Something I'm seeing more often in design feedback is an AI-generated reference arriving in place of an explanation.
A client sees something that doesn't quite feel right, generates an alternative using AI, and sends it back with, "More like this."
And to be fair, I can understand why.
It can be much easier to show someone an idea than explain it, and AI has made that possible for almost anyone.
But I think the problem starts when that generated image stops being a reference and starts being treated as the actual design solution.
When a concept has to work
Take an AI-generated sportsbook or casino lobby, for example. It might look incredible. Beautiful lighting, translucent panels, sophisticated effects and a layout that looks convincing at first glance.
But what happens when that design needs to accommodate real betting markets, changing odds, longer translations, smaller screens, accessibility requirements and interactions that actually need to work?
Suddenly, the visually impressive concept has to function as a real product.
The same applies to campaign creatives. Something that looks fantastic as a single AI-generated image doesn't necessarily translate into a campaign system that works across multiple formats, languages and production requirements.
I don't think the answer is to discourage clients from using AI. Quite the opposite.
Understand what the reference is saying
When someone shares an AI-generated reference, the first thing I want to understand isn't how to recreate it.
It's what exactly they like about it.
Is it the visual hierarchy? The colours? The overall energy? The way the information is presented?
Once you understand that, you can take the valuable parts of the concept and turn them into something that actually works for the brand, the product and the people using it.
For me, that's where experience and design judgement become even more important.
Where experience matters
AI has made it easier than ever to generate something that looks impressive.
But knowing what should be designed, why it works and whether it can actually be built is still a very different skill.
I'm curious how others are finding this.
Are AI-generated references making your design feedback conversations clearer, or are they sometimes creating unrealistic expectations about what the final product should be?