ChatGPT
61% of the scoreThe one most of your customers are actually using.
The whole formula, in public
Your score answers one question: when someone asks AI who to call, how often is it you? It takes four steps to find out. Here is every one of them, so you can check our work and ask the same questions yourself.
Step 1
Before we ask AI anything, we read your website and your Google listing. What do you sell? Who pays you for it: homeowners, businesses, patients? Which cities do you work in? Every question after this depends on getting that right.
It sounds simple. It is also where one wrong guess ruins everything after it. A company that helps owners fight their insurance company says “insurance” on every page. Go by that one word and you would ask AI for the best insurance company, and score the business against the very companies it fights. So we read your site the way a person would. Still not sure? We ask you first.
Step 2
Next we look up the real Google searches people make for what you do, and how many people make each one every month. Then we turn them into the questions people actually ask AI.
Not “is my business online”. The real ones, typed the way people type them. Who should I call to fix my furnace near me? Most people do not type their city. They say “near me”, and ChatGPT fills in the city from where they are.
So when we ask ChatGPT, it sees the question coming from your city, just like a real customer’s. The other three cannot tell where a question comes from, so for them we write your city into the question. And a search for your own name never counts: someone who types your name has already picked you.
If your site is too new to show up in any searches yet, we start from the everyday questions people ask about your line of work.
Step 3
Word for word, the way a customer would, on ChatGPT, Gemini, Claude and Perplexity. You see every answer, word for word, in your report.
We ask all four because they disagree. We tested it: we asked the same 40 questions across all four, in eight trades, and compared who each one recommended. They agreed with each other between 4% and 14% of the time.
So asking one does not give you a rough idea of the answer. It gives you a different answer. One real business we measured was named in five out of five Gemini answers and one out of five Perplexity answers. Ask only Perplexity and you would call that business invisible, and you would be wrong.
The one most of your customers are actually using.
Comes built into Android phones and Google accounts, so people land in it without choosing it.
Smaller, and leans business and professional.
The smallest of the four. Under one customer in twenty.
The weights are how many people use each one, not how much we like it. Gemini hands out more names per answer than any of the four and earns nothing extra for it, because a recommendation nobody sees is worth nothing.
Step 4
Being named is not all or nothing. Top of the list is the call. Fifth on the list is still a call sometimes. Both are worth more than being left out. So each answer is worth something between 0 and 1:
The whole prize. This is the call.
You are in the conversation.
Still named, and still gets calls. Never worth nothing.
It found you and chose someone else. A different problem, and a narrower one.
For that question, you do not exist.
Average those within each assistant, weight the four averages by how many people use that assistant, multiply by 100. That is your score, and because it is an average it is a percentage: it cannot go above 100 however many questions we ask. There is nothing else in it.
An answer that recommends nobody. AI sometimes replies with general advice and no names at all. That is not a race you lost, it is a race that never happened, so we leave it out entirely. One of the four does this in up to six answers out of ten depending on the trade. Counting those against you would be measuring its mood, not your business.
Anything we could not check. If your Google listing has no website on it, we cannot look at your site, and the report says exactly that instead of scoring you zero for it. A number we did not measure is never printed as a nought.
You can be first on Google and never named by AI at all. We have measured businesses in exactly that position. If Google rankings moved this score, the score would hide the very problem you hired us for.
The same goes for a tidy Google listing. We track it, we show it to you as your groundwork, and we work on it, because it is one of the things AI reads. But it sits beside your score, never inside it. A business AI ignores cannot buy a good number with a complete profile.
Ask AI the same question twice and you can get two different lists. That is why we ask several questions across all four and report a frequency, named in 9 of 20 answers, rather than a yes or a no. One answer is an anecdote. Twenty is a measurement.
The free report and the report a paying customer gets use the same formula, on the same scale. The paid one simply asks more questions, more often. The number you see before you buy is the number we then hold ourselves to.
Open ChatGPT and ask the question a customer would ask. Nothing on this page is anything you cannot verify in two minutes, and if your result differs from ours we would rather hear about it than not. Answers vary run to run, so ask it a few times.
Something here look wrong? support@thedohenyco.com. If we have made a mistake we will fix it and tell you what changed.
Get your free report and see where you stand.