Every AI-visibility budget is spent on faith. We settle it on proof.
Agencies want to sell results. Brands want to pay for results. Nobody can write that contract, because nobody can prove, without dispute, whether an AI mentioning a brand more this month is a real gain or noise. Nadelio is the neutral instrument that makes the deal signable.
First agreements signing now, one sector, France
The stalemate nobody talks about
The brand
Pays 1,500 to 50,000€/month for content and PR aimed at AI visibility, with no way to verify whether any of it worked.
vs
The agency
Wants to sell performance, not just hours, but has no instrument that survives a dispute over what "worked" means.
Both sides want the same thing: a number they can both trust. It has to come from someone with nothing to gain either way. That role, on this market, has been empty.
The thesis
Nadelio does not sell AI-visibility tools. It settles AI-visibility contracts, for a flat fee, never a percentage of the outcome.
Every measurement we publish, on every free audit on this site, is a bounded score: a number with its margin of error, and an honest verdict on whether it can be trusted. That same instrument, applied to a before-and-after, is what makes a performance contract possible for the first time. Read how the score is built.
How a settlement works
The brand
Fixes 25 buyer-intent questions at signature. Pays a bonus only if the gain clears the threshold.
The agency
Delivers the work, unchanged retainer. Finally has a number it can point to.
Nadelio
Measures both sides against a control group, rules on the outcome, for a flat fee.
A question set worth measuring is fixed before the campaign starts, so nobody can chase easy wins after the fact.
Competitor brands are measured in parallel as a control group, so a model update cannot be mistaken for a campaign win.
A gain is recognized only when the score moves outside the 95% confidence band, not on a single lucky sample.
At the settlement date, Nadelio publishes the verdict, timestamped, with the full measurement trail.
Example: a campaign moves a brand from 44±9 to 48±11. A bare-number tool calls that a win. Nadelio does not: the bands overlap, the gain is inside the noise, no bonus is owed. Move it to 61±5 instead, and the bands are clean apart: that one is real, and now provably so.
Every number below comes from an actual Nadelio audit run on the real Google of the brand's market. Nothing here is illustrative.
Deep Audit79€, one time
For a brand that suspects it is losing the category and wants the receipts, plus the plan to fix it.
Measured, Pennylane, comptabilité SaaS France
The read
Pennylane scores 60 ±1, behind Sage at 71 ±1 on its own buyer queries.
The blind spot
On meilleur logiciel comptabilité SaaS pour PME, the answers go to swapn.fr and legalplace.fr, not to Pennylane.
The bands do not overlap, so the gap to Sage is real, not noise. The Deep Audit hands Pennylane the exact queries it is absent from, the sites that own them today, and a written plan to earn those citations. That is what the 79 euros buys, a defensible starting line, not a vanity number.
Monitoringfrom 99€, per month
For a brand that already knows where it stands and needs to catch movement the moment it matters, not a quarter later.
Measured, Yoyaku, disquaire techno Paris
This week
Yoyaku holds 56 ±1, verdict stable, tied with Superfly Records on share of voice but behind on position.
Why weekly
AI answers drift with every model update. A one-off read goes stale. The interval is what tells a real move from noise.
Monitoring re-measures Yoyaku every week and stays silent until a change clears the confidence band. No alert on a lucky sample, no dashboard to check. The week Superfly Records pulls ahead for real, or Yoyaku closes the gap, the email lands. The card keeps charging, the cron keeps running.
Settlementflat fee, never a cut
For a brand and its agency who disagree on whether the work moved the needle. Nadelio rules on it, neutrally.
Measured, Qonto vs Shine, banque pro France
The claim to settle
Qonto at 80 ±1, Shine at 76 ±1 on shared buyer queries. Is Qonto's lead real, or inside the margin?
The ruling
The bands are 79 to 81 against 75 to 77. They do not touch, so the lead is real and provable.
Now flip it. Had the read been 78 ±4 against 76 ±5, the bands would overlap and Nadelio would rule no real gain, no bonus owed, protecting the brand from paying for noise. Same instrument, opposite verdict, decided by the interval and not by whoever shouts loudest. The referee is paid the same either way.
What it costs
A settlement needs a weekly measurement to rule on, that is the monitoring underneath it. The fee structure is simple on purpose: the referee is always paid flat, never a share of what you win.
Start monitoring a brand
Settlement
300 to 500€/mo Per brand under an active performance agreement. Flat, regardless of the verdict.
Citation Gap Report
990€ One-off: where your category gets cited, and where you are absent. Paid either way.
A settlement agreement is set up with your counsel and Nadelio directly, not a self-serve checkout. Start that conversation.
What Nadelio will never do
Take a cut of the outcome. A referee paid on the verdict is not a referee.
Execute the work. We measure and settle; agencies write the content, run the PR.
Guarantee a result. We price the measurement, never promise the number will move.
Compete on the free audit. Anyone can run a snapshot; almost nobody can settle a contract on it.
Where this stands
This is early, on purpose. The first settlement agreements are being signed now, in a single sector, in France, so the model is proven on real disputes before it is scaled. Every settlement, cleared or not, becomes part of a public record: an arbitrator's only asset is a track record nobody can fake after the fact.
Building or buying AI-visibility work right now and want the first agreement written for your case? That conversation is open.