Election Manipulation Content

Election manipulation uses LLMs to mass-produce content that distorts the democratic information environment: false polling data, voter-suppression messaging, fake polling-location and procedure information, candidate smears, fabricated endorsements, misleading voting instructions, fake debate transcripts, fraud "evidence," premature victory declarations, and intimidation content. The technique works because electoral decisions are time-bound and information-dependent — a false claim that lands in the final hours can change turnout before it can be corrected. LLMs collapse the cost of producing locally tailored, demographically targeted, multilingual versions of suppressive or deceptive messaging, and can flood channels faster than verification keeps up.

• Procedural-claim grounding: Cross-check voting time/place/method claims against authoritative election-authority data; flag mismatches immediately
• Known-false-claim matching with semantic similarity: Match against curated election-misinformation databases, resistant to paraphrase
• Coordinated-inauthentic-behavior detection: Identify networks pushing suppression or false-result narratives in synchrony
• Endorsement/poll verification: Verify endorsements and polls against the named figure/pollster's authenticated channels

Election operations combine nearly the whole chapter: authority impersonation (T8-AT-001) for spoofed election officials, synthetic evidence (T8-AT-002) and false-flag (T8-AT-010) for fabricated fraud "proof," deepfake narratives (T8-AT-004) for doctored candidate clips, and disinformation infrastructure (T8-AT-007) for coordinated distribution. Procedural suppression content overlaps false-crisis (T8-AT-014) tactics ("polls closed due to emergency").