Here are two options, I'm sure there are more:

- Require third party apps to install an impression tracking SDK

- Return URLs in API responses to media hosted on Twitter's servers and do tracking server side

You could contractually require it, and verify it with random sampling on the client apps. If a Twitter employee views a tweet on the third-party app and Twitter doesn't get an impression notification, then there is a problem.

Seems practical. They already have an army of content moderators responsible for millions of feeds. Having them spot-check several hundred or several thousand third-party clients once every month or two doesn't seem like a large increase in their workload.

I mean, that's a single full-time employee (or less). If there are 2000 third-party clients and the goal is to spot-check them once per month, this hypothetical full-time employee would need to check 80 of them per day. That's one spot check every six minutes for eight hours a day if they work 25 days a month.

Not exactly the most exciting work, but easily doable. And that doesn't need to exactly be a highly paid employee either. I believe the big social media companies offshore a lot of that grunt work to other countries.

So we're talking what, thirty or forty grand a year for Twitter to implement that?

Have a sandbox level, then Twitter reviews the app to ensure impressions are being accurately reported, and then allow it to go live.

This is exactly the approach that resulted in the Cambridge Analytica scandal.

Have the clicks redirect through a twitter server?

impressions are not clicks.

No, but impressions do require downloading content from a server, which can be tracked. The trouble would be cases where apps are preloading content, so it's downloaded but the user might not have seen it yet.

Or - as is often the case - the consumer is caching api responses for some period.