Provenance
What Normalize the News can actually establish about the outlets it tracks, and what it cannot.
Provenance here means signals that already exist to be checked, not a score and not a rating. Two of them are facts about NTN’s own database: how many of an outlet’s articles are under revision watch, how many of those the Internet Archive has captured at least once, and how many revisions NTN has published and stands behind. The third is a lookup: whether the outlet appears in a snapshot of two public news-trust programs, the Journalism Trust Initiative and the Trust Project, taken on 2026-07-11. The absence of a signal is not evidence of wrongdoing. An outlet missing from a registry has not failed a check. Nobody ran one. How each of these is produced ›
Every outlet NTN tracks
| Outlet | Trust-registry standing | Articles tracked | With an archive capture | Revisions published |
|---|---|---|---|---|
| amNewYork | Not in this snapshot | 173 | 95 | 4 |
| City & State NY | Not in this snapshot | 45 | 26 | 1 |
| City Limits | Not in this snapshot | 33 | 19 | 0 |
| Documented | Not in this snapshot | 48 | 21 | 0 |
| Gothamist | Not in this snapshot | 154 | 81 | 0 |
| Hell Gate | Not in this snapshot | 39 | 25 | 0 |
| New York Daily News | Not in this snapshot | NTN holds no articles from this outlet | ||
| New York Post – Metro | Not in this snapshot | 282 | 148 | 1 |
| Patch NYC | Not in this snapshot | 45 | 22 | 1 |
| Politico New York | Not in this snapshot | NTN holds no articles from this outlet | ||
| Spectrum News NY1 | Not in this snapshot | 134 | 71 | 6 |
| The City Reporter | In snapshot, neither program | 47 | 24 | 1 |
Across the 10 outlets NTN can read: 1000 articles tracked, 532 with at least one Internet Archive capture, 14 revisions published. 2 outlets in the set (New York Daily News, Politico New York) block NTN’s reader, so nothing has been tracked or checked for them and they are left out of these totals.
None of these outlets is recorded as a member of either program. That is a fact about which outlets have joined two specific programs, not a finding about any newsroom’s journalism. Both registries skew national and international, neither is a prerequisite for good local reporting, and an outlet absent from the snapshot has not failed a check. Nobody ran one. 1 of the 12 outlets hereappears in the snapshot at all.
What this page does not claim
Content Credentials are not verified here. The C2PA parser in this codebase is a stub that always reports no manifest, which is a statement about NTN’s reader and not about anyone’s photograph. Rather than print “no Content Credential” against twelve outlets and let it read as twelve findings, the table leaves the signal out entirely until the check is real.
AI-disclosure and sponsored labels are checked live, and not stored. Paste a link into the box above and NTN fetches that page and looks for them right then. Nothing writes the result down, so there is no per-outlet history of them to report, and the table does not pretend otherwise. That live check is labelled unreviewed on its own output for the same reason.
The registry snapshot is a file, not a live feed. It was transcribed by hand on 2026-07-11 from the two programs’ own published member lists and covers 15 outlets, most of them national or international. An outlet can join a program the day after a snapshot is taken. A standing shown here is only as good as that date, which is why the date is printed rather than implied.
Revisions published excludes revisions NTN withdrew. A revision that was published and then found to be wrong is withheld from every public read on this site, and it does not count here. That has happened once. The row is kept rather than deleted so the error stays auditable, and when the correction is written it will appear on Corrections rather than being quietly absorbed into a count.
Two of the outlets have no counts at all. They block automated readers, which is their right and which NTN respects, so nothing from them has ever been tracked or checked. Reporting them as zero would present three findings where no check ever ran, so they are named, marked as unread, and left out of the totals. This is the same rule the coverage ledger follows.