Decision integrity platform

Repetition is not
independent evidence.

Semeion helps teams see whether many reports are truly independent or simply repeating the same source—before those reports influence a decision.

Explore the Control Center Request a pilot
Doesn't judge viewpoints Compares sources Shows the decision trail
Control Center preview

A working, read-only look at Semeion using representative data. We are sharing it early to learn what is useful before a wider release.

Sample signals62representative data
Active narratives115 recent
Gate evaluations622 need review
Audit entries24835.1% quality
04 / Model previewEvidence terrain
5 narratives mapped
Evidence 01Pricing friction after policy change

Renewal objections rose after the latest packaging policy changed.

Trust
81.0%
Risk
18.0%
Corroboration
58.0%
Updated
Aug 7, 02:32 PM
Open full evidence
Evidence 02Support wait-time escalation

Independent support channels report longer resolution windows.

Trust
76.0%
Risk
12.0%
Corroboration
74.0%
Updated
Aug 7, 02:25 PM
Open full evidence
Evidence 03Migration confidence improving

Successful migration reports persist across unrelated customer cohorts.

Trust
85.0%
Risk
8.0%
Corroboration
81.0%
Updated
Aug 7, 02:18 PM
Open full evidence
Evidence 04Partner claims moving in lockstep

Several partner accounts repeated the same claim within a narrow window.

Trust
67.0%
Risk
16.0%
Corroboration
44.0%
Updated
Aug 7, 02:11 PM
Open full evidence
Evidence 05Research validation expanding

External research independently corroborates the observed performance trend.

Trust
88.0%
Risk
6.0%
Corroboration
79.0%
Updated
Aug 7, 02:04 PM
Open full evidence
Evidence gateAllow92.9%
Evidence gateAllow93.4%
Evidence gateAllow93.3%
Evidence gateHold60.0%
Evidence gateHold60.0%
Evidence gateAllow90.0%
Representative evidence terrain
Latest decisionPricing friction after policy change
0.18
Historical snapshotMigration confidence improving
0.08
Historical snapshotResearch validation expanding
0.06
Historical snapshotPartner claims moving in lockstep
0.16
Historical snapshotSupport wait-time escalation
0.12
Historical snapshotPricing friction after policy change
0.18
Time-lapse · 3s
Signal pipeline
Complete stage: Ingest62
Complete stage: Normalize35.1%
Complete stage: Score52 src
Current stage: Cluster11 sample
Review required: Gate2 review
Complete stage: Audit248

More than 2.6 million records processed in an offline evaluation

Offline evaluation

In an offline evaluation, Semeion processed 2,625,587 records from a public Harvard Dataverse research archive. All eight source files matched checksums published by the repository.

Why it matters

Ten reports may still mean one source.

Copies, syndication, aggregators, bots, AI agents, and coordinated people can all make one claim look widely confirmed. Semeion shows what appears independent, what looks repeated, and what is still uncertain.

See what is connected.

Compare source, timing, and content to find duplicates, echoes, and related reports.

Keep uncertainty visible.

When independent support is missing, Semeion can hold the case for review instead of forcing a yes or no.

Know why.

Inspect the evidence, policy, and any human review behind each allow, hold, or reject result.

More than swarms

The pattern matters, not the actor.

Swarms made the risk obvious, but the same problem can come from bots, coordinated people, news syndication, or an ordinary system receiving the same item more than once.

Semeion analyzes observable patterns. It does not claim to identify bots or infer intent.

How it works

From incoming reports to a reviewable result.

Semeion does not decide what is true. It gives people—and the AI systems they use—a clearer evidence trail before they act.

Read how it works
01CollectBring reports from feeds, APIs, and webhooks together.
02GroupCatch repeated deliveries and place related reports together.
03CheckCompare where reports came from, when they appeared, and signs of coordination.
04GateApply a clear policy: allow, hold, or reject.
05RecordKeep the evidence and any human review behind the result.
Clear boundaries

Better evidence, not an automated verdict.

Semeion helps people question the evidence before they rely on it. The final decision remains human-owned.

Semeion can show

  • Which reports are repeated or related.
  • Which reports may trace back to the same original source.
  • Where timing or behavior may point to coordination.
  • Why an allow, hold, or reject result changed.

Semeion does not

  • Declare a claim true or false.
  • Decide who is a bot.
  • Infer a person or system's intent.
  • Punish, block, or act on its own.
Start a conversation

Start with one concrete question.

Request a focused pilot, ask a product question, or talk with us about a partnership or strategic opportunity.

Request a pilot Contact Semeion