Which papers published similar text on the same day?

From 6–12 October 2026, 582 of 4,778 eligible article records (12.18%) had a similar body in another paper on the same issue date. These are algorithmic candidates, not a measure of government-authored news.

The sample has 24–25 newspapers a day. From 5,619 usable records, we excluded 161 known cut bodies, 676 bodies with fewer than 200 normalized characters, and 4 same-paper, same-day identical bodies. Different papers and dates remain separate.

We compared sets of seven-character fragments after Unicode normalization and lowercasing, keeping only letters and digits, including Chinese characters. A cosine score of at least 0.6 flags a same-day, cross-paper pair; each matched record counts once. This is a different method from the original study’s word trigrams.

Some bodies are excerpts or contain page navigation. A missing cut flag does not guarantee a complete article. The dates include the end of the National Day holiday, and coverage changes by day. The candidate share is not a national estimate.

AI-assisted close reading examined twelve pairs selected across similarity levels. This was a small diagnostic check, not independent human validation; accuracy and missed matches remain unmeasured. The example above was checked against the stored records. Original links are supplied for inspection; their current availability was not checked.

Inspired by Waight, Yuan, Roberts and Stewart (2025). Their study used additional evidence to identify government scripting and examine long-term change. This experiment does not reproduce that result.