@therealwolf's website https://steemapps.com actually uses SteemSQL as its database as he explains in his post: https://steemit.com/steem/@therealwolf/alpha-steemapps-com-the-frontpage-for-apps-on-steem
Hence that's why you have very similar numbers.
Then https://www.stateofthedapps.com/ is getting the data from his website.
@therealwof also mentioned that the user counts on the apps are so low as you can only tell which app the user used for a post or comment, but not which one he did to vote with. So the activity should obviously be higher.

Hey and machine learning could definitely be used to filter out posts by bots, but one first needs to program a software to analyze the blockchain and then also train the model with manual data sets. That's going to take some time...