How we use AI
Dental Market Radar Editorial Team · Updated
Where AI is used at Dental Market Radar, what checks apply, and what AI does not do.
Last updated:
Where AI is used
- Research. The fact database is built in AI-assisted research sessions that search and read public sources. Each figure keeps its quote and locator, so a reader can check it.
- Writing briefs. An AI pipeline writes the brief in five steps: fetch pages, extract candidate facts with verbatim quotes, write, check, and render the text.
- Review help. A second model reads the text with a fresh context. It checks the meaning: the right country, metric and date, and no overstatement.
Checks that do not rely on a model
- Every number in a brief must appear in the verbatim quote of a source it cites. The quote must be found in the saved page.
- The country, metric, period, unit and subgroup of a claim must match the cited fact.
- A quote the code cannot find in the saved page is rejected.
- A source URL the pipeline was not offered is dropped.
- A brief that fails a check is rejected and goes no further.
- A run stops at a fixed spending cap.
Human review
A human reviewer reads each brief. Nothing is published without approval, and the pipeline cannot publish by itself.
Credit
AI-assisted work is credited to the Dental Market Radar Editorial Team. It is labelled AI-assisted on the page.
What AI does not do
- It does not publish.
- It does not invent people, quotes or statistics.
- It does not change a published figure in place.
Limits
- A model can misread a source. The checks reduce this but do not remove it.
- Errors are logged and fixed as described in corrections.