Try the truth filter
Paste any AI-generated text. See what the filter catches.
Free to try, right now. Need unlimited? Bring your own LLM key. We never store it.
12 free checks per day — no API key needed
Need unlimited tests? Bring your own API key
Unlock advanced models with your own key
Your key and data are never stored.
0.38
0.42
Trend Overclaim
1
S→T
Overgeneralized
1
L→G
Unsupported Leap
2
Δe→∫de
Dimension Scores
Claim-Level Audit
“Q4 revenue is projected to grow 34% based on current pipeline trends”
Pipeline snapshot treated as confirmed revenue
“Customer churn has dropped to an industry-leading 2.1%”
Single quarter compared to unspecified industry benchmark
“Our NPS of 72 confirms best-in-class customer satisfaction”
One metric elevated to definitive proof of satisfaction
“Enterprise segment ARR reached $18M this quarter”
Verifiable internal metric, but no comparison period given
“AI integration will drive 50% efficiency gains across operations by 2027”
Speculative projection with no pilot data or methodology
Commentary
This board deck presents optimistic projections as established fact. Pipeline figures are treated as confirmed revenue, single-quarter metrics are called industry-leading without benchmarks, and future efficiency claims lack supporting pilot data.
- Pipeline-based revenue projection presented as a confirmed growth trajectory
- Churn rate compared to unnamed industry benchmark — no source or methodology
- NPS score used as sole proof of customer satisfaction without context or trend data
- AI efficiency gains projected to 2027 without pilot results or implementation plan
Example output — paste your own text above to get a live analysis.
The question isn't whether the answer
is right or wrong.
Three AIs. Same report. None of them warned you.
Runs inside Claude and ChatGPT. 300 free credits.
Runs inside your existing AI workflow.
No integration project. No architecture change.
What happens when you run TPMN Checker
Four steps. Fully automated.
Input
Paste any AI-generated text — a report, forecast, analysis, PRD, research summary.
Score
Seven dimensions evaluated independently. Each scored 0.0–1.0.
Flag
Structural logic errors detected: S→T, L→G, Δe→∫de violations.
Rewrite
Auto-compose strips overclaims. Before and after, side by side.
Seven dimensions scored
| Dimension | What it catches |
|---|---|
| Source Attribution | Claims with no traceable evidence |
| Evidence Quality | Thin or outdated supporting data |
| Claim Grounding | Assertions presented as fact without basis |
| Temporal Validity | Stale data treated as current |
| Scope Accuracy | Local findings overgeneralized |
| Logical Consistency | Internal contradictions |
| Prompt Alignment | Does the output match what was asked? |
S→T
Snapshot treated as permanent trend
L→G
Local truth presented as universal
Δe→∫de
Sweeping claim from thin evidence
AI scores. You adjust. The standard improves.
Today, AI evaluates its own output. But who decides what “correct” looks like?
Not us alone. You.
When you use TPMN Checker, you see scores across seven dimensions. If you disagree with a score, that disagreement is data. Collected with your consent, aggregated across users, and analyzed for patterns — your evaluations become the ground truth.
Human at the edge, not in the loop.
Humans set the standard. AI enforces it at scale.
Start verifying AI output now
Free access
- 1
Connect in 30 seconds (Claude.ai, ChatGPT, or Cursor)
- 2
Ask your AI tool to run
gem2_truth_filter - 3
See your first reliability report
No credit card. No setup project. 300 free credits.
Pre-GA — lock in pricing before General Availability
| Tier | Price | Credits | RPM |
|---|---|---|---|
| Free | $0 | 300 | 5 |
| Starter | $9 | 2,000 | 15 |
| Builder | $19 | 9,000 | 30 |
| Founder | $29 | ∞ | 60 |
One-time payment, not subscription. Pricing and credits subject to change during Pre-GA. Get early access →
Your data is never stored by default. Never shared. Never used for training. How we protect your data →