Comment by Animats
The big problem in open source intelligence is not in-depth analysis. It's finding something worth looking at in a flood of info.
Here's the CIA's perspective on this subject.[1] The US intelligence community has a generative AI system to help analyze open source intelligence. It's called OSIRIS.[2] There are some other articles about it. The previous head of the CIA said the main use so far is summarization.
The original OSINT operation in the US was the Foreign Broadcast Monitoring Service from WWII. All through the Cold War, someone had to listen to Radio Albania just in case somebody said something important. The CIA ran that for decades. Its descendant is the current open source intelligence organization. Before the World Wide Web, they used to publish some of the summaries on paper, but as people got more serious about copyright, that stopped.
DoD used to publish The Early Bird, a daily newsletter for people in DoD. It was just reprints of articles from newspapers, chosen for stories senior leaders in DoD would need to know about. It wasn't supposed to be distributed outside DoD for copyright reasons, but it wasn't hard to get.
[1] https://www.cia.gov/resources/csi/static/d6fd3fa9ce19f1abf2b...
[2] https://apnews.com/article/us-intelligence-services-ai-model...
The really big problem in open source intelligence has been for some time that data to support just about anything can be found. OSINT investigations start with a premise, look for data that supports the premise and rarely look for data that contradicts it.
Sometimes this is just sloppy methodology. Other times it is intentional.