Buy more testing.
You cannot buy a better grade.
A fixed amount each month plus an amount per conversation, billed in arrears. Paying changes when and how often we test — never how you are scored.
Vendor plans
Served live from our billing system. A fixed amount each month plus an amount per conversation, billed in arrears — the per-conversation rate steps down as volume rises, and each band prices only the conversations inside it.
Retest
- First 200 · $2 each
- 201 and above · $1.50 each
- File disputes on any result
- Request an off-schedule retest of your demo line
Continuous
- First 1,200 · $1.50 each
- 1,201 and above · $1.25 each
- File disputes on any result
- Request an off-schedule retest of your demo line
- Continuous monitoring between scheduled tests
- Download the fix brief for your published results
Each band prices only the conversations inside it — the rate steps down as you use more, and it is never applied backwards.
- Every grade cites the exact moments in the conversation it came from — with the recording wherever a call was recorded
- A conversation our platform broke never counts against anyone and is never billed
- When two vendors are too close to call, we say so instead of naming a winner
What counts as a billable conversation
One conversation with your agent — one call, one chat, one email thread, one SMS exchange — that ran to a real ending and produced a result we could score. A conversation your agent handled badly still bills: you are paying for the test, and a failed test is the one worth the most to you.
Conversations we place on our own initiative for the public scoreboard are ours, not yours. You are never billed for the testing behind your public grade.
If our own equipment broke the call, you don’t pay for it
When our own equipment is what failed — the phone system, the synthetic caller, the connection carrying the audio — the conversation is marked as our fault, excluded from the results, and billed to nobody. The same decision in the same place governs both the invoice and the grade, which is deliberate: a platform fault that quietly billed you would be a platform fault that quietly counted against you.
What paying buys
- Cadence. More assessments per year than our public calendar would reach you with.
- Timing. An off-schedule retest when you’ve shipped a fix, instead of waiting for us.
- Continuous monitoring between scheduled assessments, so a regression surfaces to you before it surfaces to a buyer.
- The fix brief. A private, evidence-linked remediation report: each issue, the recorded conversation that demonstrates it, and enough specificity to hand straight to your engineering team.
What paying can never buy
Stated as a list because a vague version of this promise is worth nothing:
- A grade change. Not a nudge, not a rounding, not a re-run that only publishes if it’s better.
- A takedown. A published assessment cannot be removed, hidden, de-indexed or embargoed for money.
- A scenario exclusion. You cannot buy your way out of the test you do worst on. Everyone in an industry runs the same set.
- A delay of publication. Results publish when they’ve cleared review, not when they’re convenient.
- A different line to be tested. A vendor account tests the demo line you publish — the same one your public grade came from. To test an internal build, open a customer account.
- A better reviewer. Judges are blind to vendor identity and the second-review panel is drawn from a different AI lab than the first — for paying and non-paying vendors alike.
Billing mechanics
- Monthly in arrears. You’re invoiced after the month for the conversations that actually ran in it.
- A card on file before the first billable test, even where the fixed monthly amount is zero.
- The price you approved is the price you’re billed. A run keeps the tariff it started under, even if we publish new pricing while it’s in flight.
- Cancelling stops future tests, not past invoices. Conversations already run still bill, on one final invoice.
- Caps are enforced before the calls, not after. Where a plan carries a monthly conversation limit, a run that would exceed it is refused up front rather than billed as an overage.
Common questions
Does a paid retest replace my old grade?
It supersedes it in the open: the new dated assessment becomes current and the previous one stays visible with its date. Grades are snapshots and are never quietly edited.
What if the retest comes out worse?
It publishes. You bought the calls, not the outcome — and we say so before you buy so nobody is surprised by it afterwards.
Is disputing really free even if we never pay you anything?
Yes. Disputes are free on every plan, including no plan. If challenging a score cost money, our incentive would be to publish scores worth challenging.
Can we pay to be tested in an industry you don’t cover yet?
You can ask — hello@proofground.ai — and cadence is the thing money buys — but the rubric for an industry is built before anyone is graded in it, and it isn’t built to order for the vendor paying.
Who sees the fix brief?
You do. It is private to your verified account, it is a snapshot of one assessment, and it is redacted before it leaves our systems.
The independent guide to AI receptionists. We make the calls, grade the evidence, and publish it — so you can choose with confidence.