts across three models by hand, that's what we built Bulti for — enter a brand name and get the questions you're losing, and what to publish first. Either way: check. Guessing is the only option with no upside.

Honest footnotes

• The 73-brand figures come from our own July 2026 measurement using a fixed 15-question set across ChatGPT (gpt-5), Claude and Gemini. It's a snapshot, not a permanent ranking.
• "Appeared" means the brand name was present in the answer text. Borderline cases exist.
• Results shift when models update. Treat any single check as a reading, not a verdict.
• Scoring zero says nothing about product quality or brand awareness. It's a content-structure signal.