VERIFICATION-AS-A-SERVICE

The Distiller

Catch AI hallucinations before your users do. One API call. Sub-10-second verification. 90%+ F1 score.

88.7%

Catch Rate

92.2%

Precision

90.4%

F1 Score

7.4s

Avg Latency

Starter

$49/mo

$0.049 per verification

1,000 verifications/month
REST API access
Claim-level verdicts (TRUE / FALSE / UNVERIFIABLE)
Source citations included
7.4s average response time
Community support

Start Free Trial

Professional

$199/mo

$0.020 per verification

10,000 verifications/month
Everything in Starter
Webhook & streaming support
Numerical hard-match module
Custom source weighting (Goggles)
Priority support + SLA

Get Started

Enterprise

Custom

Volume pricing from $0.008/call

50,000+ verifications/month
Everything in Professional
Dedicated infrastructure
Custom model routing
Primary source priority (.gov, .edu)
SOC 2 compliance + BAA available
Dedicated account manager

Contact Sales

Real Unit Economics

We believe in transparent pricing. Here's what each verification actually costs us.

Cost Per Verification

$0.014

GPT-4o audit ($0.008) + Brave search ($0.006) + GPT-4o-mini triage ($0.0001)

Margin at Professional Tier

30%

$0.020 price − $0.014 cost = $0.006 margin per call

Break-Even

714 calls/mo

At Starter pricing ($0.049/call), break-even on infrastructure at 714 monthly verifications

Enterprise at Scale

43% margin

At $0.008/call with Gemini stack ($0.0096 → optimized to $0.0046), 50K calls = $2,300/mo profit

Built For Teams That Can't Afford Hallucinations

⚖️

Legal Tech

Verify AI-generated case summaries, contract clauses, and regulatory citations before they reach a courtroom.

🏥

Healthcare AI

Catch hallucinated drug interactions, dosage claims, and clinical guidelines in patient-facing AI tools.

📰

Journalism & Media

Fact-check AI-assisted articles against live web sources. Flag unverifiable claims before publication.

🏛️

Government & Policy

Verify AI-generated briefings, data citations, and policy summaries against authoritative .gov sources.

🎓

Education

Audit AI tutor responses and generated study materials for factual accuracy before student delivery.

💰

Financial Services

Verify AI-generated market analysis, earnings data, and regulatory filings against real-time sources.

Benchmarked, Not Hypothetical

100 questions from OpenAI's SimpleQA dataset. Real API calls. Real results.

📊 Confusion Matrix

	Pipeline: FLAGGED	Pipeline: PASSED
AI Wrong	🎯 47 caught	💀 6 missed
AI Correct	🚨 4 false alarms	✅ 43 clean

V1 Results (with Negative Evidence + Numerical Hard-Match fixes)

🔬 Example Catch

GPT-4o said:

"The Infinity disco burned down in 1982, founded by Steve Rubell at 254 W 54th St."

Distiller found:

Fire was 1979. Founded by Maurice Brahms. Located at 683 Broadway.

Trust Score: 33% — 3 claims false, 3 true, 3 unverifiable

How It Works

One API call. Four stages. Verifiable citations for every claim.

1. Triage

Extract Atomic Claims

AI decomposes any text into independently verifiable factual statements

2. Evidence

Search Grounding

Each claim is verified against live web evidence via Brave's LLM Context API (parallelized)

3. Audit

Claim-by-Claim Verdict

AI auditor compares each claim against evidence: TRUE, FALSE, or UNVERIFIABLE with citations

4. Report

Trust Score + Corrections

Aggregate trust score, per-claim verdicts, corrections for false claims, source citations