Decoded AI Tech articles are written with AI. This page explains how, and what is checked before anything is published.
How an article is made
- Story selection. AI models survey recent developments in AI and choose stories to cover.
- Evidence collection. The public sources for each story (company announcements, research papers, filings and news reports) are fetched and stored.
- Drafting. An AI model writes the article from that evidence. Every factual claim links to the source it comes from.
- Checking. Before publishing, automated checks read the full text of every cited source and:
- compare each cited sentence with what its source actually says, including who said it, which way it points, and what a number measures;
- require every quotation to appear word for word in a source;
- recompute any arithmetic the article shows;
-
reject dead links, sections without sources, and claims no source supports.
An article that fails any check is corrected against its sources or not published. -
Editorial scoring. An AI editor scores each article for accuracy, clarity and usefulness, and articles below the bar are not published.
What we do not do
- We do not claim first-hand testing, interviews or measurements. Every fact comes from a published source linked in the article.
- Quotations come only from published sources, and attributions name the person or outlet the source names.
- Articles are not presented as written by named people.
Limits
Automated checks catch many errors, but not all of them. AI can still misread a source in ways a check does not detect, so the linked sources are the authority: follow the links to read them.
Corrections
If an article misstates a source, email hello@decodedaitech.com with the article and the sentence. We check it against the source and correct or remove the article.
Affiliate links
Some articles end with a clearly labelled block of affiliate links. We may earn a commission from those links. They never appear inside the article text, and they are never used as sources.