Stop Citation Laundering

Verify whether claims are actually supported by evidence.

CiteGuardian checks cited sources, extracts the real claims, and shows you what holds up, what falls apart, and what was never supported in the first place.

Works on AI answers, articles, reports, and any highlighted text in your browser.

5 free verifications — no credit card required

Built for people who can't afford fake confidence

Verify claims in seconds
Catch decorative citations automatically
Highlight text anywhere in your browser
Use via extension, API, or batch upload

How It Works

1. Paste Text or Highlight in Your Browser

Check AI-generated answers, articles, reports, research summaries, or any claim-heavy text.

2. We Extract Claims and Inspect the Evidence

CiteGuardian breaks the text into individual claims, reviews the cited sources, and checks whether the evidence really supports what's being said.

3. Get a Verdict You Can Act On

See which claims are supported, which are unsupported, which are contradicted by the evidence, and which citations are just there for show.

Why Trust the Results

Five independent layers cross-check every claim — no single model making a guess

Claim Extraction

Breaks text into individual factual claims so each one gets checked on its own terms.

Evidence Retrieval

Finds the most relevant passages in your sources using semantic matching, not just keywords.

AI Judge

Evaluates whether the evidence actually supports, contradicts, or fails to back each claim.

Policy Engine

Deterministic rules catch precision errors, misattributions, and contradictions the AI might miss.

Scrub Test

Removes the citation and re-runs the check. If the verdict doesn't change, the citation is decorative.

What CiteGuardian Catches

Built for the three ways evidence goes wrong

Decorative Citations

Sources that look credible but do not actually support the claim. The citation adds authority, not proof.

Contradictions

When the cited evidence says the opposite of what the text claims it shows.

Unsupported Claims

Statements that sound factual and well-sourced but are not backed by the cited evidence.

What a Verification Looks Like

Real examples of what CiteGuardian catches

Contradiction

"NHS guidance states tea and coffee do not count toward daily fluid intake."

Evidence says: NHS guidance explicitly states tea and coffee do count toward fluid intake.

Confidence:
High
Unsupported

"A WHO review in 2024 found hybrid work reduces burnout by 37%."

Evidence says: No direct evidence found for the specific percentage or attribution to WHO.

Confidence:
High

Bad citations create false confidence

That's a real problem in AI-generated content, but it doesn't stop there. Analysts, researchers, writers, and teams using internal reports all face the same risk: claims that look sourced, but don't stand up.

CiteGuardian helps you catch weak evidence before it becomes:

A bad decision
A compliance problem
A reputational hit
A published error
A false sense of trust

Where CiteGuardian Fits

Not just for AI — anywhere claims meet evidence

AI and RAG Teams

Add verification to generated answers before they reach users.

Analysts & Researchers

Check whether reports, summaries, and cited statements are actually supported.

Legal, Compliance & HR

Spot overclaims, weak sourcing, and evidence mismatches before they create risk.

Editors & Content Teams

Review AI-assisted or human-written content before publishing.

Four Ways to Verify

Deep analysis
Web App

Paste text and sources, get a full report with per-claim verdicts, evidence excerpts, confidence scores, and alignment breakdowns.

Best for: one-off checks where you want the full picture.

Instant check
Chrome Extension

Highlight any text on a web page, right-click, and verify. The page itself becomes the source — no copy-paste needed.

Best for: spot-checking articles, AI answers, and reports as you read.

Automated
API Access

Add a verification step to your RAG pipeline, content workflow, or review process. Includes a pass/fail gate for blocking hallucinated responses.

Best for: programmatic verification in production systems.

At scale
Batch Processing

Upload a JSON file of answers and sources, or use the batch API endpoint. Results delivered via polling or webhook callback.

Best for: auditing document sets, model outputs, or content backlogs.

Plans & Pricing

Monthly subscriptions with included credits. Start free, upgrade when you need more.

Free
$0

5 credits on signup

  • Basic verification
  • Web UI access
Starter
$19/mo

75 credits/month

  • Scrub test
  • API + extension
  • PDF/MD export
Most Popular
Pro
$49/mo

250 credits/month

  • Decision engine
  • Batch + webhooks
  • Gate API

Verifications cost 1–4 credits based on claim count. See full pricing details

Frequently Asked Questions

No. CiteGuardian can verify claims in AI answers, articles, reports, research summaries, and any text you highlight in the browser. If it makes a claim and cites a source, CiteGuardian can check it.

It checks whether the cited evidence actually supports the claim, contradicts it, or fails to support it altogether. It also detects decorative citations — sources that look relevant but don't move the needle on the claim.

No. It helps reviewers work faster and catch weak evidence earlier, but high-stakes decisions should still be reviewed by a human. Think of it as a first-pass filter, not a final verdict.

Yes. CiteGuardian supports API access and batch verification for automated and large-scale use cases. Add a pass/fail gate to your RAG pipeline, or batch-verify entire document sets. See the API docs for details.

CiteGuardian shows its evidence alongside every verdict — the source text, confidence scores, and reasoning. You can verify any call yourself in seconds. It's designed as a first-pass filter that catches weak evidence at scale, not a replacement for human judgment. When it's uncertain, it says so.

Don't trust a citation just because it looks convincing

Verify the evidence before you trust the claim.