Key Takeaways
- Accept that most of it is not countable. Only the buyer who clicks a link inside an answer leaves a trail. Everyone who reads your name and comes back later arrives looking like direct traffic.
- Ask the question at the point of contact. A single "how did you come across us" field on your booking form is worth more than any analytics setting, because it captures the part no tracking can see.
- Watch the referrers you do get. Visits from chatgpt.com, perplexity.ai, and similar hosts show up in ordinary analytics. That number is small and real, and it is a floor rather than a total.
- Measure the cause, not just the effect. You can count how often you are named in answers to your buyers questions. That moves before the client count does and it is fully countable, which makes it the better dial to watch.
- Distrust any tool promising exact AI attribution. Nobody can see inside a conversation you were discussed in. A vendor quoting a precise share of clients from AI is modeling it, and the model deserves your skepticism.
A new client mentions on the call that ChatGPT suggested you. You go looking in your analytics for the visit that proves it, and there is nothing there. They show up as direct traffic, on a Tuesday, from nowhere.
That gap is not a setup problem. It is a structural feature of how these conversations work, and no amount of tracking configuration closes it.
Here is what is genuinely countable, what is permanently invisible, and the number worth watching instead.
Can you see how many clients come from AI answers?
Partly. You can count the buyers who clicked a link inside an answer, because those visits arrive with an identifiable referrer. You cannot count the ones who read your name and found you some other way, and that is the bigger group.
The honest framing is that clicked-through traffic is a floor, not a total. Whatever number you can see is real, and the true figure is higher by an unknown amount.
Anyone who tells you otherwise is selling a model, not a measurement. Nobody outside the conversation can see what was said in it, which is a limit on the whole category rather than a shortcoming of your setup.
Why the trail breaks where it does
Analytics records arrivals, and an AI answer is not an arrival. If the buyer does not click a link inside the answer, nothing about that conversation ever touches your site.
Even the buyers who do act often break the chain themselves. They read your name, then search it, so the visit gets credited to search. Or they note it down and come back three days later, which lands as direct traffic. Both were caused by the answer, and neither carries any trace of it.
This is exactly the invisible research covered in how many clients ask AI before visiting your website: most of the deciding happens before you have any record that a person exists.
The same five steps happen in all three routes. Solid steps are the ones that still leave a record you can read afterwards, and outlined steps are the ones that do not. Only the left-hand route stays readable to the end, which is why the visible number is a floor. The three routes are illustrative rather than measured.
What you can genuinely measure today
Four things, in descending order of reliability: referred visits from assistant hosts, self-reported answers on your forms, unusual phrasing in inbound enquiries, and changes in direct traffic that line up with nothing else.
The first is the only one that is properly countable. Visits referred from hosts like chatgpt.com or perplexity.ai appear in ordinary analytics as referrals, and you can read them the same way you read any other source. Treat that as your confirmed minimum.
The fourth is the weakest and worth naming so you do not lean on it. A rise in direct traffic is consistent with AI-driven discovery and equally consistent with a podcast mention, a newsletter, or a seasonal pattern. It is a hint, never evidence.
| Signal | Where you find it | What it proves | What it misses |
|---|---|---|---|
| Referrals from assistant hosts | Your analytics, under referral sources | Somebody clicked a link inside an answer | Every conversation where nobody clicked |
| "How did you hear about us" answers | Your booking or intake form | What the buyer believes happened | People who skip the field or misremember |
| Enquiries quoting an assistant | Your inbox, in the buyer's own words | A specific, traceable instance | Everyone who does not mention it |
| Unexplained direct traffic | Your analytics, direct channel | Nothing on its own | Any other cause of the same pattern |
| Mentions in AI answers themselves | A repeatable set of test questions | Whether you are being named at all | Whether anyone acted on it |
The one question that beats every tracking setup
Add "how did you come across us?" to your booking form, with an explicit option for asking ChatGPT or another AI assistant. One optional field outperforms every analytics configuration available to you.
It works because it captures the part of the journey that leaves no digital trace. The buyer knows what happened. Nothing else in your stack does, and asking them costs one line of a form.
Two caveats worth holding. People misremember, so a self-reported answer is a signal rather than a fact. And listing the AI option explicitly matters: given a blank box, most people write "Google" for anything that involved a screen.
Measure the cause instead
Since the effect is not fully countable, watch the cause. How often you are named in answers to the questions your buyers actually ask is measurable, repeatable, and entirely within your ability to check.
It is also the earlier signal. Mentions move first, and client counts follow later, so a rise in how often you get named is the thing that tells you the work is landing before the bookings confirm it. The method for tracking that by hand is in how to track your AI visibility without buying software, and the way to size up what invisibility is costing you is in is being invisible to AI costing me clients.
Keep the two numbers separate in your own head. Mentions are what you can control and count. Clients are what you actually want. Confusing one for the other is how people end up either overstating a result or dismissing real progress because the booking calendar has not caught up yet.
Google's guidance on its own AI features in search is candid that there is no separate ranking system to game and no attribution layer to install, and Perplexity's help center documents citations shown to the reader rather than any reporting sent to the cited business. Neither company offers publishers a view into who was named where. That absence is the reason this post exists.
If you want the cause measured for you, the free check runs 25 real buyer questions from your niche through ChatGPT at $0 and emails a plain-English readout within the hour. It runs as soon as you ask for it.
Questions we hear the most
Can you see how many clients come from AI answers?
Not completely. You can count visitors who click a link inside an answer, because those arrive with a referrer, but buyers who read your name and look you up later are indistinguishable from any other visitor.
Does Google Analytics show traffic from ChatGPT?
It shows visits referred by hosts such as chatgpt.com or perplexity.ai when someone clicks through from an answer. It cannot show conversations where you were named and nobody clicked, which is the larger share.
Why does AI traffic show up as direct?
Because no click happened. When a buyer reads your name in an answer and types your site in later, or searches your name, the visit carries no signal that an assistant was involved.
How do I add a how did you hear about us question?
Add one optional free-text or multiple-choice field to your booking form or intake form, with an explicit "asked ChatGPT or another AI assistant" option. Self-reported answers are imperfect and still the best signal available.
Can any tool track AI attribution precisely?
No tool can see inside a private conversation, so precise attribution is not available to anyone. Products claiming an exact figure are estimating from models and assumptions, which is worth knowing before you plan around the number.
What does AnswerHalo measure instead of client attribution?
It measures how often you are named in answers to real buyer questions in your niche. That is the cause rather than the effect, it is fully countable, and it moves before your client numbers do.
What does the free 25-question check include?
It runs 25 real buyer questions from your niche through ChatGPT and emails a plain-English readout: your score, who gets named where you are not, and the first fix to make. It costs $0.
Can I run the free check today?
Yes. It runs as soon as you ask for it, and the readout lands within the hour between 7am and 5pm Central. Adding one "how did you hear about us" field to your booking form takes ten minutes and can happen today.
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 →