Direct answer
To know whether your company is visible in ChatGPT (or Gemini, Perplexity…), a surface check isn't enough. Typing your name once gives you an impression, not a reliable answer, because the AI responds differently each time and depending on whether or not it searches the web. To really know, you have to query repeatedly, on queries representative of your field, comparing modes and models — and measure how often you're cited, at what rank, and who shows up in your place.
The problem
It's the first instinct, and a healthy one: you open ChatGPT, type "best companies in [my industry]," and see if your name comes up. It appears? Relief. It's absent? Worry.
Except this test is barely better than flipping a coin. Ask the question again in an hour, the answer may change. Ask it without letting the AI search the web, it changes again. You have an impression, taken at one given moment, in a mode you don't control. That's not an answer to "am I visible."
The idea to grasp
"Being visible in ChatGPT" isn't a binary question (yes/no), it's a question of degree and conditions. To answer it seriously, you have to break it down.
Visible on which questions? You may show up on "best brands of X" but not on "cheap X" or "X for beginners." Visibility depends on the wording. You have to test a range of real queries from your field, not just one.
Visible in which mode? The AI may cite you when it searches the web, and not when it answers from memory (or the other way around). Without distinguishing the two, "visible" means nothing precise.
Visible how often? Across 20 queries, do you appear 2 times or 18 times? That's the difference between an occasional presence and an established one.
Visible at what rank? Cited first or buried at the end of the answer? Rank changes everything.
Visible against whom? Which brands does the AI cite in your place? Often, you discover competitors you weren't keeping an eye on.
A real answer to "am I visible" is therefore a set of measurements, not a yes/no. And those measurements require repetition and control over the conditions — exactly what a manual check doesn't provide.
What you hear everywhere
"I tested it, my company shows up, so I'm visible." You showed up once, in a given mode. That tells you neither the frequency, nor the rank, nor what happens in the other mode. One swallow doesn't make a summer.
"I don't show up, so the AI doesn't know me." Maybe you don't show up on that wording, in that mode. Reword it, switch modes, and the picture can be very different.
"A tool gave me my visibility score, so it's settled." Computed how, over how many queries, in which mode? A score without those details is an impression dressed up as a number.
My stance: only the facts. And a visibility fact is a repeated and contextualized measurement — not a glance.
My take: turning "am I visible" into measurements
From here on, the register changes: we describe the instrument.
Answering "is my company visible in ChatGPT" seriously requires a method:
- A range of queries representative of your field (not a single wording).
- Repetition (n=20) to measure a frequency, not a stroke of luck.
- Distinguishing the modes (with / without web search) and the providers (ChatGPT, Gemini, Mistral…).
- Indicators computed mechanically: presence frequency, average rank, share of voice, competitors cited.
- Dating and sealing, to compare over time.