I Tried the AI Tool That Just Raised $5.3M: Artificial Societies Review

Imagine knowing how your post will perform before you hit publish. That’s what Artificial Societies promises, so I tried it. 

How it works 

You connect your LinkedIn and X accounts, and AS pulls your social graph to simulate your network.   With its LLM-based infrastructure, it generates: 

• Variants → different versions of your post. 

• Insights → summaries of how “personas” (via their AI agent Amos) would react. 

• Conversations → simulated dialogue between these personas, sometimes even phrased as if it came from your own connections.   

I ran some posts through it. The variants suggested didn’t outperform the originals by much in AS’s internal testing. Still, the platform gives each version an Impact Score (out of 100), where anything above 80 is considered exceptional. I think that 80-cap is intentional, it makes the evaluation look fair and realistic. If posts were constantly scoring 100, it would feel fake. 

The insights are useful but, honestly, pretty similar to what a well-structured ChatGPT prompt could also generate. Although the platform claims to be 30% more accurate in predicting engagements compared to standard LLMs (up until GPT4-o), doesn’t seem to me like it is so different. It’s just the fact that not many people would come up with the idea of creating a new project under their ChatGPT and prompting it well enough for it to come up with these variants, critiques and simulated conversations that feel real. Isn’t entrepreneurship all about this? Coming up with an idea that not many could think of, molding it into its best version and executing well. Very well thought and executed, I’ll give that to them.   

Same with the conversations, they surface relevant talking points, but again, nothing you couldn’t get elsewhere. What does make them convincing is the design: the animations and dialogue snippets are presented so well that it feels like a real interaction is happening. 

What stood out to me  

• The design is excellent. The way personas interact makes the simulation feel alive. 

• Direct integration with LinkedIn/X makes the feedback feel personalized. 

• But… since you only post the “improved” version, there’s no way to verify if the feedback truly boosts engagement. Even A/B testing would be shaky because algorithm timing, audience mood, and other factors skew the numbers. 

Honestly, AS looks smartly designed and has the potential to be a very useful tool. But right now, many of the comments and suggestions feel on par with what ChatGPT could already give, just wrapped in a slick, social-simulation package.  Link to AS