Methodology
How MAP documents the observable information environment around a URL or publisher — across seven dimensions, with no truth, bias, or credibility judgment.
MAP documents seven dimensions of observable information structure. There is no composite score and no pass/fail threshold — each record shows what was found for each dimension, plainly, including when a dimension has little or nothing to report. Publisher certification is a separate application process reviewed by MAP's founder (see below), not an automated calculation from these dimensions. No publisher has been certified yet.
Origin
Publisher, author, and publish date, drawn from the page's metadata and structured data where available. MAP reports what's documented — not whether the source is reputable.
Attribution
Structural phrasing patterns associated with attributed claims ("according to," "officials state") and outbound references to named sources.
Specificity
Body content depth (word and paragraph counts), named dates, and name-like sequences suggesting people, places, or organizations.
Timing
Whether the headline and description structurally lean toward past, present/ongoing, or future/anticipated framing.
Framing
Structural presentation patterns — listicle, opinion, live-update, or standard format — and framing indicators like urgency or high-engagement phrasing.
Context change
A record's Context Change frame states how many feed observations of the URL MAP holds and points to the differences, if any. MAP compares consecutive stored records of the same URL from the same pipeline, skips observations where the page could not be read, and lists the differences on the What Changed page and in the record. Coverage is opportunistic: only URLs observed more than once can be compared.
Information relationships
Outbound references to other domains, surfaced as clickable material — shown to demonstrate how the topic connects elsewhere, not to vouch for what's on the other end.
Intake
A URL is submitted for analysis — by a reader, via API, or through MAP's syndicated content pipeline.
Extraction
MAP's analysis engine fetches the page and extracts metadata, body content signals, and structural patterns — no external judgment inputs, no manual scoring.
Record assembly
The seven dimensions are assembled into plain-language sentences — each one traceable to something genuinely extracted, with gaps disclosed rather than papered over.
Ledger entry
The record is written to the MAP Verification Ledger as a SHA-256 hash-chained, append-oriented entry. Corrections create additional entries rather than silently replacing earlier observations.
Publisher certification (separate process)
Certification is a distinct application process, with the decision made by MAP's founder — not derived from any URL's context record. No publisher has been certified yet. Once a publisher is certified, its business category and certification date are recorded on the same ledger.
How automation and AI are used
Context records are generated automatically, usually within minutes of MAP seeing the page (a queue processes them), and written to the ledger with no per-record human review before publication. AI tools may assist in extracting structured context from publisher content; MAP's founder maintains the extraction code. AI does not issue scores or make final editorial decisions. See the AI Disclosure Policy.
Repeat observations of the same URL
MAP compares consecutive stored records of the same URL — headline, headline characteristics, author, publish date, referenced domains and presentation pattern — and lists any differences as observed differences. A URL is re-observed when it reappears through MAP's feed ingestion or is submitted again, so these are mostly short-window comparisons, not long-term tracking. A difference shows that two MAP observations differed; it does not establish that the publisher edited the content or which version is correct. See What Changed.
Errors in a record
If a context record contains an error, MAP writes a new ledger entry that notes the correction and leaves the original in place. Corrections are handled manually; the ledger does not yet link a correction to the entry it corrects. See the Corrections Policy.
What a record can show about its own origin
Each record has an ID, the source URL, the observation time, a SHA-256 hash of the record and the hash of the record before it. For records MAP ingests from publisher feeds, the record page shows these values, and MAP checks daily that each entry's previous hash matches the entry before it, and publishes the result, including the count of exceptions it finds. An entry's previous hash can belong to a check submitted by a visitor, which MAP does not publish. MAP operates the ledger, does not publish the data needed to recompute the hashes and does not yet offer a lookup by hash, so readers cannot yet verify continuity independently; they rely on MAP's own report. Provenance shows where MAP's information came from and when; it does not show whether the content is accurate.
What MAP does not do
MAP does not determine whether content is true, false, biased, or trustworthy, and it produces no composite credibility score. A record reflects the page as MAP fetched it at one moment; pages change, and paywalled, login-gated, or script-rendered pages may expose less. Extraction is structural, so it can miss or misread metadata. Differences can be listed only for URLs MAP has observed more than once, and the ledger shows what MAP observed, not everything that was published.
Version 1.0
Last revised October 3, 2026. Material changes to this methodology will be listed here with their date.