# HD-INC-037: CNET Published 77 AI-Written Articles. More Than Half Needed Corrections. > From late 2022 CNET ran financial explainers generated by an AI tool under a 'CNET Money Staff' byline that revealed the automation only when readers hovered over it. After Futurism reported it, a review found errors in 41 of the 77 articles, some of them substantial. - Source: https://useheadlights.com/library/HD-INC-037/ - Publisher: Headlights (useheadlights.com), author Ellie Harris - Published: 2026-06-20 - Company: CNET (Red Ventures) - Sector: Technology - Country: United States - Year: 2023 - Behaviour in the record (Headlights reading): Automation bias; also Normalisation of deviance, Concealment - Governance domains: Consumer protection, Identity & authenticity - Licence: CC BY-NC 4.0. Cite as: Harris, E. (2026). CNET Published 77 AI-Written Articles. More Than Half Needed Corrections.. Headlights AI Incident Library, HD-INC-037. https://useheadlights.com/library/HD-INC-037/ ## What happened It was reported that between November 2022 and January 2023, technology publisher CNET published 77 personal finance articles written using AI, without making it immediately obvious to readers that the content had been generated by a machine. The articles appeared under the byline "CNET Money Staff", with a disclosure about automation visible only when readers hovered over the author's name. The arrangement attracted attention in January 2023 when journalists discovered the practice and began examining the accuracy of the articles. One article explaining compound interest claimed that a 10,000 USD investment earning three per cent annually would generate 10,300 USD in interest during its first year, confusing the total account balance with the 300 USD actually earned. Following an internal review, CNET acknowledged that 41 of the 77 articles required corrections, ranging from relatively minor wording problems to incorrect figures and more substantial factual errors. Some articles also contained passages that were not entirely original, raising additional questions about how the content had been produced and reviewed. CNET subsequently paused its AI-writing programme and revised its approach to disclosure, while maintaining that AI could still play a useful role in journalism when appropriate editorial safeguards were in place. ## What an auditable version would have shown An auditable system would have recorded which parts of each article were generated using AI, what factual claims had been checked and who approved the final version before publication. This would have allowed editors to identify unverified financial information before readers encountered it, while providing a clear record of how AI had contributed to the published work. ## Where the gap was Headlights identifies **overconfidence and a rules-behaviour gap**, where organisations become comfortable relying on technology without ensuring that the normal standards of accuracy and accountability are being maintained. CNET was using AI to produce financial information more efficiently, but the number of articles requiring correction suggests that the editorial checks were not consistently sufficient to catch errors before publication. There is a familiar human behaviour behind this failure because when technology makes a task faster and easier, people can gradually become less attentive to the work they previously performed themselves, particularly when the results appear polished and professionally written. The problem was not that AI helped produce the articles, but that readers were given financial information under the authority of an established publisher without adequate assurance that the facts had been checked. ## What governance should have looked like Using AI to assist with journalism does not change a publisher's responsibility to verify information, particularly when readers may rely on financial advice to make decisions about their money. A VerificationGate could check figures and factual claims against reliable sources before publication, while a ConductRecord could document which content was AI-generated, what had been verified and who was responsible for approving it. Readers should also be told clearly when AI has been used to produce an article, rather than having to discover that information through a disclosure that is easy to overlook. The broader lesson is that producing information faster has little value if the time saved in writing is later spent correcting mistakes, particularly when those mistakes undermine the trust readers place in the publication. **Failure Pattern:** CNET published 77 AI-generated personal finance articles, with 41 subsequently requiring corrections, exposing weaknesses in factual verification and transparency around AI-generated content. **Governance Principle:** AI-assisted publishing should maintain clear editorial accountability, independent verification of factual claims and transparent disclosure of how automated tools contributed to the final content. The reference implementation of VerificationGate and ConductRecord is open source. It lives at [github.com/saffronandindia/headlights-oss](https://github.com/saffronandindia/headlights-oss), Apache 2.0 licensed and free to install. The repository is public now. ## Sources - [Plagued with errors: A news outlet's decision to write stories with AI backfires (CNN Business)](https://www.cnn.com/2023/01/25/tech/cnet-ai-tool-news-stories) - [Publisher breaks news using bots to write inaccurate stories (The Register)](https://www.theregister.com/2023/01/19/cnet_reviewing_ai_authored_stories/) - [CNET Is Quietly Publishing Entire Articles Generated By AI (Futurism)](https://futurism.com/the-byte/cnet-publishing-articles-by-ai) ## Related - [HD-INC-038](/library/HD-INC-038) - [HD-INC-141](/library/HD-INC-141) - [HD-INC-010](/library/HD-INC-010)