AI Visibility Trackers Are Nearly Useless – And What to do Instead

There are now more than 18 dedicated AI visibility tracking tools on the market.

Profound, SE Visible, Semrush AIO, Otterly, AIclicks, Sight AI, BrandLayer, Brandwatch AI, Authoritas and a dozen more all launched or expanded their AI tracking features in the past 12 months.

The marketing copy across all of them sounds nearly identical. They all promise to show you whether your brand comes up when someone asks ChatGPT or Perplexity a relevant question.

They’re not exactly useless, but for most coaches, they solve the wrong problem.

Here’s what all of them actually do. You feed the tool a list of prompts you want to track. Something like “best longevity coach online” or “recommend a fitness coach for busy professionals.” The tool runs those prompts through various AI engines on a schedule and tells you whether your brand appeared, what position it was mentioned in, and how that compares to competitors.

Some tools, like SE Visible, start at $189 a month for coverage of just two engines, with others still “coming soon.” The more expensive tiers go to $489 a month and beyond.

The fundamental problem is the prompt list.

When you sit down to write a list of queries you want to rank for, you’re doing the same thing SEOs have always done with keywords: guessing at how people phrase things when they search.

With traditional search, that guess was reasonably close to reality because search queries were short and predictable. “Longevity coach online,” “fitness coach seniors” – people typed in fragments.

AI search doesn’t work that way. The whole point of AI tools is that you can talk to them like a person, and people do exactly that.

A prospective client of yours doesn’t open ChatGPT and type “sleep coach.” They type something like: “I’ve been waking up at 3am every night for six months and I feel exhausted all day, is there someone who specialises in this kind of thing?”

Or: “My doctor told me my cortisol is chronically high and I need to change my lifestyle but I don’t know where to start, who could help?”

Or: “I just turned 47, I’m perimenopausal and my sleep has completely fallen apart, what kind of coach should I be looking for?”

None of those queries are on your prompt list. AI prompts tend to be longer, more conversational, and far harder to predict than traditional search queries.

Every person asking asks differently. The same person asking on different days asks differently.

What this means practically is that your tracker is measuring your visibility only for the handful of tidy queries you thought to type in. As one reviewer noted, most AI visibility tools give you a number and stop — you learn that your brand appeared in 14% of tracked ChatGPT answers, but you don’t learn whether the queries anyone is actually typing look anything like the ones you put in the tool.

The criticism has been specific about this: tool reliability issues, the gap between what the dashboard reports and what actually drives results, and general scepticism that “AI visibility optimization” is as tractable as the tools make it sound.

Traffic estimates

The other limitation nobody talks about loudly: traffic estimates from AI tracking tools are modelled, not measured.

Google and the AI engines don’t expose click data. So any “traffic from AI” number in a dashboard is an estimate built on assumptions, and it won’t reconcile with what your actual analytics show.

AirOps put it sensibly: visibility metrics work best as a compass for content decisions, not as an exact revenue counter.

If you run 50 prompts in your niche every month and your brand shows up in progressively more of them, that’s a useful directional signal, just don’t confuse the signal for the full picture.

For coaches specifically, the better use of time and money at this stage is building the underlying presence that drives AI citations in the first place: cornerstone content, off-site placements, a consistent entity across platforms, and content that sounds like real people describing real problems rather than a list of keywords stuffed into sentences.

That work produces results across every unexpected query, including the unpredictable ones you’d never think to put in a tracker.

How to Actually Track AI Traffic Without Paying for Another Dashboard

None of what follows requires a subscription to anything. These are free methods using tools you already have access to.

Google Analytics 4: Building an AI Channel Group

By default, GA4 dumps most AI referral traffic into “Referral” or “Direct” with no label.

A visitor who clicked a link inside a ChatGPT answer looks identical to someone who typed your URL directly. Google introduced a native “AI Assistant” channel in late 2025, but it misses roughly 30% of active AI sources and isn’t retroactive, meaning historical traffic that arrived before the update was recognised stays miscategorised.

The reliable fix is a custom channel group. In GA4, go to Admin, then Channel Groups, and add a new channel.

Name it “AI Traffic” or similar, and set the condition to Session Source matching the following regex:

chatgpt\.com|chat\.openai\.com|openai\.com|perplexity\.ai|claude\.ai|gemini\.google\.com|copilot\.microsoft\.com|deepseek\.com|grok\.com|meta\.ai|you\.com|poe\.com|phind\.com

This catches the platforms responsible for over 99% of measurable AI referral traffic as of mid-2026. Review and update the list quarterly: Grok and DeepSeek were negligible a year ago and are now measurable sources.

Once your channel group is live, the most useful report is landing page by AI source.

It tells you which specific pages on your site are being cited and which aren’t, that’s far more actionable than a share-of-voice percentage in a paid tracker.

A few things worth knowing about what the data shows and doesn’t show: ChatGPT began appending utm_source=chatgpt.com to outbound links in June 2025, so desktop ChatGPT referrals are now reasonably trackable.

Perplexity passes referrer headers consistently from its web interface. Claude and Gemini are inconsistent.

And Google’s own AI Overviews use a noreferrer attribute by design, meaning AI Overview traffic is completely invisible in GA4 regardless of what you set up.

A SparkToro study from January 2026 found that only 12 to 18% of Perplexity citations result in an actual click-through visit, and the rest are zero-click interactions your analytics will never see.

The conversion case for tracking this is strong regardless. Seer Interactive’s analysis found ChatGPT referral traffic converting at 15.9% compared to 1.76% for Google organic on the same site.

When volume is still small, even a few AI-referred visits per month are worth knowing about.

Google Search Console: Finding Conversational Queries with Regex

Search Console won’t tell you whether a query came from AI Mode, a regular Google search, or a user who typed something long and specific the old-fashioned way.

But query length is a reasonable proxy for conversational AI-style search, and the data is sitting there waiting to be filtered.

The logic is simple: traditional Google searches average one to four words, while conversational AI prompts are long.

When you filter your Search Console queries to show only those with ten or more words, what surfaces reads very clearly as prompts rather than search terms.

To do this: go to Search Console, Performance, Search Results. Set your date range to at least six months (from June 2025 onward is most relevant for AI Mode data). Click Add Filter, select Query, choose Custom (Regex), and paste this:

^(?:\S+\s){9,}\S+$

That filters to queries of ten or more words. If your site is smaller and this returns too few results, drop to seven words by changing {9,} to {6,}.

What you’re looking at when the results come back is something like a window into how people actually phrase the problem your coaching practice solves.

Not “sleep coach” but “I keep waking up at 3am every night and I’m exhausted all the time what could help.” Not “burnout coaching” but “how do I recover from burnout when I can’t afford to take time off work.”

These queries read like prompts because they probably are prompts, or at minimum they reflect the same conversational style that people use when talking to AI tools.

Either way, they’re a content brief just like your client support calls. Pages that directly answer these specific questions, in the first paragraph, are the pages that get cited.

A few additional regex filters worth running in Search Console alongside the word-count filter:

For question-led queries (useful for FAQ and content gap analysis):

(?i)^(how|what|why|when|where|who|is|can|should|do)\s

For comparison and consideration queries (high commercial intent):

.*(best|vs|versus|compare|alternative|difference|review).*

For buyer-stage queries:

.*(price|cost|hire|worth it|how much|find|recommend).*

Each of these surfaces a different layer of intent, and together they give you a fairly complete picture of where your prospective coaching clients are in their decision process when they find you.

What the Data Can and Can’t Tell You

Putting both methods together gives you a clearer picture than any paid tracker, but it’s still partial.

GA4 shows you who clicked through from an AI tool after being cited.

Search Console shows you what conversational queries are finding you via Google.

Neither shows you the most common zero-click scenario: someone asks ChatGPT or Claude a question, your name comes up in the answer, they note it down, and they come back to you later via a branded search or by typing your URL directly.

That visit shows up as Direct in GA4. The AI citation that caused it is invisible.

Only 16% of brands currently track AI search performance systematically, which means most of the coaching practices you’re competing with have no idea what’s happening in this channel.

Setting up the channel group in GA4 and running the Search Console filter takes about 20 minutes. It’s not a complete picture, but it’s a real one, and it’s free.

If you want help setting this up, interpreting what your data shows, and what to do about it, – get in touch.