Skip to content
Start free

Truth,
engineered.

The verification layer for content you can't afford to get wrong. Every claim is cross-examined by a panel of frontier AI models and the live web, then adjudicated into a single verdict you can defend.

Start verifying freeSee a real report

1,000 free credits on signup, no card required

Q3 mobility briefing.docxTier 3 · Live researchChecking claim 1 of 12

Document

Global EV adoption keeps accelerating: EV sales passed 17 million worldwide in 2024, roughly one in five new cars sold. Supply chains are following the money, since coffee is the world's second-most traded commodity after oil, and logistics firms are electrifying those routes first. Meanwhile employers cite research showing remote workers are 13% more productive to justify smaller fleets.

Claim 1 · panel of 4

"EV sales passed 17 million worldwide in 2024"

  • Claude Sonnet 5True96%
  • GPT-5.4 MiniTrue94%
  • Gemini 3.8 FlashTrue92%
  • DeepSeek V4.1 FlashMostly true81%
VerdictTrue

Confirmed by the IEA Global EV Outlook 2025. The figure includes plug-in hybrids.

Unanimous, challenger could not refute5 sources cited

Orchestrating frontier models from

Anthropic
OpenAI
Google
DeepSeek
Meta
Mistral
The Great Wall of China is visible from space with the naked eyeFalse
Honey never spoilsTrue
We only use 10% of our brainsFalse
Bananas are botanically berriesTrue
Goldfish have a three-second memoryFalse
The Eiffel Tower grows ~15 cm taller in summerTrue
Mount Everest is the tallest mountain on EarthMisleading
Lightning never strikes the same place twiceFalse
The Great Wall of China is visible from space with the naked eyeFalse
Honey never spoilsTrue
We only use 10% of our brainsFalse
Bananas are botanically berriesTrue
Goldfish have a three-second memoryFalse
The Eiffel Tower grows ~15 cm taller in summerTrue
Mount Everest is the tallest mountain on EarthMisleading
Lightning never strikes the same place twiceFalse

Why a panel, not a model

One model guesses. A panel has to agree.

Any single model hallucinates. Different providers fail in different places, so we make them vote and then try to break the result.

6

frontier providers on the panel, so no single vendor's blind spot decides

7-pt

verdict scale. Misleading and Unverifiable are their own answers, never rounded to True or False

100%

of claims anchored to the exact sentence in your document

3

tiers of scrutiny, priced per claim and quoted before you submit

Worried one model will confidently make something up?

No single model ever decides. Every verdict is the consensus of a panel drawn from different providers. A challenger then tries to refute any unanimous call before it is confirmed.

Read the methodology
  • ClaudeFalse97%
  • GPTMostly false88%
  • GeminiFalse94%
  • ChallengerCould not refute

Need to show your work to an editor or a lawyer?

Sourced and auditable. Live-research verdicts cite the open web for every call. Every vote is on the record, and each report exports as a certificate you can hand over.

See a real report

  • Global EV Outlook 2025

    iea.org · supports · audited

  • Electric car sales, 2012 to 2024

    ourworldindata.org · supports · audited

  • EV Volumes annual review

    ev-volumes.com · supports

Moving faster than a research desk can?

Checks run in the background. You watch them land. Submit a draft, watch verdicts stream in, and get a signed webhook when the check completes.

Explore the API
// POST → your callbackUrl, HMAC-signed
{
  "type": "check.completed",
  "check": { "status": "completed", "claims": 12 }
}

How it works

From draft to defensible, in minutes.

Checks run in the background and stream results as they land. No single model ever decides.

01 / 04

Submit

Drop in a document or paste text and pick the depth of scrutiny. You see the claim count and the exact price before anything runs.

Q3 mobility briefing.docx12 claims
Tier 1 + Tier 3, default panel580 credits

Capabilities

Three tiers of certainty.

Match the depth of the check to the stakes. Every check is priced per claim and quoted before it runs, so the price you see is the price you pay.

Tier 1

Spell & Grammar

Spelling and grammar across the whole document in one fast pass, with every fix shown as a diff.

  • Every fix shown as a diff
  • Your style guide applied
  • One fast pass

5credits / check

See pricing
Tier 2

Model Fact-Check

A panel of frontier models checks every claim against what they know, then a judge adjudicates.

  • Panel from several providers, then a judge
  • A challenger attacks unanimous calls
  • No web round trip, so it is fast

25credits / claim

See pricing
Tier 3Most adopted

Live Research

Each claim is researched on the live web and returned with citations you can audit.

  • Live web research, cited
  • Every citation audited for faithfulness
  • Connectors to your own sources

40credits / claim

See pricing

Developers

Verification as an API call.

Estimate, submit, poll and fetch checks over a token-authenticated API. Signed webhooks tell your CMS the moment a verdict lands.

Read the docsAPI reference
cURL
# Check a draft against live research
curl -X POST https://fact.engineering/api/v1/checks \
  -H "Authorization: Bearer $FACT_ENGINEERING_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "text": "EV sales passed 17 million in 2024",
    "tiers": ["t1", "t3"],
    "callbackUrl": "https://example.com/webhooks/fact-engineering"
  }'

FAQ

Questions, answered.

Everything you need before your first check.

How is this different from just asking one AI model?

A panel of models from different providers checks every claim, and each model answers on its own. A separate AI judge then weighs their reasoning and evidence into one verdict. No single model's blind spot, bias, or hallucination decides the outcome. Read the full methodology

Which AI models do you use?

Models from Anthropic, OpenAI, Google, DeepSeek, and others, routed through one gateway. Organization admins can change the panel and the judge, and we add models as providers ship them.

Does it really check against the live internet?

Yes. Tier 3 researches each claim on the live web and returns a verdict with the sources it relied on, so you can open every citation yourself instead of trusting a model's memory.

How does pricing work?

It's credit-based and priced per claim. Before you submit, we count the claims in your content and quote the exact price: the number of claims times a per-claim rate set by the tier and the models on your panel, plus a flat fee per check. The quote is the charge. Every signup includes 1,000 free credits, and we refund failed checks automatically.

What can I upload?

PDF, Word (.docx), plain text, and Markdown. Submit one document or a batch. Processing runs in the background, and results appear in the report as each claim finishes.

Is my content private?

We use your documents only to run your checks and build the reports in your account. We don't sell your data or train models on it.

Stop betting your reputation on a single model.

Get a panel of frontier models to agree, with sources, before you publish.

Start free

Contact

Talk to a human.

Questions about tiers, the API or volume pricing? We read every message. You can also email support@fact.engineering.