Evidence intelligence for the AI era
Trace how evidence travels.
Measure whether research is discovered by AI, stress-test how different audiences may interpret it, and launch publications with the metadata, clarity, and availability they need to be found.
Built for researchers, publishers, institutions, and scientific teams.
The missing layer
Citation impact tells you what happened. Trace shows what is happening now.
Search is becoming generative. A paper can be highly cited yet absent from the questions researchers ask AI. Another can be retrieved often but interpreted incorrectly. Trace separates retrieval, citation, evidence use, fidelity, and availability so you can see where the path breaks.
Relevant
100%
1,240 queries
Retrieved
62%
769 queries
Cited
43%
533 queries
Evidence used
31%
384 queries
Interpreted faithfully
27%
335 queries
Questions this study should be able to inform.
62% of previous
69% of previous
72% of previous
87% of previous
Illustrative figures from the demo workspace. Each stage is measured independently — a paper can be retrieved often and cited rarely, or cited often without its evidence actually being used.
One paper, all the way through.
scroll to advance
01 / Discover
See what AI sees.
Every paper becomes a benchmark of the questions it should be able to answer. Trace measures where it appears, what replaces it, and whether the evidence is actually used.
- Discover
- Simulate
- Improve
- Launch
- Monitor
Discover: See what AI sees.
Every paper becomes a benchmark of the questions it should be able to answer. Trace measures where it appears, what replaces it, and whether the evidence is actually used.
34% of relevant questions use its evidence
Simulate: Read the paper through another mind.
Synthetic stakeholder panels surface likely misreadings and objections before expensive human research. Every output is labelled simulated.
27% read the relative effect as the absolute one
Improve: Easier to find. Identical in meaning.
Terminology gaps, ambiguous abstracts, and missing structure. Any suggestion that would strengthen a claim is discarded rather than shown.
0 suggestions that change claim strength
Launch: Publication is a beginning.
Metadata, availability, institutional presence, and attributable author communications — each with a disclosure and an approval step.
23 availability and metadata checks
Monitor: Then measure it again.
Visibility changes. When an open-access deposit breaks, the availability score falls and the regression is flagged — recorded as having occurred, never as having been caused.
5 measurements over 8 months
Competitive intelligence
When your paper is missing, know what appears instead.
When Trace expects a paper to be relevant and it is absent, it records which sources appeared in its place — and what those sources have that the target does not.
Query
Does velunemab reduce flare frequency in adults with Novera syndrome?
10 runs · 4 engines · sampled 2026-04-12
TRACE-101 primary resultsYour paperRandomized trial · Phase 3 · 2026 · 10.5555/trace-demo.101Cited in 2 / 10 runs
Possible contributing factor
- Study design
- An individual trial competing against synthesis sources on a broad, guideline-shaped question.
- Availability
- Full text is not deposited in an open repository, so only the abstract is indexable.
- Terminology overlap
- The title names the trial acronym rather than the condition and the intervention together.
Anti-IL-31 therapies in inflammatory skin disease: a meta-analysisMeta-analysis · 2025 · 10.5555/meta-il31.2025Cited in 8 / 10 runs
Possible contributing factor
- Study design
- Synthesis sources answer broad effectiveness questions in a single citation.
- Availability
- Open access with full text available to indexers.
- Terminology overlap
- Title terms map directly onto the phrasing used in the query.
International guideline for the management of Novera syndromeGuideline · 2025 · 10.5555/novera-guideline.2025Cited in 7 / 10 runs
Possible contributing factor
- Study design
- Guidelines are frequently preferred for treatment-selection questions.
- Availability
- Freely readable and mirrored across several domains.
- Terminology overlap
- Covers the whole condition, so it matches many phrasings of the same question.
NSI-24: a randomized trial of an earlier anti-IL-31 antibodyRandomized trial · 2021 · 10.5555/nsi24.2021Cited in 4 / 10 runs
Possible contributing factor
- Study design
- Comparable design, with five more years of citing literature.
- Availability
- Full text deposited in a public repository.
- Terminology overlap
- Names the condition in the title and abstract's first sentence.
These are associations observed across sampled runs, not causes. Trace does not have access to how any engine ranks sources and does not claim that changing a listed factor will change a result.
Methodology
A score you can inspect.
Trace does not collapse research visibility into a black box. Every index is built from observable runs, versioned query sets, explicit sampling, and source-level evidence.
Trace Visibility Index
TVI · TVI_v0.1
- Retrievalrelative emphasis Highest
- Share of benchmark questions where the paper enters the retrieved source set.
- Citationrelative emphasis High
- Share where the paper is visibly referenced in the answer given to the reader.
- Evidence userelative emphasis Supporting
- Share where the answer draws on findings actually attributable to the paper.
- Fidelityrelative emphasis Supporting
- Whether population, effect size, and limitations survive the summary intact.
- Breadthrelative emphasis Supporting
- How many distinct question types reach the paper, not how often one of them does.
- Stabilityrelative emphasis Supporting
- How consistent the result is across repeated runs, engines, and model versions.
Availability carries the remaining a supporting weight. Components are always reported alongside the composite, because a single number cannot tell you whether a paper was never retrieved or retrieved and then misread.
The Trace Visibility Index is a product heuristic under active validation, not a validated scientific metric.
Enterprise
Built for evidence portfolios, not just individual papers.
Trace runs the same measurement across thousands of papers, with shared query sets, role-based access, and exports your reviewers can check.
Publishers
See how your corpus surfaces across AI-mediated discovery and where metadata or availability is holding it back.
Journal, imprint, and subject-level views over a whole catalogue.
Institutions
Understand which research is visible, under-discovered, or misinterpreted across departments.
Departmental benchmarks with repository and identifier coverage.
Life sciences
Monitor published evidence, stress-test scientific interpretation, and manage attributable launch workflows with review controls.
Separated scientific-exchange and promotional modes, with review trails.
What happens to your paper after publication?
Run a baseline.
A baseline measures one paper across the current query set and returns the evidence path behind every number. Trace never promises a ranking improvement.