Try this. Open ChatGPT and Gemini side by side. Type the same buyer-intent prompt into both — something like "best HVAC company in Tucson" or "top commercial roofing contractors near Charlotte." Don't optimize the wording. Just ask the same plain question a real customer would ask.
Then read the two lists of named businesses.
If your experience matches what I keep seeing, the lists barely overlap. ChatGPT names five companies. Gemini names five. Maybe one shows up on both. Sometimes none do. The query is identical. The intent is identical. The answers come from different rosters entirely.
That's the whole point of this post. There is no single "AI visibility" number for your business. Each assistant is its own channel, with its own list, and the only way to know where you stand is to measure them in parallel.
A single score hides the thing you need to know
It's tempting to want one number. "We're visible in AI" or "we're not." It feels like a search ranking — one position, one source of truth.
But the assistants don't share a source of truth. ChatGPT, Claude, Gemini, Perplexity, and Google AI each pull from different data, weight it differently, and phrase results differently. They are built by different companies on different models with different training cutoffs and different live-retrieval setups. Expecting them to agree is like expecting a radio buy and a paid-search buy to deliver the same customers because they're both "marketing."
So when someone hands you a blended "AI visibility score," ask what it's averaging. If you show up strong in ChatGPT and invisible in Gemini, a blended 50% tells you nothing useful. It hides the fact that half your prospects — the ones who happen to use Gemini — are being handed your competitors and never hearing your name.
A blended score is a comfortable fiction. The real picture is five separate scoreboards.
Why the lists diverge
You don't need to know the internals to plan around this, but a few reasons are worth naming:
- Different underlying data. Gemini leans heavily on Google's index and Google Business signals. ChatGPT pulls from its training data plus whatever it retrieves live, which is a different mix. They're literally reading different things.
- Different recency. One assistant might know about a company that rebranded last quarter; another might still be working from older data and name a business that's since closed or merged.
- Different phrasing of "best." Each model has its own internal sense of what makes a business worth naming — reviews, mentions across the web, directory presence. Those weightings aren't the same, so the rankings aren't the same.
The result: the same question, asked the same way, returns different competitors. Not slightly different. Often a completely different set.
What this looks like in practice
Picture a local-services business — say a plumbing company that's been running Google Ads for years and knows its cost per lead cold. The owner asks ChatGPT "who are the best plumbers in my city" and sees their own name in the top three. Relief. They conclude they're in good shape with AI.
Then they ask Gemini the same thing and they're nowhere. The list is three competitors and two directory sites. A customer using Gemini in that market would never encounter this business at all.
Both of those facts are true at the same time. Neither one is "the" answer. If the owner had only checked ChatGPT — which is what most people do, because it's the one they've heard of — they'd have walked away with a false sense of security and no idea that an entire channel was running without them.
This is the same discipline you already apply to paid acquisition. You don't report one blended CPA across Search, Social, and radio and call it a day. You break it out by channel because the channels behave differently and you'd make bad decisions otherwise. The assistants deserve the same treatment.
Measure first. That's the whole job right now.
I want to be careful here, because this is where a lot of conversation jumps the rails. The useful work at this stage is not changing what the assistants say. It's finding out what they say — accurately, per assistant, and tracked over time so you can see it move.
That distinction matters. You can't have a sensible conversation about influencing anything until you have a baseline. If you don't know that you're strong in ChatGPT and absent in Gemini, you can't prioritize, you can't set a target, and you certainly can't tell whether anything you try later made a difference. Measurement is the step before optimization, and skipping it is how people end up chasing a problem they never actually confirmed.
So the practical move is simple and you can start it this afternoon:
- Write down three to five prompts a real buyer would type. Buyer-intent, not your brand name.
- Run each one through ChatGPT, Claude, Gemini, Perplexity, and Google AI.
- Record which businesses get named in each, and in what order.
- Do it again next month and compare.
You'll have five lists, not one. Where you sit on each one, and how that shifts over time, is the real signal. (This is exactly what tools like LLMClarity automate — tracking share-of-voice and competitive position per assistant, over time, so you're not running the grid by hand. Pricing runs from $89 to $800. But the manual version above works fine for getting your bearings.)
The takeaway
ChatGPT and Gemini name different competitors for the same query because they are different channels, not one channel with some noise. A single "AI visibility" number papers over the only thing worth knowing: where you actually stand, assistant by assistant.
Run the prompts in parallel. Look at the gaps. You'll almost certainly find at least one assistant where your strongest competitors show up and you don't — and that's a problem you can't see, let alone address, until you've measured it.