Conspiracy Theory Amplification

Conspiracy amplification uses LLMs to manufacture the dense web of "supporting" material — articles, fabricated study citations, insider testimony, leaked-document mockups, expert analyses — that gives a fringe claim the *texture* of a well-evidenced position. The technique works because conspiratorial belief is sustained less by a single proof than by volume and apparent convergence: many seemingly independent sources pointing the same way. LLMs collapse the cost of producing that volume to near zero, and they are adept at the rhetorical moves conspiracies rely on (just-asking-questions framing, selective citation, connecting unrelated facts into a narrative).

• Coordinated-amplification detection: Identify clusters of accounts pushing the same "independent" evidence in near-synchrony (shared timing, phrasing, link sets)
• Citation and source grounding: Resolve cited studies and "leaked" documents; fabricated or unresolvable sources are strong signals
• Stylometric clustering: Detect that supposedly distinct authors/experts share a single generative fingerprint
• Narrative-tracking against fact-check corpora: Match emerging claims to known debunked narratives and authoritative consensus

Conspiracy amplification consumes synthetic evidence (T8-AT-002) and authority impersonation (T8-AT-001) as inputs and feeds disinformation campaigns (T8-AT-007) and radicalization pipelines (T8-AT-009) as outputs. It chains with T9 synthetic media when a "leaked" video or photo is fabricated to anchor the narrative, and with T15 human-workflow exploitation when the manufactured consensus is used to pressure institutions or journalists.