Skip to main content

How it works

The methodology behind
every colored square.

ASURIQ doesn't ask another AI if the first one was right. It checks claims against the same databases a researcher, journalist, or compliance officer would use. In seconds.

The verification pipeline

What happens when your AI answers

Detect

The extension detects an AI response in ChatGPT, Claude, Gemini, or any supported provider.

Classify

A local classifier scores the response for check-worthiness: statistics, citations, entities, claims, and domain signals.

Extract

Verifiable claims and entities are extracted from the response text.

Verify

Each claim is checked against authoritative databases: PubMed, CrossRef, BLS, FDA, and more.

Score

A trust score is computed across seven axes: factual accuracy, source independence, recency, balance, domain sensitivity, coherence, and completeness.

Show

A colored square appears on the response. Click it for the full analysis: checks, confidence breakdown, and what to do next.

Where we verify

21 databases and growing

Every claim is checked against the databases most relevant to its domain. We add new sources regularly as we expand coverage.

New databases are added as we expand into additional domains. Medical claims check PubMed and OpenFDA. Financial claims check FRED and BLS. Citations check CrossRef and Semantic Scholar. Every query is routed to the sources that matter most for that specific claim.

How we score

Seven axes of trust

We don't just check if an answer is right. We measure how it could be wrong, and where to look if you want to verify it yourself.

Factual accuracy

Do the specific claims check out against authoritative sources?

Source independence

Do multiple independent sources confirm, or does everything trace back to one?

Recency

Is the information current, or based on data that's been updated since?

Balance

Does the answer acknowledge relevant counterpoints and limitations?

Domain sensitivity

Is the answer appropriately cautious for high-stakes topics like medical, legal, or financial?

Coherence

Is the answer internally consistent, or does it contradict itself?

Completeness

Does the answer cover the key aspects, or leave important gaps?

The combined score determines the badge color. But you can always click to see the breakdown and decide for yourself.

What you see

One small square. Everything you need.

A colored square appears on every AI response. The color tells you the headline. Click it for the full story.

Verified

Claims checked out. Sources confirmed. You can rely on this one.

Review

Partially verified. Some claims need a closer look. Tap for details.

Flagged

Significant concerns found. Check the analysis before acting on this.

Check-worthy (hollow)

The classifier detected verifiable content. Tap to run a full check.

Low signal (hollow)

No strong verifiable claims detected. You can still tap to run a check if you want one.

It works the same everywhere

Every domain. Every question.

"Ibuprofen is safe to take daily with no significant long-term risks."

ASURIQ checks dosage claims against FDA-approved labeling, verifies cited studies in PubMed, and flags when an answer contradicts official safety warnings.

Checked against: OpenFDA Drug Labels · PubMed · ClinicalTrials.gov · WHO

See it for yourself.

Install the extension, ask your AI anything, and watch the square appear. Free badge on every response. Detail card for 1 credit. No account needed for the badge.

Install Free. No Account Needed.