Having worked for these companies some of the data is murky (e.g. did these ads they saw earlier lead to them buying the product later, perfect attribution is obviously impossible) but a lot of it is unambiguous where they tracked people straight from clicking on the ad directly to a purchase. People have their conspiracies but I've seen it in black and white, it's very very clear. The only way I could see the data not being clear is in the case of outright fraud, which I'm fairly certain wasn't happening within our own metrics (as it would not only lead to legal liability but even more importantly fuck up the machine learning models).

Edit: to be clear I would believe the effect of ads is overstated, it's just the idea they are ineffective is wrong and people claiming that you can get more effective ads without tracking people at all doesn't seem plausible based on what I know of the industry. I could see contextual ads working in niche use cases (which again we already see when searching for products. YouTubers have relevant sponsors all the time. We even have affiliate marketing, where it's not only contextual but part of the content).