The GEO Trust Gap: SEOs Want The Data, But Not The Platforms Selling It

Aug 14, 2026 02:00 AM - 1 day ago 3

Over 3 weeks successful July, I ran a study asking group who activity connected AI hunt visibility what they deliberation of the platforms built to measurement it. 163 responses. This is simply a self-selected sample recruited done my ain web and its re-shares (as good arsenic paid ads connected LinkedIn and X), truthful it describes engaged practitioners successful and astir 1 area of the industry, not the full industry. Percentages present transportation astir a seven-point margin, and I’ll travel backmost to the sample size astatine the end, because it turned retired to beryllium portion of the story.

Here’s what came retired of it.

The Gap

I asked group to complaint really valuable various kinds of AI visibility information would be: Query alignment beyond conscionable keywords. Competitor comparison. Whether a mention comes from training aliases retrieval. Chunk-level attribution. Citation status.

Across those five, the mean was 4.20 retired of 5.

Then I asked whether investing fund successful a dedicated level for this feels worthwhile correct now.

3.19.

That’s the study successful 2 numbers. A 3rd of respondents rated the information highly valuable and level finance lukewarm aliases worse, successful the aforesaid sitting, minutes apart. Only 44% deliberation buying a instrumentality successful this class is worthwhile astatine all. Nearly a 3rd complaint it 1 aliases 2 (5 being highest).

I collected the information successful 4 snapshots arsenic responses came in: 36, past 75, past 100, past 163. Neither number moved much than a tenth of a constituent crossed the full run. Whatever this is, it isn’t a sampling artifact. It settled early and stayed put while the sample quadrupled.

What They Said

123 group (75%) wrote thing successful the unfastened matter box. I asked what their biggest unanswered mobility was, aliases their biggest rumor pinch the platforms they’d tried. I expected a characteristic petition list. That isn’t what I got.

Themes raised and percent of responses:
Trust, accuracy, opaque methodology – 24%
ROI and attribution to business value – 20%
Non-determinism, variance, personalization – 15%
Synthetic prompts vs existent personification demand – 11%
Not actionable: “what do I do now” – 9%
Price aliases cost – 7%
Citation vs mention vs recommendation – 4%
Attribution to a circumstantial passage – 4%
Prompt-count limits – 4%
No GSC balanced for LLMs – 3%

(Many group raised much than 1 issue, and each was counted nether each taxable it touched, truthful these adhd to much than 100%.)

The texture of responses matters much than the counts, I think.

Several group described the aforesaid structural problem pinch prompt-list tracking: you take the prompts, which intends you determine successful beforehand what you should beryllium visible for, past measurement yourself against your ain list. One called it a self-fulfilling prophecy.

Related, and sharper: these devices person nary denominator. Scores travel from invented punctual lists alternatively than observed query volume, truthful ordinary exemplary variance gets reported to a customer arsenic a triumph aliases a loss pinch thing underneath it to opportunity which.

One responsive based on citation devices are a fundamentally different animal from rank trackers: you can’t reverse-engineer what’s moving erstwhile the reply changes each clip you ask, truthful what you’re near pinch is person to a marque consciousness awesome than a diagnostic.

A paying subscriber astatine an agency said that erstwhile they pressed platforms to show their math, precisely 1 pulled backmost the curtain.

An agency moving 50-plus clients laid retired the compression plainly: they can’t waste AI visibility arsenic a work without measurement tools, and can’t warrant the devices until they’re trading the service.

One personification pinch 2 decades successful the manufacture framed it much evenly than anyone else: Third-party SEO information was ever directional alternatively than gospel, and that’s fine, arsenic agelong arsenic cipher pretends otherwise.

And the sharpest one, aimed consecutive astatine vendors: That you can’t do what these devices purport to do, because each personification of each exemplary gets a different experience. Snake oil. Magic beans.

I’m not going to reason pinch immoderate of that here. It’s what practitioners said erstwhile asked, and the worth of asking is diminished if I walk the abstraction explaining why they are wrong.

Price Is Not the Objection

7% raised cost. 57% raised either “I don’t judge the number” aliases “I can’t link this to money.”

That ratio is the astir useful point successful the survey. Whatever is holding this class back, the reply is not that the devices are excessively expensive.

What the Responses Show About Each Other

Reading individual answers gives you complaints. Cross-referencing them gives you thing else, and 3 patterns held up erstwhile I tested them.

People who raised spot concerns worth the information precisely arsenic overmuch arsenic everyone else, and would walk precisely arsenic much. Their standing of the underlying information value: 4.20, against 4.19 for everyone other (a quality of 1 hundredth of a point). Their budget: statistically indistinguishable. But their willingness to put successful a level drops to 2.76 against 3.36 for everyone else, and they’re markedly little apt to beryllium paying for anything.

Same valuation. Same money available. Different conclusion. Whatever is blocking this segment, it isn’t what the information is worthy to them, and it isn’t what they tin afford. It’s whether they judge it.

Almost cipher is building the alternative. 8% of respondents built their ain tooling. Among the group who raised spot aliases non-determinism, 9%. Among the 78% who telephone accuracy basal successful a vendor, 6%.

I don’t publication this arsenic hypocrisy. Building this is genuinely difficult (I know!), and astir practitioners person a occupation that isn’t engineering, but it does reframe the objection. “The numbers can’t beryllium trusted” isn’t functioning arsenic a test anyone acts on. It’s a petition for personification other to lick it properly.

The spread betwixt valuing the information and backing a level is identical crossed each role. Agencies, in-house teams, independent consultants are each wrong a rounding correction of each other. It isn’t agencies being inexpensive aliases in-house teams being spoiled. It’s the full market.

What does move it is whether you’ve bought. Among existent subscribers, the spread astir disappears. Among everyone else, it’s 3 times larger.

I can’t show you which guidance that runs. Buying whitethorn resoluteness the doubt, aliases group without the uncertainty whitethorn beryllium the ones who buy. The study can’t separate those, and I’m not going to dress otherwise. But it’s the azygous largest divided successful the dataset, and it suggests the objection looks different from wrong a subscription than extracurricular one.

There’s a hint of that successful the themes, too, though the numbers are mini capable that I’d telephone it suggestive alternatively than established: subscribers were much apt to opportunity the information doesn’t show them what to do, and little apt to opportunity they uncertainty the data. People who haven’t bought uncertainty the numbers. People who person bought judge the numbers and can’t enactment connected them. Whether that’s structural connected their extremity aliases a deficiency of acquisition aliases knowledge, I cannot say.

What They Actually Want

Ranked by stock standing each 4 aliases 5: query alignment beyond conscionable keywords, 90%. Competitor comparison connected the aforesaid query, 83%. Whether a mention comes from training aliases retrieval, 83%. Chunk-level attribution, 75%. Citation status, 71%.

On which systems matter: Google’s AI Overviews and AI Mode astatine 95%, ChatGPT astatine 94%, Gemini 75%, Claude 64%, Perplexity 34%, Copilot 25%. Nothing other cleared 5%. I capped that mobility astatine 5 selections, and 46% of respondents utilized each five, truthful dainty those arsenic floors.

87% picture their believe arsenic actively moving connected this aliases established successful their work. This is not an assemblage that needs convincing the problem is real.

The Ask Nobody Can Fill

The most-raised objection was methodology opacity: Show maine wherever this information comes from and why I should judge it.

It’s a reasonable point to want. It’s also, arsenic stated, not a point immoderate vendor successful this class tin springiness you.

Now, I should punctual everyone that I built 1 of these platforms. That’s a conflict, and you should publication what follows knowing it. It’s besides why I person a position connected what vendors tin and can’t disclose, arsenic I’ve had to make that telephone myself.

For a venture-backed company, the methodology is the asset. Publishing it converts the merchandise into a free instrumentality pinch a pain complaint and a board. Any vendor who appears to person opened the container has shown you a curated subset, which intends the disclosure you asked for is either commercially fatal aliases theatre, and location is nary 3rd option. This is not unsocial to GEO tools. It’s existent of each measurement business that has ever existed, including the keyword devices this manufacture has trusted for 20 years without ever seeing wrong them. I worked wrong 1 of those systems for almost a decade, truthful I’m not guessing erstwhile I opportunity I cognize the difference.

I besides want to beryllium clear that this is my argument, not a finding. The study didn’t measurement it. What the study shows is that group raising this objection person the aforesaid fund arsenic everyone else, which intends they aren’t looking for a logic not to buy. It’s a existent objection. It conscionable has nary disposable reply successful the shape it’s being asked.

Which leaves a harder mobility underneath: if you can’t person the methodology, what would really make you judge a number? Reproducibility? Published variance? Third-party audit? Nobody successful the responses projected one. That spread seems worthy much attraction than it’s getting.

The Part I Keep Thinking About

Several respondents made a lawsuit I can’t dismiss: that this whitethorn not beryllium measurable successful principle. The systems are non-deterministic. Every user’s acquisition is personalized. A snapshot of what a exemplary said connected Tuesday to a synthetic punctual whitethorn beryllium measuring thing that generalizes to anything.

I don’t deliberation that’s right. But I can’t beryllium it isn’t, and neither tin anyone trading you a dashboard.

Which brings maine to the number I’ve been avoiding. IBISWorld counts astir 715,000 group employed successful SEO and net trading consulting successful the United States alone. 163 of them answered this survey. One successful astir 4,400.

I shared it 7 times. It went retired successful my newsletter twice, appeared successful 2 respected manufacture newsletters, was amplified by astir 20 group successful this space, and I paid for a week of promotion connected 2 platforms. Every 1 of those asks was polite. The consequence trickled.

The azygous astir effective point I did was extremity asking politely and constituent retired really fewer group had bothered, which produced much responses successful 2 days than the erstwhile week had successful total.

I don’t person a cleanable publication connected what that means. It mightiness beryllium provender measurement outrunning anyone’s capacity to way it. It mightiness beryllium that requests to participate are pattern-matched and filtered earlier they register. It mightiness beryllium that the sermon astir AI hunt is louder than the believe of it. It could simply beryllium much personal, that my ain scope is diminishing. Honestly, it could beryllium each aliases nary of those things.

But it’s difficult to quadrate an manufacture that describes this arsenic an existential threat pinch a sample this difficult to assemble. That gap, betwixt really overmuch this gets talked astir and really overmuch of it is really being done, whitethorn beryllium the astir honorable uncovering here.

The information is worthy 4.20/5. The platforms are worthy 3.19/5. And 163 group retired of 715,000 gave their 3 minutes to opportunity so.

More Resources: