How to Check if ChatGPT Recommends Your Brand (Free, 15 Minutes)

More buyers now start their research inside an AI answer than most brands realise: 53% of professionals globally use AI search tools for research (Stanford AI Index 2026), and 38% of Google results in Singapore already show an AI Overview. So the question every founder eventually types is some version of: does ChatGPT recommend us? Most people test this wrongly, get a false answer in either direction, and make decisions on it. Here is how to test it properly, for free, in about fifteen minutes.

Step 1: Ask Buyer Questions, Not Your Brand Name

Searching your own brand name is the most common mistake. If you ask ChatGPT "what is [YourBrand]?", it will usually summarise your website back to you — that is retrieval, not recommendation, and it tells you nothing about whether you appear when a buyer who has never heard of you asks for options. Instead, write down five to ten questions a real buyer would ask before they know you exist: "best [category] in [city]", "which [product type] should I buy for [use case]", "[competitor] alternatives", "is [category] worth it". These buyer-intent questions are the surface where recommendations are won or lost — and where your competitors may already be winning.

Step 2: Repeat Every Question at Least Five Times

A single answer is a coin flip, not a measurement. Generative engines are probabilistic: the same question, asked twice in fresh sessions, routinely returns different brand lists and different citations. Independent research on AI answer stability has found source overlap between same-day repeat runs can drop below half. In our own testing we treat anything below five runs per question as anecdote. Practical protocol: open a fresh chat (or incognito session) for each run so earlier turns cannot contaminate the answer, ask the identical question, and tally results as a frequency — "named in 4 of 5 runs" — never as a yes/no. If a vendor or a tool ever shows you a single screenshot as proof of AI visibility, apply the same scepticism.

Step 3: Test Each Engine Separately — They Do Not Agree

"AI visibility" is not one number. ChatGPT, Perplexity, Google AI Overviews and Gemini retrieve from different indexes and prefer different sources: published analyses of hundreds of millions of citations show ChatGPT leaning on Wikipedia and authority sites, Perplexity drawing heavily on Reddit, and AI Overviews favouring Reddit and YouTube — with only around one in ten domains cited by both ChatGPT and Google for the same query. We see the same divergence in our own probes: in August 2026 we asked two independent engines which Singapore GEO agencies they would recommend, and their shortlists agreed on almost nothing. The practical consequence: run your question set separately on ChatGPT (with web search on), Perplexity, and Google (watching for AI Overviews), and keep three separate scorecards.

Step 4: Record Three Things Per Run

For each run, record: named — did your brand appear in the answer text at all; cited — did the engine link your domain as a source; and who instead — which brands and which source pages filled the answer. The third column is the one most self-tests skip and the one that matters most, because it is your action list. If the engines keep citing a "best of" listicle, a Reddit thread, or a review platform instead of anyone's own website, that page — not your homepage — is the surface you need to appear on. In our published dataset of 50 cross-border DTC brands (DOI 10.5281/zenodo.22103178), the overwhelming pattern was exactly this: engines answer buyer questions from third-party surfaces first, brand-owned pages second.

Step 5: Read the Result Honestly

Four outcomes, four different problems. Named and cited consistently: you are winning — now defend it, because answers refresh continuously. Named but never cited: engines know you exist (usually from third-party mentions) but will not send traffic; your own pages are likely hard to parse or thin on citable facts. Cited on brand questions but absent on buyer questions: the tautology trap — you only appear when the asker already typed your name, which converts nobody new. Absent everywhere: usually a structural problem — the engines' preferred sources (listicles, communities, review sites, Wikipedia-grade references) simply do not mention you, and no amount of homepage rewriting fixes that alone.

The Traps That Invalidate Self-Tests

Five traps produce false readings. Logged-in contamination: your ChatGPT account has memory and custom instructions; it knows who you are and flatters accordingly — test logged out or in a clean account. Asking once (covered above — a coin flip is not a measurement). Leading prompts: "why is [YourBrand] the best?" manufactures the answer you want. API-vs-app confusion: results from the developer API and the consumer app overlap far less than people assume; what your buyers see is the app, so the app is what counts. One-time testing: answers drift as engines refresh sources — a clean bill of health in March can be an absence by June, which is why serious measurement is a cadence, not an event.

What to Do With a Bad Result

The fix follows the diagnosis from Step 5, but the general order of operations is stable. First, make sure the basics are not blocking you: AI crawlers allowed in robots.txt, content server-rendered, key facts stated in plain extractable prose. Second, restructure your money pages so each buyer question has a self-contained, quotable answer block with concrete numbers — peer-reviewed research (KDD 2024) found adding statistics, quotations and cited sources lifted content visibility in AI answers by up to 41% on position-adjusted measures. Third — and usually decisive — earn presence on the third-party surfaces your engines actually cite: the listicles, communities and review platforms from your "who instead" column. Brand mentions on those surfaces correlate with AI visibility far more strongly than backlinks do in every major study published since 2025.

Frequently Asked Questions

How do I check if ChatGPT recommends my brand?

Ask ChatGPT the questions your buyers ask before they know your name — "best [category] in [city]", "[competitor] alternatives" — in fresh, logged-out sessions. Repeat each question at least five times and record how often your brand is named and whether your domain is cited as a source. Report the result as a frequency ("named 3 of 5 runs"), and repeat the test on Perplexity and Google AI Overviews separately, because engines cite largely different sources.

Why does ChatGPT give different answers each time?

Generative engines are probabilistic systems: each response is sampled, and retrieval results shift as sources refresh. Same-day repeat runs of the same question routinely return different brand lists, with source overlap sometimes below half. That is why single-screenshot "proof" of AI visibility — positive or negative — is unreliable, and why measurement protocols use repeated runs and report frequencies.

Can I pay to be recommended by ChatGPT?

No. There is no advertising product that places brands inside ChatGPT, Perplexity or Google AI Overview answers, and any vendor selling a guaranteed "AI ranking" is selling something they do not control. What can be influenced is the evidence the engines retrieve: machine-readable pages with citable facts, and presence on the third-party sources engines prefer to cite. That work is measurable, but it is optimization, not placement.

Or Have It Measured Properly

The manual method above is a real measurement — it is a lighter version of exactly what we run as a service. Canlah's probe fleet runs your buyer questions on a weekly or daily cadence, in real browser sessions, across ChatGPT, Perplexity, Gemini and Google AI Overviews, with every answer timestamped and archived so any result can be re-run in front of you. If you want the baseline done for you, request an AI visibility audit or start with the free 48-hour snapshot described on our pricing page. To see how this methodology ranks an entire market, read our probe-data ranking of Singapore GEO agencies.