A survey of 163 SEO and GEO practitioners, conducted over three weeks in July and published on Search Engine Journal, scores AI visibility data at 4.20 out of 5 for perceived value. Yet only 44% consider investing in a dedicated measurement platform worthwhile. Nearly a third rate platform investment 1 or 2 out of 5. The gap is not about price: just 7% raised cost as an objection. It is about trust.
What the Survey Measured
Duane Forrester, founder of UnboundAnswers.com, recruited respondents through his own network, re-shares, and paid promotion on LinkedIn and X. The self-selected sample carries roughly a seven-point margin of error. Forrester discloses he built one of the platforms being evaluated, a relevant conflict noted in his original piece. Respondents rated five data types on a 1-to-5 scale:
- Query alignment beyond just keywords, 90% rated 4 or 5
- Competitor comparison on the same query, 83%
- Whether a mention comes from training or retrieval (retrieval: the AI fetching live web content at the time of the query, as opposed to knowledge baked in during model training), 83%
- Chunk-level attribution (tracing which specific source passage a model drew from when generating its answer), 75%
- Citation status, 71%
Average across all five data types: 4.20/5. Average willingness to invest in a platform to track them: 3.19/5. That gap held stable from the first 36 responses to the final 163, suggesting it is not a sampling artifact.
What Is Holding Back the Investment?
Of the 123 respondents (75%) who left open-text answers, objections broke into clear themes, per Forrester’s analysis:
- Trust, accuracy, opaque methodology, 24%
- ROI and attribution to business value, 20%
- Non-determinism (AI systems that produce different outputs each time they are queried) and personalization, 15%
- Synthetic prompts versus real user demand, 11%
- Distinctions between citation, mention, and recommendation, 4%
Combined, 57% raised either “I don’t believe the number” or “I can’t connect this to money.” Just 7% said cost was the problem. Affordability is not what is stopping purchases.
Why the Methodology Problem Has No Clean Answer
Multiple respondents identified what Forrester describes as the “no denominator” problem: platform scores are built from invented prompt lists, not observed real-world query volumes. When model variance shifts a score between reporting periods, a client may be told they won or lost, with no underlying data to indicate which. One respondent called it a self-fulfilling prophecy. Another compared citation tracking to a brand awareness signal rather than a diagnostic tool, noting that because AI answers change each time the question is asked, there is nothing to reverse-engineer.
Forrester argues, and flags explicitly as his own view rather than a survey finding, that full methodology disclosure is commercially impossible for vendors: publishing it converts a paid product into a free tool. He draws a parallel to keyword research platforms, trusted for two decades without practitioners ever seeing inside them. The sharper question, he suggests, is not “show me your methodology” but “what evidence would actually make you believe a number?” No respondent proposed a specific answer.
What Changes After Buying a Subscription
Among current subscribers, the gap between data value and platform value nearly disappears, according to the survey results. Among everyone else, the gap is roughly three times larger. Forrester does not assign causation: buying may resolve doubt, or people who already lack doubt may simply be the ones who buy. The survey cannot distinguish the two.
A pattern emerged in the open-text responses: non-subscribers tend to doubt the numbers, while subscribers tend to accept the numbers but struggle to act on them. Whether that reflects a training gap, a product gap, or a structural limit of the medium, Forrester does not conclude.
Only 8% of respondents built their own measurement tooling. Among those raising trust or non-determinism concerns, the rate rises to 9%, a statistically negligible difference. The “I don’t trust it” objection, Forrester notes, is functioning more as a request for someone else to solve the problem than as a direction the market is actively pursuing.
Which AI Platforms Practitioners Are Watching
Respondents were capped at five platform selections. Results by share who included each, per the July survey:
- Google AI Overviews and AI Mode, 95%
- ChatGPT, 94%
- Gemini, 75%
- Claude, 64%
- Perplexity, 34%
- Microsoft Copilot, 25%
46% of respondents used all five selections, so these figures are floors rather than full-distribution rankings. Nothing outside these six cleared 5%.
What This Means for AI-Search Visibility
The clearest way to read these results: practitioners trust what they are trying to measure, and do not trust the ruler doing the measuring.
Picture a shop owner who wants to know how often ChatGPT recommends their business when someone asks for a local supplier. That question is genuinely worth 4.20/5 to know. But the tool giving an answer built its score by picking its own list of hypothetical questions, then running them through the model on your behalf. It might not have included the exact questions real customers are actually typing. And because AI systems give different answers to the same question depending on the user, the time of day, and dozens of other factors, the score you receive is a snapshot of a moving target that nobody can freeze. That is the 3.19/5 reality.
In Hingewise’s view, the most structurally important finding is the “no denominator” problem, not the methodology opacity. Traditional rank tracking works because search results have enough consistency to anchor a score. AI systems are non-deterministic by design: the same prompt can produce materially different answers across users and sessions. This means an AI visibility score may be measuring the platform’s own prompt selection as much as it measures a brand’s actual presence in model outputs.
That does not make AI visibility measurement useless. It means the appropriate mental model is different from rank tracking. A rank tracker is diagnostic: you rank position 4 for a keyword, and that is a specific, reproducible fact. An AI visibility signal is closer to a brand sentiment panel: directional, noisy, and most useful when tracked over time across competitors rather than read as an absolute number on any given day.
The finding that subscribers move past doubt but then struggle to act on the data points to a second, less-discussed gap: practitioners who accept the scores have not yet developed consistent frameworks for translating them into concrete content or technical decisions. That, rather than trust, may be the more tractable problem to solve first.
One further data point worth noting: Forrester’s survey captured 163 responses from an estimated population of 715,000 SEO and internet marketing consultants employed in the United States alone, according to IBISWorld figures cited in his analysis. That is roughly 1 in 4,400, despite seven distribution attempts, two newsletter sends, amplification by approximately 20 industry voices, and paid promotion on two platforms. The talk-to-action ratio in AI search visibility, based on this evidence, skews heavily toward talk. Hingewise’s view is that this gap matters as much as the trust gap: a category described by 87% of its practitioners as an active, real concern should produce more than a 0.02% survey participation rate.
Before Evaluating an AI Visibility Platform: Seven Things to Check
- Ask to see the platform’s prompt list. Platforms that don’t disclose what prompts they use are measuring your visibility for queries someone else chose.
- Request variance data. Ask how scores change across repeated runs of the same prompt before trusting a single data point as a win or a loss.
- Confirm it distinguishes training mentions from retrieval citations. These have different optimization implications.
- Verify it covers Google AI Overviews and ChatGPT at minimum. These two alone account for 94-95% of practitioner tracking priority, per the survey.
- Clarify how the platform defines citation, mention, and recommendation. They are different outcomes with different values, and conflating them distorts reporting.
- Check whether competitor data appears on the same queries, not just your own scores in isolation.
- Look for documented examples of how other subscribers translated scores into actual content or technical decisions.
The tension the survey surfaces, high perceived value of the underlying AI visibility data alongside persistent skepticism about the infrastructure measuring it, is likely to persist as long as AI systems remain non-deterministic and prompt lists remain proprietary. What moves next is not obvious. But the most useful reframe may be Forrester’s own: stop asking vendors to prove their methodology, and start defining what evidence would be sufficient. That question has no industry answer yet.
Source: Duane Forrester, “The GEO Trust Gap: SEOs Want The Data, But Not The Platforms Selling It,” Search Engine Journal. Survey conducted over three weeks in July, 163 respondents, ~7-point margin of error. IBISWorld US SEO/internet marketing employment figure cited by Forrester in the same piece.