Thirteen companies sell AI visibility software. The median one is invisible.
We sealed a measurement method, signed it, and then measured how often thirteen companies in the AI-visibility category appear in AI-generated answers about their own market. This page states what we found. The method is on the Method page. The evidence is on the Receipts page.
All measurements were taken on Perplexity, in a single window, under one sealed protocol.
What this is, and is not.
This is one population of thirteen companies, measured on one engine, under one sealed protocol, in one window of time.
It is not a measurement of every AI system. It is not a ranking. It is not a claim about your category, your company, or your assets. It is not a general statement about how often businesses appear in AI-generated answers.
It is a demonstration that the question can be measured honestly: with the method fixed before the evidence exists, absences counted rather than discarded, and every value audited against the basis underneath it.
Where a figure carries a limitation, the limitation is published beside the figure and is not removed for brevity.
Measured on Perplexity. Across the population, the median rate at which a company appeared in an answer that did not name it was zero. Eight of the thirteen were never named at all, in any run.
Measured on Perplexity. The same query, repeated, did not return the same answer set. Between roughly sixty and eighty percent of the set held steady from one run to the next. The remainder changed with no change to the input.
In four cases the engine returned a value with no supporting basis. We removed all four.
The composite score did not move. Not approximately — the affected axis values were 100 before the correction and 100 after it, exactly.
One 100 was fabricated: produced by a scoring rule applied below the floor it was written for, projecting an aggregate from a component that held no measured sub-metric at all. The other was earned, summed from real measurements. The number is identical. What produced it is not.
An audit that grades the value without auditing the basis underneath it records all four of those corrections as no change at all.
The best-performing company in the population reached an unaided appearance rate of 0.333 on Perplexity — one query in three. That is the ceiling of the measured population, not its average. Two out of every three times the question was asked, the strongest company in the category was still absent from the answer.
We ran this instrument on ourselves before we ran it on the market.
When we corrected our own instrument, our own composite score fell from 62.5 to 50.0. Every composite we had produced before that correction was inflated. We published the corrected figure and retired the earlier one.
The findings on this page have not been re-measured under the corrected instrument. The run that produced them was never sealed, and we are not going to call a re-score of a different population a reproduction of it.
What we did instead was seal a specification and run the corrected instrument from scratch against our own estate. That run is on the Receipts page. It measures one subject, not these thirteen, and it is a first measurement rather than a repeat of anything. It does not confirm the figures above and is not offered as though it did.
What happens when we run the instrument on ourselves?
We ran the full loop on our own site and we publish it, bad baseline and all.
Baseline. Our own first reading was 0 of 144 — and it sampled 0 pages of our own estate. A defective baseline. It is published intact, not re-run.
Prescription. The instrument prescribed four changes it could support and refused nine of thirteen it might have wanted — more than two-thirds refused.
Remedy. We executed the four: structured data, a knowledge-graph entity, self-contained answer blocks, and source attribution. It reached four of eight sampled pages; four evidence pages were barred at the time.
Re-measurement. Under the identical sealed method the value verdict is NO RESULT: 0 of 144 → 0 of 144. Four structural changes on one property do not create a category, and we say so. What moved is provenance — every component that was previously unmeasurable became measurable.
Ceiling, declared before the run: that re-measurement reached four of eight pages because four were barred. Disclosure: the bar has since been lifted and the remaining four pages remediated — any future re-measurement covers a larger share of the estate and is not directly comparable to this one.
Every stage is sealed to the append-only ledger and verifiable without contacting us: prescription, remedy, re-measurement, verdict.
Measured on Perplexity. Population: 13 companies. Protocol sealed prior to measurement; see Method. Underlying records: see Receipts.
Jet Fyul Dynamics LLC · Wyoming
What does the Findings page report?
The Findings page reports Jet Fyul Dynamics' sealed measurements of how brands appear in AI answer engines. Every figure traces to an append-only ledger entry sealed before the evidence it describes was captured.
Sources — entity catalogued at Wikidata; U.S. service mark at the USPTO (Reg. 8009699). Last modified 2026-08-17.