Rating and ladder placement
Where the brand sits on R1–R7, what is holding it there, and the specific conditions that would move it up or down a rung.
Independent ratings · AI representation
Every AI answer about your brand is a claim. We rate whether it holds.
Buyers now meet your company inside an answer, not on your site. What that answer says is unverified, unstable and unattributed. Vizaible runs a fixed, published test across ChatGPT, Gemini and Claude, scores what comes back against your own public record, and issues a rating with every failed check named. The test never bends, whoever pays.
Rating action
Illustrative sample
- Subject
- Example NV — B2B software
- Rating
- R4 accurate, under-surfaced
- R1
- R2
- R3
- R4
- R5
- R6
- R7
Accuracy gates R1–R3Corroboration gates R4–R7
- Providers
- GPT · Gemini · Claude
- Checks
- every failed check listed
- Conditions
- model versions logged · temperature 0
- As of
- 15 Jul 2026
The reason the rating is worth anything
The measurement never bends.
Same test for everyone
Every subject is measured by the same fixed procedure, whether or not it buys anything else from us. A rating costs the same whatever it says, and no one is paid more when a subject rates well.
Fixes kept apart, and disclosed
When a company asks us to improve how AI represents it, that work is a separate, fixed-fee project, declared in the report and to anyone who asks. Its result is judged only by a fresh run of the same unchanged test.
Nothing suppressed
Every check a subject failed is printed in the report, named, with the model response that produced it. Subjects cannot buy an omission, and we hold our own instrument to the same rule.
What you receive
A rating, and the evidence that produced it.
A dashboard tells you a number moved. A rating tells you what was tested, what the models actually returned, which claims failed against your own published record, and what would have to change for the rating to move. Every figure traces back to a logged response you can read.
Full response log
Every prompt, provider, model version and raw answer retained and reproducible, with the date and conditions of the run.
Failed checks, in full
Every check that did not pass is named — including checks on our own instrument, where the run fell short of its stated tolerance. This is the part most reports leave out.
Share of category answers
How much of the answer space for the category's real buying questions the brand holds, measured against a competitor set agreed before the run.
The scale
- R1
- R2
- R3
- R4
- R5
- R6
- R7
Accuracy gates R1–R3Corroboration gates R4–R7
R1–R3 · Accuracy
Is what the models say about you true? Facts, ownership, footprint, capabilities, and whether the brand is confused with another entity or its parent. Failing here caps everything above it.
R4–R7 · Corroboration and recall
Can what you say about yourself be verified anywhere outside your own site, and do the models raise you unprompted when a buyer asks a category question? Each rung above R3 has a stated threshold.
Being right is not enough. A brand whose every claim checks out can still be held at R4, because nothing outside its own website corroborates it and the models never raise it unprompted. Accuracy opens the scale; external corroboration and recall are what move you up it.
How a rating is produced
Fixed procedure, in this order, every time.
The order matters and is not negotiable per subject. Scope is frozen before anything is run, so nobody chooses the questions after seeing the answers.
01
Scope and freeze
Brand, category, competitor set and the buying questions to be tested are agreed and locked before the first prompt is sent.
02
Run
The same prompt battery is put to each provider with model versions recorded and temperature at zero, so the run can be repeated and compared.
03
Score
Responses are checked against the brand's own published record — its site, filings and verifiable third-party sources — not against our opinion of it.
04
Review
Every flagged item is read by a person before issuance. Borderline calls are recorded as borderline rather than rounded away.
05
Issue
The report goes out with the rating, the evidence log and the complete appendix of failed checks. The subject sees it in the same form anyone else would.
Snapshot
A first read on whether the models know you, and whether what they say is right. Returned in minutes.
Free
Rating report
The full audit: rating, ladder placement, share of category answers, evidence log and the appendix of every failed check.
On request
Fix
A fixed-fee project to correct what the models get wrong or miss about you, scoped from your report. Proven only by a re-measure with the same test.
Fixed fee · scoped per report
Ratings and Front Desk are kept apart. Vizaible also sells Front Desk, an assistant and AI-visibility service for small business websites. Its checks are working checks for the customer's own use, never ratings. Buying Front Desk has no effect on any rating, the rating test is the same for Front Desk customers as for everyone else, and a rating report states when its subject is a Front Desk customer.
Request a rating or fix
Tell us the brand and the category. We'll come back with scope and price.
We also rate on behalf of acquirers and investors doing diligence on how a target is represented in the answer layer.