Fact parity & AI-trust audits · by Neural Command
Can machines
trust the facts
on your page?
Paste a URL. ingest measures crawler access, rendered DOM, JSON-LD vs schema truth, and atomic fact parity — evidence for every finding.
How it works
- 01
Fetch like a machine
Raw HTML, crawler identities, robots policy, and rendered DOM — kept separate.
- 02
Extract atomic facts
Only propositions the source supports. Questions are not claims. Qualifiers survive.
- 03
Compare representations
Visible HTML vs JSON-LD vs other machine layers. Contradictions stay visible.
- 04
Preserve evidence
Every finding links back to source text, selector, and measurement confidence.
For operators
See whether AI-facing answers can misstate your offer, pricing, or policy from conflicting page layers.
For practitioners
Audit crawler access, structured data integrity, fact parity, and entity identity with permanent fixtures.
What we check
Eight measurement categories, run before the first score appears — each finding linked to its evidence.
Crawler access
13 public crawler identities, each asked what it is allowed to read.
Robots & access policy
One policy per agent — what blocks machines before content is ever seen.
Raw HTML vs rendered DOM
What changes after JavaScript runs, and whether crawlers see the same page.
JSON-LD parse validity
Whether the structured data parses at all — before asking what it claims.
Structured-data vs visible-copy parity
Fact tension between page layers: machine claims vs what a human reads.
Atomic fact extraction
Only propositions the source supports. Questions are not claims. Qualifiers survive.
Entity identity
One name, one identifier, one definition wherever the entity appears.
Regression fixtures
Permanent fixtures, so two runs are comparable instead of anecdotal.
Questions operators ask
What does ingest measure?
ingest measures whether search engines and AI crawlers can retrieve, parse, extract, verify, and trust the facts on a page — crawler access, raw HTML against the rendered DOM, JSON-LD against schema truth, atomic facts, entity identity, and parity between page layers. Every finding carries its evidence: source text, selector, and confidence.
Why is ingest not an SEO score or a citation guarantee?
ingest measures comprehension and fact parity, not rankings. A page can be perfectly machine-readable and still not rank, and no audit can guarantee an AI system will cite a page. ingest reports what it can prove from the page itself and leaves UNKNOWN as UNKNOWN.
How does ingest differ from agent-readiness checklists?
Agent-readiness checklists ask whether the standard exists — robots access, llms.txt, structured data presence. ingest asks whether the page tells the truth: whether the structured data agrees with the visible copy, and whether a fact survives extraction without being distorted. Both layers matter; ingest owns the evidence layer.
Who runs ingest?
ingest is built and run by Neural Command, with research and implementation led by Joel Maldonado. The methodology and the measurement contract are published openly, and the instrument is free.
Deep enough for experts
The instrument behind this page is the same harness used for crawler matrices, structured-data integrity, fact parity, entity graphs, and measurement-validity CI. UNKNOWN stays UNKNOWN.
- 13 public crawler identities
- Raw HTML vs rendered DOM
- JSON-LD parse validity ≠ schema truth
- Atomic facts with assertion mode
- Permanent regression fixtures