Skip to content

Product

Research does not end at publication.

Trace is one loop with five modules. Measure how a paper is found and used by generative engines, model how the finding is likely to be read, improve the language and metadata costing you clarity, publish with a fair path into the record, and then keep measuring.

DiscoverMeasure what is foundSimulateModel how it is readImproveClarify without changingLaunchPublish with attributionMonitorMeasure again
Discover: Measure what is found. Simulate: Model how it is read. Improve: Clarify without changing. Launch: Publish with attribution. Monitor: Measure again. The loop returns to measurement.

The loop

Five modules, one measurement.

Each module answers a question the previous one raises. The loop closes because monitoring changes what you measure next.

  1. 01

    Discover

    Is the right evidence being found at all?

  2. 02

    Simulate

    How will the finding be read once it leaves your hands?

  3. 03

    Improve

    Which specific sentences and fields are costing you clarity?

  4. 04

    Launch

    Does the published record give the work a fair chance of being found?

  5. 05

    Monitor

    What changed, and when?

01

Discover

AI visibility and research discoverability.

Run a versioned question set across generative engines and record, per question, whether the paper is retrieved, cited, drawn on, and represented faithfully. Retrieval and citation are measured separately because they fail separately.

Discover in detail

Outputs

What you get back

  • Evidence path across five stages
  • Per-engine and per-question detail
  • Competitor papers answering the same questions

Demo workspace figures: 62% of benchmark questions retrieved, 43% cited, 31% with evidence used, 27% interpreted faithfully.

02

Simulate

Synthetic stakeholder panels and artificial societies.

Synthetic stakeholder panels read the manuscript or abstract and respond in role — clinician, methodologist, payer, editor, journalist. The output is directional: it surfaces the misreadings worth pre-empting, not a forecast of opinion.

Simulate in detail

Outputs

What you get back

  • Panel responses by stakeholder role
  • Predicted misinterpretations, ranked
  • Objections a reviewer is likely to raise

03

Improve

Pre-publication clarity and discoverability optimization.

Pre-publication review of title, abstract, keywords, structured metadata, and claim phrasing. Every recommendation names the failure it addresses and the stage of the evidence path it affects. Nothing is rewritten for you without review.

Improve in detail

Outputs

What you get back

  • Clarity and specificity findings
  • Metadata completeness checklist
  • Claim-strength and hedging review

04

Launch

Post-publication availability and attributable distribution.

Availability, identifiers, machine-readable structure, and attributable author and institution assets. Distribution is transparent and attributed, or it does not happen.

Launch in detail

Outputs

What you get back

  • Availability and indexing checks
  • Structured metadata and schema output
  • Approval workflow before anything is published

05

Monitor

Longitudinal visibility and interpretation monitoring.

Scores are re-run on a schedule and kept as history. Model versions change, competitor papers appear, and the questions people ask drift. Changes are annotated as occurring after an event, never as caused by it.

Monitor in detail

Outputs

What you get back

  • Score history with methodology version
  • New competitor detection
  • Query drift and alerting

Trace Visibility Index

TVI_v0.1

Every module reports into one versioned composite, and every composite ships with its components. A single number cannot tell you whether a paper was never retrieved or retrieved and then misread, so the components are never hidden behind it.

How the score is built