Methodology

This page explains how Taipei Closing Bell is produced: where the data comes from, how AI is used, and how we handle mistakes.

Data sources

  • Taiwan Stock Exchange (TWSE) official statistics — daily trading value of the three major institutional investor groups (foreign investors, securities investment trusts, dealers), index data, and per-stock institutional flows, via TWSE’s public data APIs
  • Market data feeds — index levels, individual stock prices, and USD/TWD from established global market-data services
  • Primary documents — company announcements and exchange filings where relevant

Every specific figure in our articles is date-stamped. If a figure cannot be traced to one of the sources above, it does not get published.

How AI is used — full transparency

Articles are drafted by AI language models working from the verified data described above. Before anything is published, every draft passes through an adversarial review gate: a separate AI model from a different vendor cross-checks each specific figure against the source data, flags unsupported claims, checks for advice-like language, and can block publication entirely. Posts that fail review are held as drafts for human review. Editorial standards, topic selection, and this methodology are set by a human editor.

Corrections policy

When we get something wrong, we correct the published article and mark the correction visibly in place, with a note explaining what changed. We do not silently rewrite history.

What we are not

We are a data-driven information publication. We do not provide investment advice, price targets, or recommendations to buy or sell any security. Nothing on this site should be read as a solicitation.