Someone asks an assistant who to call. It names you. They search your business name, look at your reviews, and phone you.
In your analytics, that person arrived from a branded search or came in direct. There is no trail back to the recommendation that started it. The channel that produced the job is invisible, and the channel that gets the credit did nothing except be the last click.
This is the central measurement problem of AI-mediated search, and it is not going to be solved soon. Assistants do not pass referrer data the way search engines do. Google's own AI reporting in Search Console currently shows impressions but not clicks. There is no clean attribution path, and anyone selling you one is overstating what exists.
What you can do is find the signal that moves anyway.
Why branded search is the signal
Branded search is people searching your business name rather than your category. "Meridian Dental" instead of "dentist near me".
It works as a proxy for one reason: nobody searches your name unless something put it in their head. They did not wake up knowing it. Someone recommended you, they saw you somewhere, or an assistant named you. Branded search volume is a measure of how often your name is entering circulation.
That property makes it uniquely robust right now. It does not care whether the recommendation happened in ChatGPT, in a Facebook group, or over a fence. It only registers that it happened.
And crucially, it is not distorted by the two things wrecking everything else: crawler traffic does not search your name, and zero-click answers do not suppress it. If anything, an AI-mediated market increases branded search, because the assistant hands over a name rather than a link.
How to actually track it
Google Search Console, Performance report. Filter queries containing your business name. That is your branded impression and click volume. Watch the trend, not the absolute number.
Then filter it out and look at what remains. That is your non-branded, category-demand visibility. These two lines tell different stories and averaging them together hides both.
Watch the ratio. Branded rising while non-branded is flat means your name is circulating without your category visibility improving — which is what being recommended looks like. Non-branded rising while branded is flat means you are being found but not remembered.
Google Business Profile splits this for you already. It separates "direct" searches, where someone looked up your name, from "discovery" searches, where someone searched a category and found you. Most owners never open this report. It is the same distinction, free, and specific to local.
The three-line report worth more than a dashboard
Most agency reporting fails now because it leads with sessions, and sessions have become a number containing several unrelated stories. Replace it with:
1. Branded search volume. Is your name circulating more than last quarter?
2. Non-branded impressions for the queries you care about. Are you visible for the things people search when they do not know you yet?
3. Inquiries, with a source question on the form. Actual work, plus the only first-party data you will get about where it came from.
Add "How did you hear about us?" to your inquiry form as a free-text field, and read the answers. It is unfashionable, it is unscientific, and it is currently more informative about AI-driven demand than your analytics platform is. When people start writing "ChatGPT said you were good", you will know before any dashboard tells you.
What this does not solve
Branded search is a proxy, not attribution. It will not tell you which assistant named you, or which query, or how often you were passed over. It moves for reasons that have nothing to do with AI — a good month of word of mouth looks identical.
So do not over-read a single quarter, and do not build a case on it alone. Pair it with the manual citation test: ask the assistants the questions your customers ask, and record who gets named. That gives you the cause. Branded search gives you the effect. Neither is sufficient by itself.
Being honest about that is more useful than pretending the gap is closed. The businesses that will get this wrong are the ones sold a confident number by someone who does not have one.
What to do this week
- Open Search Console and filter queries by your business name. Note the figure. That is your baseline.
- Filter it out and note the non-branded figure separately. Keep both.
- Open Google Business Profile insights and look at the direct-versus-discovery split.
- Add a "how did you hear about us?" field to your inquiry form, as free text rather than a dropdown. Dropdowns cannot tell you something you did not already think of.
The bottom line
The measurement problem is real and it is not fixed. Sessions have stopped meaning what they used to, and AI recommendations arrive with no return address.
Branded search is the closest thing to an honest signal, because it registers the one thing that matters: whether your name is getting into people's heads. Track it, separate it from non-branded, and ask people directly.
Most businesses are not measuring any of this, which is why so many of them cannot tell whether their visibility work is producing anything.
Want to see where you currently stand across search and AI answers? Request a visibility check.
PresenceKit
Practical growth guidance for small and midsize businesses