What the record cannot see

Text held

The writer and several studies can use an article's own text only where the publisher lets it be fetched. Everything else is a headline, sometimes with a short summary. 26 Sep to 2 Oct:

Google News and GDELT supply headlines only. A story written mostly from those rests on how outlets headlined it, not on what they reported; each story page says how many of its inputs carried text.

Copies set aside before counting

When two channels carry the same item (the same address, or the same headline), Argus keeps one and counts it once. In the run of 2026-10-02 00:00 UTC the feeds, Google News and Reddit returned 4028 items; 232 were copies of 201 items already held. GDELT's files for that window held 35201 English articles: 2034 repeated a headline Argus already had, 12230 repeated another outlet's headline word for word, and 823 were kept as new outlets for stories already found.

Which outlets carried each copy, and when, is recorded from 29 Sep 2026; before that only the first carrier was kept.

What only Google listed

Google News's lists are one of Argus's channels. For each item they carried at the run of 2026-10-02 00:00 UTC: did any newsroom feed or subreddit Argus reads carry the same story? Where none did, the story is in the record only because Google listed it, and its breadth is Google's selection, not Argus's.

53 items could not be matched to a story and are left out. "Only via Google" says where Argus's own channels are thin; it says nothing about how widely the story was reported elsewhere. Recorded from 29 Sep 2026.

Always true

• "First" anywhere on this site means first in this record, not first in the world: Argus does not read the wires directly, so a wire story appears through whoever republished it.
• GDELT's times are when GDELT crawled a page, not when it was published. They add outlets to a story, never recency.
• Paywalled, broadcast, print-only and non-English reporting is outside the record.