The most common AI visibility mistake we see has the same shape every time. Someone Googles their business, checks ChatGPT, sees themselves mentioned, and relaxes. ChatGPT is one of four major assistants doing recommendation work right now. The other three behave differently, and they're not optional.
The two-thirds finding
Across a scan of 1,296 answers about 36 Northern Irish businesses, the four major assistants agreed on whether to mention a given business only two-thirds of the time. The disagreement rate, where at least one assistant reached for a different name than the others, was 33%.
That means on roughly one in three buyer queries, the answer your prospect gets depends entirely on which AI they happen to ask. Win the wrong assistant and you can still be invisible on the other three. Win the right one for the wrong audience and you're invisible to the audience that actually buys.
The four are not the same product
Gemini was the most generous in the scan, naming a business on 25% of open buyer questions. Claude came in at 18.9%. Perplexity, 17.2%. ChatGPT was the most reserved at 15.6%. Same businesses, same questions, four different answers.
Each one weighs sources differently. Perplexity leans hard on editorial citations and tends to read more like a search engine. Claude favours measured, multi-source consensus. ChatGPT reaches for fewer, higher-authority signals. Gemini surfaces more breadth and reads more like a directory. A strategy that wins one can lose the other three.
Why this matters for who you're trying to reach
Different audiences are choosing different assistants. ChatGPT is closest to the mainstream default for white-collar work and a lot of consumer search. Gemini ships with every Pixel phone and is increasingly woven into Google's products. Perplexity skews toward research-leaning, AI-fluent users. Claude has the deepest hold on developers, writers and Anthropic-adjacent professional segments.
If your buyers are developers and you only check ChatGPT, you're checking the wrong thing. If they're tourists picking a hotel, Gemini is going to matter more than Claude. The audience-by-assistant mapping is rough but real, and ignoring it is the most common gap in early AI visibility work.
What to do about it
Check all four. Always all four. The asymmetry between them is the whole point. Single-assistant checks understate the gaps you actually have, because the assistants where you're invisible are exactly the ones you wouldn't have thought to check.
Then look at the gap pattern. If you're strong on one and weak on the others, the difference will tell you which sources you're missing. If you're weak on all four, the work is more structural. The pattern points at the fix.
Read the full research for the per-assistant breakdowns and how each model weights sources differently in the recommendation pass.
See where you stand
Run the free AI visibility check. It runs your domain across all four major assistants in about a minute and shows you where the gaps are, assistant by assistant.