If you are an operating partner and one of your portfolio companies has just put an AI visibility tool in next year’s budget, you are about to approve a line item in a category where more than 20 vendors sell measurement on more than 20 different methods, and where the trade body that reviewed them concluded most of what they produce is not fit to allocate budget against. Here are the seven questions I would ask before signing.

TL;DR

• The category has no shared method. The IAB counted 20-plus vendors, each measuring differently. Two tools pointed at the same brand in the same week routinely disagree.

• Only 16% of brands track AI visibility at all. So most of your portfolio is buying its first program, with no internal benchmark and no prior year to sanity-check against.

• The headline lift usually does not survive a second run. A portfolio company reporting a 10-point improvement often cannot reproduce it, because the first reading and the second are two draws from a wide distribution.

• Portfolio-wide is where the upside sits, and where the trap is. One method across twelve companies is a real comparative asset. Twelve different vendors across twelve companies is twelve incomparable dashboards and a roll-up nobody can defend.

• Seven questions, one hour. They apply at the portfolio company, not at the fund, and any competent CMO can answer them.

I run operations at a company that sells in this category, so read this with that in mind. I am not going to tell you which vendor to buy. I am going to tell you which questions separate a defensible line item from a plausible one, because I have sat on both sides of that conversation and I have been wrong about it before.

Why the headline lift usually does not survive

Start with the mechanic, because every one of the seven questions follows from it.

Think of flipping a fair coin 50 times. Sometimes you’ll get 23 heads. Sometimes 27. Sometimes 30. Same coin. Just different runs. The variation isn’t because the coin changed, it’s because 50 flips is a small sample, and small samples bounce around a lot. AI visibility measurement at 50 prompts works the same way.

Generative answer engines do not return a fixed answer to a fixed question. They sample. We have watched a brand appear in an answer at 9:00 AM and be gone from the same prompt five minutes later, with nothing changed on the brand’s side. So when a portfolio company reports that visibility went from 12% to 22% after a content sprint, there are two possibilities, and the dashboard usually cannot tell you which one you are looking at. Either the content worked, or you took two draws from a distribution that was always ten points wide.

The IAB’s guidance, published August 3, draws the line at the same place. Under 50 queries is exploratory and cannot characterize a category. Directional data answers trajectory questions and is explicitly not sufficient for budget allocation. Decision grade clears a higher bar on six things at once: sample size, query volume, prompt-type coverage, testing cadence, reproducibility, and documented method.

Our own estimate is that the industry burns roughly $2 billion a year acting on AI visibility numbers that do not hold up. That is IQRush’s figure, not the IAB’s, and I would rather you discount it as a vendor estimate than repeat it as a trade-body number.

The seven questions

Question 7 is the one I would not skip. If nobody can name the decision the number is supposed to inform, the line item is not a measurement program. It is a reporting habit, and it will cost the same every year.

Portfolio-wide is the actual opportunity

Here is the part I think most sponsors are underweighting. The single-company version of this is a marketing tool decision worth somewhere between $30,000 and $120,000 a year. The portfolio version is something else.

If twelve companies run the same measurement method, with the same definition of a citation, the same repeat-run design, and the same convergence gate, you get a comparative asset that no single company could build alone: a cross-portfolio read on which categories are being reshaped by answer engines fastest, which companies are structurally invisible in their own buying queries, and which content investments actually reproduce. That read is a diligence input for the next deal in the same vertical, and it is a line in the exit narrative.

If twelve companies run twelve vendors, you get twelve dashboards that cannot be added together, and an operating partner in a portfolio review explaining why Company A’s 30% and Company B’s 30% are not the same 30%. They will not be. Different prompt sets, different engines, different denominators, different definitions of what counts as a mention.

The fork is a governance decision, not a procurement one, and it has to be made before the first company signs.

Sponsor-Side Impact

Model it at the fund level. Twelve portfolio companies each spending $60,000 a year on AI visibility is $720,000 of annual portfolio opex, plus the content and agency spend it directs, which in our experience runs three to eight times the tool cost. Call the directed spend $2 to $5 million a year across the portfolio. If the measurement underneath is directional rather than decision grade, you are not risking the $720,000, you are risking the allocation of the $2 to $5 million, plus the credibility of every visibility number that reaches an LP update. And the failure is quiet. Nobody writes a memo when a content budget was pointed at the wrong query cluster. The number just does not reproduce next quarter, and the CMO explains it as market noise, which is the one thing it genuinely is.

Three things to put in the value creation plan

1. One method across the portfolio, decided before the first signature. Name the standard, not the vendor. The IAB’s six decision-grade criteria are a serviceable standard and they are public.

2. A reproduction test at 90 days. Re-run the baseline program with nothing changed. Write down what moved. That single test tells you more about the vendor than the entire sales cycle did.

3. Question 7, in writing, per company. What decision changes if this number moves five points? Put the answer in the plan. Revisit it at the next portfolio review.

Tracie Kambies is co-founder and COO of IQRush, which builds measurement infrastructure for probabilistic AI search. Here is the falsifiable bet: by mid-2027, at least one mid-market sponsor will run a standardized AI visibility spec across its portfolio and say so publicly. If nobody has by June 2027, the category is still selling dashboards rather than infrastructure, and I was early.

Book a walkthrough →

Published October 8, 2026.