I want to describe a meeting instead of a statistic, because the statistic only means something once you've sat through the meeting. The client opens their laptop, pulls up a screenshot of a ChatGPT answer that mentions a competitor and not them, and asks the question you knew was coming: "Are we showing up in this?" You have an answer for their Google rankings. You have an answer for their backlinks. For AI search visibility, you have a shrug dressed up in professional language, and everyone in the room knows it.
That scene isn't a hypothetical I built to open this piece. It's close to verbatim from a conversation on our own team a few weeks back, and it turns out we're nowhere near alone in having it. A survey of 494 agency professionals published by Search Engine Land on July 24, 2026, put a number on a feeling most of us in this industry have been carrying around quietly. We went looking for the rawest version of this story where it usually lives, in the practitioner threads on Reddit, and came up empty; the relevant subreddits weren't reachable today across more than a dozen query attempts. So instead of a thread, you're getting a survey. It's a more honest source than it sounds, because it's not a handful of loud opinions, it's nearly 500 agencies describing the same wall.
The client call every agency is having right now
Here's the thing about that meeting I opened with: it's not really about ChatGPT. It's about trust. A client who's been paying for SEO for three years has a mental model of what that buys them, rankings move, traffic follows, leads show up in the CRM with a channel attached. AI search visibility breaks that mental model in a way they can feel before they can articulate it. They saw their competitor mentioned in an AI answer. They didn't see themselves. They want to know why, and "we're working on it" isn't an answer, it's a stall.
What makes this uncomfortable rather than just difficult is that the client isn't wrong to ask. AI Overviews and chat-native search are pulling real query volume away from the ten blue links, and the businesses paying us to be found have every right to expect we can tell them whether they're being found in the places their buyers have started looking. The problem isn't the question. The problem is that most of our reporting stacks were built for a world where a landing page, a ranking position, and a conversion event could be strung together into one clean line, and AI search doesn't hand you that line.
What the survey actually found
The numbers come from AgencyAnalytics' 2026 Marketing Agency Benchmarks Report, based on responses from 494 agency professionals, as covered by Search Engine Land's July 2026 agency survey. Four numbers matter most, and they fit together in a way that should worry anyone running an agency reporting stack in 2026.
AgencyAnalytics CEO Joe Kindness put the client-side pressure plainly: "Clients want to know if they show up in ChatGPT, they want answers the moment they ask, and they expect proof tied to revenue." Read that quote next to the 48% and 47% figures and the shape of the problem gets obvious fast. Two-thirds of agencies are being asked, insistently, for a service that roughly half of them can't currently instrument. That's not a niche capability gap. That's the industry's most-requested new line item running ahead of its most basic reporting infrastructure.
It's also worth sitting with the fact that this isn't a story about a handful of laggard shops. Nearly 500 professionals answered this survey, and close to half of them, by their own admission, can't do the thing 66% of their clients are asking for. When a gap that wide shows up in a sample that size, it's not a training problem at individual agencies. It's a tooling and methodology problem across the industry, and pretending otherwise just delays fixing it.
Why the old dashboard can't answer the new question
I've written before about AI search hitting its own "not provided" moment, and this survey is the client-facing consequence of exactly that shift. When Google encrypted search queries in 2011, the industry lost query-level attribution and spent years building workarounds, Search Console data, rank tracking, page-level inference. AI referral traffic is repeating that pattern, except the referrer data was often never built out for SEO-style attribution in the first place, and it's arriving faster than the 2011 transition ever moved.
Practically, that means three things break at once in a standard reporting stack. First, the click that arrives from an AI answer frequently carries little to no data about which prompt or citation produced it, so it lands in an undifferentiated direct-traffic bucket instead of a labeled channel. Second, the research journey itself has stretched, a buyer might ask ChatGPT three questions over two days before ever landing on your client's site, and none of that upstream behavior shows up anywhere in a standard analytics view. Third, and this is the one that actually generates the client-call panic, visibility inside the AI answer itself, whether your client's brand got mentioned, cited, or recommended, isn't something Google Analytics or a rank tracker was ever built to see. You can be doing everything right and still have zero rows in the report that prove it.
| WHAT CLIENTS NOW EXPECT | WHAT MOST REPORTING STACKS STILL SHOW |
|---|---|
| Are we mentioned in AI answers for our category? | No visibility signal at all, or a manual spot-check screenshot |
| Which AI engine is sending us traffic? | Collapsed into 'direct' or 'referral, unknown' |
| Did that AI-assisted research turn into a lead? | No connective tissue between the AI touch and the conversion event |
| Is our AI visibility trending up or down? | A one-time answer, not a tracked trend |
None of that is a reason to throw out the reporting stack you have. Rank tracking and GA4 still tell you true things. It's a reason to stop treating AI search visibility as a bolt-on metric you check occasionally and start treating it as its own tracked surface, the same way AI visibility tracking has had to account for real market share swings between engines rather than assuming one tool's numbers hold steady quarter to quarter.
What attribution-resilient reporting actually requires
I'm not going to pretend we've solved this cleanly, because we haven't, and anyone telling a client they have a fully-closed-loop AI attribution model in July 2026 is overselling. What we have found is a set of things that move the needle from "shrug" to "defensible answer," even without perfect data underneath them.
This is also where dedicated reporting and analytics work earns its keep rather than being a line item nobody questions. Building a dashboard that survives a missing-data era isn't decoration on top of the SEO and GEO work, it's the difference between a client trusting the next twelve months of reporting and a client quietly shopping your replacement because a competitor's agency had a slicker answer to one uncomfortable question. We've laid out the broader model for wiring rankings, citations, and revenue into one view in attributing pipeline to organic and AI-cited traffic, and a lot of the sequencing work in the 2026 AI discovery readiness playbook exists specifically because this reporting gap doesn't close on its own.
Where this leaves you Monday morning
I don't think the fix here is a new tool, though tools help. I think the fix is a change in what we're willing to say to clients before they ask. Most of us have been treating AI search visibility reporting as a feature we'll build once the platforms hand us better data, the same posture the industry took toward query encryption back in 2011, waiting for someone else to fill the gap. That wait cost years last time. It doesn't have to cost that long again, but only if the reporting gets built on triangulated signals now instead of on the hope that ChatGPT eventually ships an equivalent of Search Console.
So here's the reframe I'd offer, to clients and to ourselves: stop asking whether you can measure AI search visibility perfectly, because that answer is no for basically everyone right now, all 494 agencies in that survey included. Ask instead whether the client can see your reasoning, your mention tracking, your branded-search lift, your modeled attribution, laid out clearly enough that they trust the process even where the data itself is incomplete. That's a lower bar than perfect measurement. It's also, right now, the only honest one available.
See where you are cited today
A free snapshot audit of your rankings and AI citations before we ever talk.
Josh leads work at the intersection of SEO and generative engines at Something Inc., helping B2B brands get ranked and cited across every major AI engine.