Skip to content

Method

This site is a data desk. Stories start from official open data, never from another news outlet.

Sources

Sources whose terms we could not verify, or whose robots.txt disallows access, are not used. MTA data is downloaded to our own server; MTA does not endorse this site, and data may not be real time.

Steps

  1. Collectors copy the data into our own database and keep its history.
  2. Code computes the comparisons. Each value stores its source, query and retrieval time, and is recomputed on a separate connection to verify it.
  3. No story is written unless it adds at least 3 verified facts not already in the source. If a similar story ran in the last 30 days, no new story is written.
  4. An AI-assisted writing tool drafts the text from those facts only; it does no arithmetic.
  5. Code checks: every number must trace to a fact or to a dated source record, no long quotes from the source, no repeated templates across stories.
  6. A different AI model checks every claim against the source. Stories with a wrong or misleading claim are not published.
  7. Stories naming private people, or involving crime, health, deaths or elections, are never published automatically.
  8. At most 6 stories a day, at least 45 minutes apart.

Images

Charts are drawn by code from the data. Some stories carry an AI-generated representative illustration, labeled as such and tagged with IPTC DigitalSourceType metadata. We never use images that look like documentation of a real event.

Human review

The editor responsible for publication is Utku Eroğlu. Stories that name a private individual or involve a crime allegation, health, death or an election are not published without the editor's approval. All other stories are published after passing automated checks. To report an error or request a correction: info@newyork24.news.