How AI Hurts Graphic Design — and How It Helps
·Caitlyn York
I have a folder on my computer of AI-generated logos clients have sent me over the last year with some version of the message “could you just clean this up?”
I love that they are excited. I really do. It is a lot more fun to start a conversation with a client who has been playing with ideas than one who says “you’re the designer, surprise me.” But I want to be honest about what I see in that folder, and then equally honest about the parts of my job this stuff has genuinely made better. Because it is both, and anybody telling you it is only one is selling something.
Start with the hurt, because it is the part nobody says out loud
It produces images, not identities
A logo is not a picture. A logo is a thing that has to work as a 24-pixel favicon, embroidered on a polo, cut in vinyl on a truck door, in one color on an invoice, and reversed out of a dark background on a website. It has to still be legible when the sign guy scales it and the print shop converts it.
What comes out of an image generator is a picture of a logo. It usually has soft gradients, a stray fifth color, letterforms that are almost but not quite a real typeface, and detail that vanishes below about 200 pixels. It looks great in the square it was generated in. It falls apart the second it has to live in the world.
I cannot “clean that up.” I can look at it, understand what you liked about it, and build the real thing. That is a different job and I am always glad to do it — but it is worth knowing that is what is happening.
It is confidently average
These models work by learning what things generally look like. That means the output is, structurally, the average of everything that already exists in that category. Ask for a logo for a landscaping company and you will get a leaf, a swoosh, and a green-to-darker-green gradient, because that is what landscaping logos look like.
Which is the exact opposite of the assignment. The whole point of a brand is to not look like your competitors. A tool whose core competency is producing the most typical version of a thing is structurally bad at differentiation.
It kills the thinking that makes design work
The valuable part of a design project is rarely the drawing. It is the conversation where a business owner tells me their customers are mostly retirees who moved here from up north, and that changes everything about the typography. It is realizing halfway through that the thing they are proudest of is not on their website anywhere.
When a client arrives with a generated logo, that conversation is at risk of not happening. We skip to evaluating an artifact instead of understanding a business. I have to work to drag it back, and sometimes I lose.
The legal and originality problem is real
I do not know what a generated mark is derived from, and neither does the client. That is a genuinely uncomfortable place to be when you are about to put something on a building. This is a big part of why our logo work is done by a human being from a blank page — not as a marketing position, but because I can tell you where every line came from.
Now the help, which is bigger than I expected
I was a skeptic. I am much less of one, but about completely different parts of the job than I would have guessed.
It ended the blank page
The worst hour of any project is the first one. Now I can generate twenty terrible directions in ten minutes, look at them, and know immediately which two are interesting and why. Almost none of that survives. It does not need to. Its job is to get my brain moving, and it is very good at that.
Mood boards and reference in minutes
Finding visual reference used to eat real hours. Now I can rough out three completely different directions — the warm one, the technical one, the playful one — and put them in front of a client in the same meeting where they described what they wanted. Watching somebody point at one and say “that, but not the blue” is worth an entire round of revisions.
Critically: these are conversation pieces, and I say so out loud. They never become the deliverable.
The tedious production work
This is the quiet one. Masking, background removal, cleaning up a client photo where somebody parked a truck in the shot, generating the eleven sizes every asset needs. That work used to eat whole afternoons and none of it was creative. Getting those hours back is the single largest change in my week.
Placeholder content that is not lorem ipsum
Designing with fake text has always been a lie you tell yourself. Real-ish copy in a layout tells you immediately that the headline is too long, that nobody will write four testimonials that length, that the card breaks with a two-line name. I catch layout problems in the design now instead of in the build.
Where I have landed
Here is my rule, and it has held up for about a year now.
I use it for the parts of design that are labor. I do not use it for the parts that are decisions.
Generating options is labor. Choosing one is a decision. Removing a background is labor. Deciding what the photograph should be of is a decision. Producing eleven sizes is labor. Deciding the whole thing should be simpler is a decision, and it is usually the decision that makes the work good.
The risk in our field is not that the machine takes design jobs. It is that designers stop making decisions because there is always another button that produces something acceptable. Acceptable is genuinely the enemy here. Nobody remembers a brand that was acceptable.
If you are a business owner reading this
Bring me the generated logo. Seriously. It tells me more about what you like than an hour of adjectives will.
Just do not put it on the truck yet. Let me ask you the boring questions first — who your customers are, where this is going to live, what you want people to feel in the two seconds before they read a word. Then I will build you something that survives contact with a vinyl cutter.
Chad has a version of this same argument about code that works but should not. Same disease, different department.
Come talk to us. Ask for Caitlyn — and bring the folder.
Ready to grow your business?
Let’s talk about your website, marketing, or next project. We don’t bite.