Key Takeaways
- Anchor on the number that actually exists. Roughly half of US adults report having used an AI chatbot, and about a quarter use one daily. That is measured. Almost everything else circulating about AI and buying is not.
- Treat any precise "AI referral" percentage as invented. No public dataset reports what share of coaching clients consulted an assistant before enquiring. A confident figure means someone guessed and rounded.
- Notice that fitness sits inside the measured data. One in five chatbot users report using it for medical, diet, or fitness guidance, which puts your category squarely inside the behavior rather than adjacent to it.
- Your analytics will never show you this. Someone who asks an assistant, hears your name, and searches for it arrives looking like direct or branded traffic. The first step is invisible by design.
- Ask new clients how they found you, and write down the answer. One question at intake produces better data about your own business than any published study, because it measures your clients rather than the population.
You have probably seen the statistic. Some enormous share of buyers now consult AI before they contact anyone, quoted to the decimal place, usually by someone selling a fix for it.
That number does not exist. Nobody has measured it, and the studies that get cited for it measure something else.
What does exist is good survey data about how many people use these tools and what for. It supports a real conclusion. It just does not support the precise one you keep being sold.
How many people use AI to find services?
About half of US adults report having used an AI chatbot, and roughly a quarter use one daily. How many use it specifically to choose a service provider has not been measured publicly.
The Pew Research Center survey of 5,119 US adults, fielded in February 2026, puts chatbot use at 49%, up from 33% in 2024. Around 24% use one daily. ChatGPT specifically reaches 44% of adults.
Two of its findings matter more to a coach than the headline. Among users, 42% report using a chatbot to search for information, and 20% report using one for medical advice or diet and fitness guidance. Your category is inside the measured behavior, not adjacent to it.
What is actually measured, and what is inference
Three tiers of confidence sit behind every claim in this area. Keeping them apart is the whole discipline.
- US adults who have used an AI chatbot 49%
- US adults who use one daily 24%
- Chatbot users who search for information with it 42%
- Chatbot users seeking medical, diet, or fitness guidance 20%
- Some share of your enquiries has already asked an assistant
- That share is larger among clients under fifty
- It is growing rather than shrinking
- The exact share of coaching clients who ask AI first
- How often an assistant names any given local business
- What proportion of bookings began in a chat window
In text, so nothing lives only in the figure. Measured: 49% of US adults have used an AI chatbot, 24% use one daily, 42% of users search for information with it, and 20% use one for medical, diet, or fitness guidance. Reasonable to infer: some share of your enquiries has already asked an assistant, that share is larger among clients under fifty, and it is growing. Not published anywhere: the exact share of coaching clients who ask AI first, how often an assistant names any given local business, and what proportion of bookings began in a chat window.
Why nobody can tell you your own number
Because the step happens somewhere you cannot see, and it does not end in a click. The conversation leaves no trace in your reports.
Picture the actual sequence. Someone asks an assistant who helps with a problem you solve. It names three businesses. They recognize one, type the name into a search bar, and land on that website. Your analytics records a branded search or a direct visit, and the moment that decided the outcome is nowhere in the data.
This is not a tracking gap that better software will close. There is no referrer to capture, because most of the time there was no link. The person read a name and remembered it, which is how word of mouth has always worked, and word of mouth has never been attributable either.
So treat any vendor quoting your AI referral percentage with suspicion. What can be observed honestly is the other side of it: whether assistants name you when buyers ask. That is measurable today, and it is the whole premise of the fifteen-minute check.
The numbers, and what each one actually covers
Every figure in circulation measures something specific. The mischief happens when a number about one thing gets quoted as a number about another.
| Figure | What it measures | Source | What it does not tell you |
|---|---|---|---|
| 49% of US adults | Have ever used an AI chatbot, up from 33% in 2024 | Pew Research Center, February 2026 | Nothing about buying, hiring, or choosing a provider |
| 24% of US adults | Use an AI chatbot daily | Pew Research Center, February 2026 | What they use it for on any given day |
| 44% of US adults | Have used ChatGPT specifically | Pew Research Center, February 2026 | Whether they trusted or acted on what it said |
| 42% of chatbot users | Use one to search for information | Pew Research Center, February 2026 | Whether that information included local businesses |
| 20% of chatbot users | Use one for medical advice or diet and fitness guidance | Pew Research Center, February 2026 | Whether they then hired anyone |
| 66% and 23% | Chatbot use among 18 to 29 year olds, and among those 65 and over | Pew Research Center, February 2026 | How your own client age mix compares to the population |
| Any "% of clients who ask AI first" | Nothing verifiable | No public dataset | Treat as invented until a source is named |
The last row is the useful one. When a figure has no source column, that is the finding.
How to measure your own, which is the only number that matters
Ask every new enquiry how they found you, and record the answer verbatim. Within a quarter you will have better data about your business than any survey can give you.
Keep the question open rather than offering a list. "How did you come across me?" produces "I asked ChatGPT for coaches who work with runners and you came up", which is the sentence you are looking for. A checkbox menu produces whichever option is closest, which is usually wrong.
Run the other half at the same time. Ask the assistants what they say about your niche, note who gets named, and date it. Do that quarterly and the two records together tell you whether the picture is moving, which is the approach set out in how often to check.
Worth being clear about the size of the change. This is not the end of search, and a coach whose clients are mostly in their sixties is looking at a slower version of this than one whose clients are thirty. Pew's breakdown by age puts chatbot use at 66% of 18 to 29 year olds and 61% of 30 to 49 year olds, against 42% of 50 to 64 year olds and 23% of those 65 and over. Same shift, very different speed depending on who you coach. The honest framing is in whether people trust AI recommendations and in how clients find a coach in 2026, neither of which argues that everything has changed at once.
What has changed is that a first pass now often happens before any website is opened, and that pass produces two or three names. It does not only happen in a chat window either: Google's guidance on AI features describes generated answers appearing inside search itself, which is the same behavior arriving by a different door. Being one of the named is the whole game, and you can find out where you stand today for nothing.
The free check is twenty-five questions, $0, and a plain-English readout within the hour. It runs as soon as you ask for it.
Questions we hear the most
How many people use AI to find services?
About half of US adults report having used an AI chatbot, and roughly a quarter use one daily, according to Pew Research Center survey data collected in February 2026. How many use it specifically to choose a service provider has not been measured publicly.
How many clients ask AI before visiting my website?
Nobody can tell you that number, and anyone quoting a precise one is guessing. The honest answer is that a meaningful and growing share of your enquiries has already asked an assistant something before reaching you.
Can I see AI referrals in my website analytics?
Mostly no. Someone who hears your name in a chat and then searches for it arrives as direct or branded traffic, so the conversation that produced the visit leaves no trace in your reports.
Is AI actually replacing search for finding a coach?
Not replacing, adding. Traditional search still carries most discovery, and the change worth acting on is that a first pass now often happens in a chat window before any search runs.
Does this matter if my clients are older?
Less, but it is not nothing. Adoption is markedly higher under fifty, so a coach whose clients are mostly in their sixties faces a smaller and slower version of the same shift.
How do I find out if AI mentions my business at all?
Ask it directly. Open a fresh chat, type the questions a prospective client would ask, and record every business it names. That takes fifteen minutes and tells you more than any survey about your own position.
What does the AnswerHalo free check include?
It costs $0. We ask ChatGPT 25 real buyer questions from your niche, then email a plain-English readout within the hour: how many answers named you, who was named instead, and the first fix. It runs as soon as you ask for it.
Does AnswerHalo measure how many clients came from AI?
No, and we will not pretend to. We measure whether assistants name you when buyers ask, which is the part that can be observed honestly. Attributing a booking back to a chat window is not something anyone can currently do.
Your buyers are already asking. Find out what AI tells them.
The free check asks ChatGPT 25 real buyer questions about your niche and sends you the report within the hour, at $0. If it shows you are already getting named everywhere, we will say so in plain words.
Get my free check →