An article that doesn't sit right.
A quote with no source.
A data point that seems made up.

SlopSleuth checks every article you read — verifying whether quoted experts exist, whether cited studies are real, and whether claims match their sources. It shows you what it found, what it couldn't check, and why. The conclusion stays yours.

AI isn't the problem. Deception is.
Request beta access — invite only
Private beta — installs are approved personally, and capped by design
globalhealthtribune.com/research/nxr7-cognitive-decline
Global Health Tribune · Research
Breakthrough Study Links New Compound to 60% Reduction in Cognitive Decline
By J. A. Hartwell · March 28, 2026 · 5 min read
With more than 55 million people living with dementia worldwide, researchers at the European Centre for Neuroscience have published findings that a newly synthesised compound may significantly slow cognitive decline. Lead author Dr. R. L. Thornbury, described as a pioneer in neural inflammation research, stated that the compound was tested across 14 clinical sites in six countries. The study, reportedly published in Nature Medicine, found a 60% reduction in measurable cognitive decline.
3
SlopSleuth
globalhealthtribune.com
Check firstChecked —
nothing found
Verify before sharing
Claims checked · 4
3 worth checking before you share it — 1 confirmed against its source.
Unverified expert
No academic profile, publications, or institutional affiliation found for Dr. R. L. Thornbury.
"Lead author Dr. R. L. Thornbury"
No record of citation
No matching article found in Nature Medicine archives. This citation may not exist.
"published in Nature Medicine"
Couldn't verify
Clinical trial claim with no trial registration number or named institution to check against.
"14 clinical sites in six countries"
Matches source
WHO reports about 55 million people living with dementia — the article's figure matches.
"more than 55 million people living with dementia worldwide"

What SlopSleuth actually checks

  • Fabricated quotes and unverified experts Direct quotes and named authorities we can't confirm are real.
  • Fake citations Studies and journals we can't find in academic databases.
  • Unsourced statistics Numbers that look fabricated or can't be traced to a real source.
  • Unsupported health and financial claims Consequential claims made without evidence.
  • And every claim gets a named answer Confirmed, contradicted, no record, no source covers it, or not checked this scan — we tell you which, every time.

Every check runs against a growing registry of authoritative sources — 15 live today, with industry‑specific source packs on the roadmap.

Talk to us — hello@slopsleuth.com

The documents look finished.
The claims inside them were never checked.
Why it matters

This is already happening

And it isn't a few headline cases. An audit of 2.5 million biomedical papers found fabricated references have climbed roughly twelvefold in two years — reaching one paper in every 277 by early 2026.

Receipts, naturally — The Guardian · Fortune · ACS Information Age · Semafor · STAT / The Lancet

For teams

Check the report before it goes out.

Scan a report in your browser today. Open the Documents tab in the SlopSleuth panel and drop in a PDF or Word document — currently in beta — and every checkable claim gets the same treatment as a web page: checked against external sources, each landing in a named, auditable state. A report you commissioned, a supplier's white paper, research you're about to act on — checked before you rely on it, whether AI wrote it or not.

The check has to be independent of whatever produced the document — the system that writes a claim is the wrong system to certify it, the same reason the auditor is kept separate from the accounts. As more documents are drafted by machine, an independent, claim-level evidence trail is what lets a team show its work held up before anyone relied on it.

Pilot access — talk to us
We're talking to our first partners now.

A pre-send verification API is in development. Submit a document's text and get back the same per-claim evidence cards as a structured record: what was checked, against which sources, and which of six honest states each claim landed in. Not a score you have to trust — receipts you can audit. We're onboarding pilot partners first.

// one entry from checked_claims[] — example of a returned result
{
  "claim_text": "At least 117.8 million people, or one in 70 worldwide,
                 remain forcibly displaced",
  "kind": "statistic",
  "state": "confident",
  "evidence": "web-search: UNHCR Global Trends confirms 117.8 million
               forcibly displaced…",
  "sources": ["world-bank", "owid", "claim-review", "web-search"]
}
// states: confident · contradicting · entity_verified · no_match
//         no_source_for_type · not_checked_budget
Fair questions

Asked and answered

Is this a truth detector?
No — and be wary of anything that says it is. We check whether the evidence behind a claim holds up: does the cited study exist, does the named expert have a record, does the figure match the official series. Whether the article's conclusion is right stays your call.
Does it rate political bias?
No. Political leaning isn't what we measure. A fabricated source is a fabricated source whoever publishes it — the checks run identically on every outlet.
What happens when no source covers a claim?
We say so, plainly. "No record" means the sources we check have no matching entry — the claim may still be accurate. We'd rather hand you an honest gap than a confident guess.
Can it be wrong?
Yes — which is why every finding shows its working, and why each update must pass the same exam before it ships: claims we already know the answers to, fakes it has to catch and truths it has to leave alone. One miss and it doesn't ship. You can also challenge any flag from the panel.
Does it check images or video?
No. That's a different problem with different tools. We stay where we're strong: written claims, checked against the record.