OMISPHERE

Evidence-based Campaign Intelligence

OMISPHERE detects coordinated account groups — influence operations, engagement farms, artificial amplification networks — on YouTube and X (Twitter). Every finding is a probability with its confidence, evidence for, evidence against, and corroboration status; a maximal verdict requires either a discriminative signal (shared fingerprints, co-engagement, co-tag networks) or multiple independent methods agreeing. Built for OSINT researchers, investigative journalists, and trust-&-safety teams.

What OMISPHERE is not

OMISPHERE is not a binary “bot / not-bot” classifier. Every result is a probability with an explicit evidence chain. We never accuse, we never claim certainty, and we never tell you a person is a bot. We tell you that certain observable patterns are consistent with synthetic or coordinated behavior — and we show you exactly what those patterns are.

Self-improving

Every scan adds a behavioral fingerprint to OMISPHERE's database. Future scans pull priors from that growing set. The platform sharpens with every use — and the intelligence belongs to its users collectively.

The omi detection engine

Under the hood, the detection engine is called omi. It runs locally with no LLM calls in the per-scan path — pure Python heuristics, embeddings, and graph algorithms. LLMs are reserved for optional report generation, never the core scoring.

Scope, plainly

Today OMISPHERE scans YouTube only — videos and channels. Every “scan” covers a video's comment thread or a single channel's recent activity, and consumes one credit. X / Twitter ingestion is the next platform on the roadmap; pricing for X scans will reflect that platform's higher API cost when it ships.

Ethical use

OMISPHERE is for researchers, journalists, brand-safety teams, and platform-integrity professionals. It is not a tool to harass individuals. Probabilistic patterns are not proof.

Platform roadmap

Depth over breadth — deep coordination intelligence on the platforms that matter most.

  • YouTube — live: full comment analysis, channel intelligence, coordination detection
  • X / Twitter — live: account fingerprinting, hashtag/amplification networks, campaign detection
  • Reddit — planned: subreddit coordination, post + comment analysis
  • TikTok — planned: comment-section analysis, creator audience intelligence