I doubt it's any of the two. At the scale they operate, they will have multi level checks - automated and manual. Some automated checks would approve the ad, an

I doubt it's any of the two. At the scale they operate, they will have multi level checks - automated and manual. Some automated checks would approve the ad, and then if the ad is getting a huge reach and engagement, they may have a manual user to look at it again, and probably disapprove it. Same with how some images can be posted, and then removed later if they get good engagement > The group pulled the ads before they were posted or paid for This is just a way to create buzz about an upcoming study/report at Facebook's expense. I don't like Facebook, but if they/AP were really serious, we could see the report and look at the way they executed the whole experiment. The other data point is that how many did they try, cos as far as I know, they use AI to detect text on images. Probably not that effective in Thailand's local language. One other aspect here, while Facebook deserves no sympathy, let's not forget misses happen even with editorial oversights. Let's not forget how NYT exclusive about nuclear weapons in Iraq was the final trigger for the war, and that report was false.