How QuotedWrong checks AI pricing claims

We record the exact sampled answer, then compare it with dated first-party facts. Every observation keeps its own audit trail.

Method explanation updated 26 September 2026.

How does QuotedWrong check an AI pricing answer?

QuotedWrong records the exact prompt, sampled assistant response, platform or model when observable, country, language, time, and citations. It compares pricing claims with the vendor's first-party source as observed on a recorded date. Machine extraction may flag a potential discrepancy; a person must review the evidence before a finding is described as confirmed. Answers can vary by prompt, model, location, and date.

Run free scan

The observation contract

Every observation keeps three sets of facts.

[ What was asked ]

Exact prompt
Country and language
Observation date, time and timezone

[ What was answered ]

Platform and model, where observable
Assistant response or relevant excerpt
Assistant citations

[ What it was compared with ]

First-party source URL
First-party source observation date
Comparison status
Review status and reviewer notes
Correction-draft and approval status

Observations are not merged across dates. Answers change, so the time boundary is part of the result.

Answers come from the assistants' provider APIs. They are sampled answers, not rankings of consumer apps and not the answer every user sees.

Truth levels

Every finding carries one of five states. Each state pairs a color with a shape and a written label.

Potential discrepancy - machine extracted

Machine scan

Machine extraction suggests a difference. No person has confirmed it.

Needs human review

Review pending

Evidence is incomplete, conflicting, or outside the current scope.

Confirmed discrepancy - human-reviewed

Paid audit

A reviewer confirmed a difference within the recorded scope.

No discrepancy found in this sample

Clean sample

The sampled answer matched the available fact. Wider accuracy is not established.

Could not verify this comparison

Unavailable

The first-party source was missing or insufficient.

Level	Label	Meaning
Machine scan	Potential discrepancy - machine extracted	Machine extraction suggests a difference. No person has confirmed it.
Review pending	Needs human review	Evidence is incomplete, conflicting, or outside the current scope.
Paid audit	Confirmed discrepancy - human-reviewed	A reviewer confirmed a difference within the recorded scope.
Clean sample	No discrepancy found in this sample	The sampled answer matched the available fact. Wider accuracy is not established.
Unavailable	Could not verify this comparison	The first-party source was missing or insufficient.

What claims are compared with

Each claim is compared with a first-party source, usually your public pricing page, recorded with its URL and the time it was read. When the source is missing or unclear, the result says so instead of guessing.

How VerifiedPricing checks pricing facts

What QuotedWrong does not claim

Intent, negligence, legal liability, or platform-policy violations
Future model behavior, or that any AI answer will change after a correction
That an assistant is consistently wrong, from a machine scan alone
Accuracy beyond the stated sample, assistants, countries, languages, prompts and dates

QuotedWrong reports observed AI outputs and compares them with cited first-party sources.

See the method applied to your pricing

Free scan with up to two sampled model responses.

Run free scan