It began with a small but important realization:

I had been looking at the screenshots on my Steam page for months, convinced they were “fine.”

What I never really questioned was what players actually see in those images - not what I intended them to see.

Original shot

Motivation

The motivation came from this X.com conversation:

It brought me some feedback, but less actionable than I hoped. Still, it pushed me to think more seriously about what my Steam page was actually communicating. And once that thought appeared, I figured: why not ask ChatGPT to analyze it? Maybe, just maybe, I could get a few useful pointers.

AI Assisted Analysis

What followed was far more eye-opening than I expected.

I had been looking at my Steam screenshots for months, believing they were “fine.” What I had never really questioned was what players actually see - not what I intended them to see, but what is literally on the screen.

The AI analyzed my screenshots, evaluated how well they matched what the game is actually about, and suggested a different ordering. This was probably the most important part, but it is not what I want to talk about today. Apart from the ordering, the AI pointed out that none of the screenshots really captured what the game is about.

When analyzing my woodcutting screenshot, shown at the beginning of this article, the result could be summarized as:

This isn’t a screenshot of chopping wood.

It’s a screenshot of a very large barrel.

Oh, really? I had been so familiar with my own game that my brain filled in the missing meaning.

But players only see what is literally on the screen - and sometimes that’s a barrel.

This moment was the start of a surprisingly useful experiment.

Iteration: Adjusting the Camera Like a Film Crew

Once the obvious issues were out in the open, we (the AI and I) began iterating. Not dramatic changes - just tiny, specific adjustments I had never given enough attention:

AI became an art director and I a camera operator, doing what I was told to do and sending images back for analysis. By the way, the image analysis capabilities are astonishing. It feels as if the AI really understood what it is shown. As much as I hate the “art” AI produces, this is something it does really well.

  • Move the camera 20-30 cm to the right.
  • Raise it slightly.
  • Tilt it down by a few degrees.
  • Avoid foreground objects that overpower the scene.
  • Make sure the player character isn’t glued to the edge of the frame.
  • Let NPCs add life, but not chaos.

And most importantly:

  • Capture the axe at the most readable moment - just before the swing hits the log.

I stepped through the animation frame by frame, discovering how different a tenth of a second looks when you really inspect it. The AI wasn’t giving orders; it was giving reasons. And those reasons made me look at my own game more critically, as if I were seeing it through someone else’s eyes.

The Final Shot

The final shot

Eventually, we captured a screenshot that actually communicates what Bayaya is:

  • A clear, readable action,
  • A visible wooden-toy aesthetic,
  • A lively village behind the player,
  • No overpowering foreground objects,
  • And a composition that makes sense even at thumbnail size.

It’s not “perfect,” whatever that would mean - but it’s undeniably more effective than what I had before.

What I Learned

This experiment taught me something simple but important:

  • Even when you know your game intimately, you don’t automatically see what others see.
  • Familiarity creates blind spots.

AI, with its relentless focus on what is literally present in the image, helped me notice things I had been overlooking before. It didn’t take control away from me, but it did help me step outside of myself and evaluate my own Steam page more objectively.

The updated screenshot and rewritten text are now live.

Final thoughts

Will they actually improve engagement? AI expectation is 20-50% increase in “conversions”. (“Conversion” is what marketing people call it when a user does something we want - buys a product, or, in the case of Steam, wishlists the game.)

We’ll see. This is an experiment - and the real result will come from the players.

You can check the updated page at Bayaya on Steam

… and if the new screenshot finally does the job, you’re very welcome to wishlist the game. :)