Deception Flags — Reviewing Coordinated Manipulation
Corroboration is the platform's strongest credibility signal — which makes it the obvious thing to fake. A coordinated campaign can push many "sources" repeating the same false claim to manufacture the appearance of consensus. The deception detector watches for that pattern and raises a flag for analyst review, so manufactured agreement never quietly inflates a finding.
This article is the review workflow. For how a flag affects scoring, see Credibility & Validity — in short, a confirmed flag acts as a hard ceiling that holds a cluster down regardless of how many sources back it.
What gets flagged
Two patterns are the usual triggers:
- Narrative coordination — multiple sources publishing near-identical text in a tight window, suggesting coordinated messaging rather than independent observation.
- Timing anomaly — several distinct sources publishing the same event almost simultaneously, faster than independent reporting would plausibly allow.
Both are signs that what looks like many witnesses may be one orchestrated push.
Opening the panel
Click ⚠ Deception Flags (the Threats panel) in the header. The button carries a badge with the number of unreviewed flags, and it updates in real time as new flags are raised — you don't need to refresh.
Reviewing a flag
Each flag card shows what triggered it: the type, when it was flagged, a measure of how strong the signal is (text similarity for coordination, source count for timing), and the number of intel items involved. If you need to look closer, open the items in the intel feed before deciding.
Then record one of three verdicts:
| Verdict | What it means | Effect |
|---|---|---|
| Confirm deception | You agree this is coordinated manipulation | Applies a credibility penalty to the sources involved |
| False positive | Innocent — independent reporting that merely looks similar | No scoring change; the flag is archived |
| Inconclusive | Can't tell either way | Marked reviewed, no scoring change |
Your verdict and name are recorded on the card. Confirming deception is how the platform learns which sources participate in manipulation — repeated involvement degrades a source's standing over time.
Staying on top of them
Filter by status (all / unreviewed / confirmed / false positive) to focus on what still needs attention, and by type to look at coordination or timing flags separately. New high-confidence flags are pushed to every connected analyst the moment they're raised.
Treat a heavily-corroborated item carrying a flag with suspicion, not confidence. The whole point of the flag is that the corroboration may be the manipulation.
Where to next
- Credibility & Validity — how a confirmed flag caps a cluster's credibility.
- The Analyst Review Queue & Triage — the other surface where you record decisions.