The only honest way to know is to ask it the same questions your buyers would, in a fresh chat, and read what comes back. There are three ways to do that: ask ChatGPT yourself with five real buyer questions, read the sources it cites when it doesn't name you, or run a free check that does both across ChatGPT, Claude and Perplexity at once. A single answer is a snapshot, not a measurement — engines vary run to run, which is why this matters as much as the check itself.
Don't type your company name — type the question a buyer would actually ask. "Is Reflexa good?" tells you nothing about how AI shops for you. Open a fresh ChatGPT chat (logged out or in a new conversation, so history doesn't skew it) and ask five variations of the same buying question:
This is the same method behind the Reflexa AI Visibility Index, built from real Buyer Questions run across ChatGPT, Claude and Perplexity rather than a single branded prompt.
Do this: write down whether you're named, not named, or named but wrong (wrong pricing, wrong feature). Each outcome points to a different fix, covered on the what is GEO pillar.
When ChatGPT doesn't name you, the sources it does cite tell you why. ChatGPT's search-backed answers show the pages it drew on. If none of them are your site, that's the finding: you're not in the pool of pages the engine considers credible enough to quote for that question, whatever the reason — blocked crawler access, thin content on that exact topic, or simply being outranked by a comparison page you don't own.
Rule of thumb: if you're absent from the sources, not just the recommendation, the fix starts with content and access, not brand messaging.
Manually asking five questions on ChatGPT alone misses Claude and Perplexity, which often name different companies for the same question. Reflexa's free check runs the same kind of Buyer Questions across all three engines in about three minutes, and shows the per-engine count — named you in 2 of 5 on ChatGPT, 0 of 5 on Claude — plus the full answer text and cited sources behind each one, not a single blended score.
Shortcut: if you only have three minutes, this is the fastest way to see all three engines side by side instead of repeating Way 1 three times by hand.
The same question, asked twice, can get two different answers — engines are non-deterministic and their web results change between runs. A company named in a ChatGPT answer on Monday can be absent on Wednesday, not because anything on their site changed, but because the underlying search results shifted or the model sampled a different answer. Treating one lucky mention as "we're visible on AI" is the same mistake as judging your Google ranking from a single search on one device.
Do this: ask each question at least three times, ideally on three engines, before drawing a conclusion — the same three-runs, three-engines baseline the Index uses, and what re-checking regularly is built to catch.
None of this requires buying anything — it takes about fifteen minutes:
Bottom line: checking is free and takes minutes — the only mistake is stopping after one question on one engine and calling it an answer.
The free check runs real buyer questions across ChatGPT, Claude and Perplexity — 3 minutes, evidence included.