Stable scope
The consumer must know whether a record represents a root/domain scope, one normalized exact URL, or another explicitly defined object.
AI systems need stable evidence primitives, not another authority they must blindly trust
Retrieval and generation systems are good at finding and transforming information. The harder problem is preserving enough context to know what was observed, where it came from, whether the scope changed, and when the available evidence is insufficient for a confident conclusion.
What is an evidence primitive?
A useful web evidence primitive is a small, stable, machine-readable building block that answers one bounded question without pretending to answer every trust question at once. A timestamp can identify when an observation was made. A hash can help establish record integrity. A verification event can establish control under a defined method. A deterministic comparison can establish that supported state changed.
None of those primitives is a universal truth oracle. Their value comes from being explicit, reproducible, composable, and difficult to confuse with a different class of evidence.
Requirements for a useful machine-readable evidence layer
Provenance
The record should identify what source was observed, when it was observed, and which system produced the record.
Integrity
Where supported, fingerprints and hashes can help establish consistency of stored evidence without being misrepresented as proof that the content is true.
Verification context
Proof of control or authorization should be represented separately from passive observation and separately from content-quality assessment.
Longitudinal state
Repeated observations should preserve earlier evidence so consumers can inspect continuity, change, and gaps.
Deterministic comparison
When mechanical evidence can answer a question reproducibly, the system should not replace it with an opaque model judgment.
Explicit uncertainty
Unknown, unavailable, insufficient-history, observation-gap, and insufficient-evidence states should remain first-class outcomes.
Bounded interpretation
Derived scores and model-assisted analysis should expose enough context for consumers to distinguish them from underlying observations.
How GoGuides implements that model today
| Public surface | Primitive or layer |
|---|---|
| Evaluate | Authoritative GoGuides assessment for a supported normalized scope. |
| Web State V3 | Stored objective state with provenance and availability semantics. |
| Change Intelligence | Deterministic observation-pair comparison. |
| Verification | Control/authorization evidence represented separately from passive observation. |
| History | Longitudinal observations that preserve earlier recorded state. |
| Claim Support | Bounded model-assisted analysis against a fixed stored source observation. |
| Content Relationships | Conservative relationship evidence for two exact URLs. |
| URL Request | Explicit attributed acquisition intent for one public URL. |
| machine-use.json | Machine-readable capability discovery. |
| OpenAPI 3.1 | Formal public interface contract. |
A primitive should support disagreement
An outside system should be able to inspect a GoGuides record and decide that a different weight, threshold, ranking policy, citation rule, or safety policy is appropriate. That is a feature, not a defect. Infrastructure is more useful when it exposes evidence cleanly enough for different consumers to apply different downstream policies.
Why time matters
A new crawler can observe today's page. It cannot recreate an observation that was never made months or years earlier. Longitudinal evidence therefore becomes more informative with time — not because age makes a source true, but because continuity, changes, and observation gaps become independently inspectable.
The goal is composable evidence
GoGuides is not asking machines to replace their judgment with a GoGuides judgment. The goal is
to provide stable evidence and bounded interpretation that a machine can combine with its own
retrieval, ranking, citation, safety, and decision systems.