Honestly the AI product in its current form is shit, authenticity is king in this space, interactivity is also valuable, how many youtube channel sign off with "what do you think, leave a comments below" beyond the algorithm people like interacting with something that feels real. If the output feels AI then you might as well use a robot voice. I assume the point of this is to push out content on the cheap. But to make it good/worthwile to consume would require non-content improvements.
I think you have the wrong use case/customer for the short term sentiment data.
If bringing in more traffic is your primary goal then using highly targeted advertisement on Facebook or equivalent would probably be more effective (the offering is quite scary, a campaign at a board level rather than a site level would probably quite effective though the mods probably won't appreciate the influx, and you'd risk a cultural shift of the site if too successful), what percentage of AFL supporters would be aware of the site in the first place. How many do you actually want to capture.
Looking into causes of friction for new users, one average there would be a substaintial gap of time between when users create the account and when they first post. You'd probably see a decent boost by adding a positive affiirmation encouraging people to post that appears the next time someone logs in after 10 unique times of lurking without posting.
