This is a hot take. I am putting that on the label so nobody is surprised. The whole post lives in opinion-piece territory, and if you came here for a balanced takedown of AI in design, you are about to be disappointed.
Here is where I land. AI-generated design work is fine when the alternative is no design work at all. It is not fine when it crosses from polishing the thing into manufacturing the thing.
The criticism is real and I do not dismiss it
Let me start by giving the critique its due. There are real problems with the current AI boom and they deserve to be said out loud.
- Data centers and water. Training and serving these models is energy-hungry in a way that is not invisible. Local communities are noticing.
- Environmental load. The whole stack of chips, datacenters, and cooling has a real footprint, and pretending otherwise is silly.
- Training data. A meaningful chunk of these systems was built on the public output of working artists who were never asked. That is not a small thing.
- Designers losing work. Junior design work is the work most easily replaced by a decent prompt and three iterations. That is a real career impact, especially for people just starting out.
If you stop reading here and walk away thinking I am pro-AI without reservation, you misread the post. The rest of this piece is about what happens next, given that the tech exists and is not going away.
The part where I push back, gently
Here is the wrinkle. I keep seeing the same pattern in 2026: a working musician posts a flyer they made with an AI tool, or a small restaurant posts a banner image they generated, and a chorus of people online dunks on them. Could not be bothered to hire a designer. Soulless slop. Tells me everything I need to know about your business.
I want to put this gently. The US economy is rough right now and the world economy is rougher. Touring musicians are not making touring money. Independent restaurants are running on margins that would horrify you if you saw the books. A lot of the small businesses I work with are choosing every week between paying themselves and paying a contractor.
Those people do not have a design budget. They have never had a design budget. The choice for them was never AI flyer vs hire a designer. The choice was AI flyer vs no flyer. Or AI flyer vs a Word doc in Calibri.
In that context, dunking on them for using the tool is a strange thing to do. It is dunking on people who finally got access to the kind of polish that used to require an agency retainer. I would rather see a working musician put out a hundred shows with AI-assisted flyers than see them quit because the flyer made them feel like they were not professional enough to deserve a poster.
The line in the sand
OK, here is the part where I do plant a flag.
I am fine with AI used as a polish layer. Filters on a real photo. Background cleanup on a real flyer. Color-correcting a real video. Touching up a real logo someone already drew. These are all the digital equivalent of editing a photograph in a darkroom, except faster.
I am not fine with AI used to manufacture the thing the customer is buying. That is the line.
The clearest example is restaurant food photography. There is a clean version of using AI here: a restaurant takes a real picture of a real plate of their real food, and AI helps with the lighting, the cleanup, the color cast from the bad overhead bulb. Great. The picture is still a picture of the food they serve. The customer is being shown a real promise.
The other version is: a restaurant generates an image of a steak that does not exist in their kitchen, off a stock prompt, and posts it on Instagram. That is not polish, that is a lie. The customer is being shown a promise the kitchen cannot deliver. When they walk in and order the steak, they will get a different thing. That breaks the contract a menu is supposed to keep.
Same logic everywhere else:
- Musician using AI for a flyer poster: fine. The flyer is the polish around the show. The show is real.
- Musician using AI to generate the live recording they are charging you for: not fine. The thing you are buying is the performance.
- Restaurant using AI to color-correct a plate of pad thai: fine. The plate exists.
- Restaurant using AI to invent a plate of pad thai: not fine. The plate does not exist.
- Agency using AI to generate a stock illustration for a blog header: fine. The illustration is decoration.
- Agency selling you a brand identity generated by AI in twelve seconds, and billing you for a brand identity: not fine. The thing you are buying is the brand.
The line is whether AI is the wrapping or the thing inside the wrapping. Wrapping, fine. Inside, not fine.
The other reason I am soft on AI for small businesses
One more thing, then I will let you go. The visual ecosystem on the internet has a charm problem already, and AI is not the only thing making it worse. The default Squarespace template is making it worse. The Canva-with-the-same-five-fonts post is making it worse. The agency-default brand system with a startup-blue palette is making it worse. We have been losing visual personality for a decade.
An AI-flashy logo is a new version of that same problem. It is smooth, it is generic, it has the same gradient as every other one, and you forget it the second you scroll past it.
Given the choice between a smooth AI logo and a homemade Comic Sans one, I will pick the Comic Sans every time. The homemade one has a human in it. It is honest about what it is. It tells me something true about the business. The AI logo tells me only that the business asked a tool for a logo. I wrote more about this argument over at substance over style, where the larger point is that the words are doing most of the work anyway.
The summary
Use AI for the polish layer. Do not use it to manufacture the thing the customer is buying. Be kind to small businesses figuring this out in real time on a budget of zero. And if you are choosing between Comic Sans and an AI gradient, pick Comic Sans. The future will thank you.