How we use AI
Archived articles are not touched by AI beyond the technical work of stripping out old navigation bars, ad slots, and broken layout code. The text you read in a 2017 or 2020 post is the text that was published then, preserved as a record of what was reported at the time.
New evergreen guides are a different story. AI drafts structure, suggests phrasing, and speeds up the assembly of explanatory articles. A human editor sets the scope, checks every claim, rewrites anything that sounds wrong, and makes the final call on what gets published.
AI does not invent facts. It cannot interview a district officer, dig up a budget line, or verify a timeline. When a guide cites a number or describes a process, that information comes from documents, prior reporting, or contributor knowledge. The AI is a drafting tool, not a reporting tool.
We do not use AI to rewrite old posts. If an archive article feels dated or incomplete, we publish a new guide on the same topic rather than silently updating the original. The old piece stays in place with its original wording, and the new one links back for context.
AI helps with translation and localization when we adapt material for different audiences, but a fluent human always reviews the output. Machine translation is fast and cheap, but it misses nuance, flattens tone, and occasionally inverts meaning. We catch those problems before anything goes live.
The site structure, navigation, category pages, related-article links, is partly AI-generated and partly hand-tuned. The goal is to make it easy to move from a breaking-news archive piece to a broader explainer without getting lost. AI suggests connections; editors decide which ones make sense.
We do not use AI to generate fake quotes, synthetic interviews, or imaginary case studies. If a guide includes an example, it is either drawn from prior reporting or presented as a hypothetical. We will never dress up an AI hallucination as a real person or event.
This approach will evolve. Tools improve, and so do our editing processes. What will not change is the firewall: AI drafts and suggests, humans verify and decide. If that balance ever shifts, we will update this page and explain why.