AI can already produce clean, convincing letterforms. So where does that leave the type designer? For me, the opportunity lies in physical materials, stronger ideas and making work with a history that AI can't touch, at the moment anyway.
When perfect started looking generic
AI is learning at a ridiculous rate. It can, apparently take a handful of sample glyphs and turn them into the beginnings of a clean, coherent typeface.
I don't have access to that sort of software so haven't tested it myself, but I know it exists from chat within the industry.
It could be very helpful for all designers who may have created one weight but don't have the time or knowledge to create new weights.
However its a bigger problem when its taking several designers works and trying to create a blended font using machine learning—but lets put that on hold for one minute.
The uncomfortable paradox
That creates an uncomfortable paradox for type designers. We need font companies and marketplaces to help sell our work. Those companies built their catalogues using fonts created by individual designers.
For years, the arrangement was fairly simple: they sold the font, then the income was divided.
We hate to love you
Now those same catalogues can become training material—and have become training material. A company can learn from thousands of typefaces, generate its own alternatives and remove the original designer from the transaction.
Why continue paying someone a 50% royalty when software offers the possibility of keeping 100%?
It's a strange way to treat the people who gave the company a reason to exist—but that's where we are.
Can AI truly think like a human?
The technology can already reproduce clean, orderly letterforms. What it can't convincingly reproduce—at least not yet—are fonts that contain random details, like handmade fonts and the impulse behind a human making something in an unexpected way.
AI needs a lot of learning material
I might decide to take a rusty screwdriver and scratch an alphabet into a metal plate. Not because research suggested a market for rusty-screwdriver typography, but because the idea entered my head and refused to leave. Turns out us humans are a weird lot.
In 1997, British comedian and author Tony Hawks accepted a £100 drunken bet and hitchhiked around Ireland with a table-top fridge. The journey became a bestselling book and, later, a film.
That’s the useful comparison. AI can study the finished result and imitate its surface. What it doesn’t naturally have is the daft human impulse that starts the whole thing.
As a human—and I think I still qualify—that’s where I believe the opportunity lies. Not in trying to make cleaner, safer fonts than a machine, but in making work that could only have begun with a peculiar human decision.
A font can begin with almost anything
I don't always begin with a carefully planned alphabet or a gap in the market. Sometimes the material itself becomes the challenge.
Could I carve the letters from lino and print them several times to capture the uneven ink, changing pressure and small differences between impressions? That question led to Higgs Boson Blues.
Other fonts begin with a particular tool or technique. Choose Love was made using a print roller. Wild Irish Rose began with acrylic paint. Autobahn was created using spray paint and hand-cut stencils.
The material becomes part of the idea
Each method leaves different marks. Lino resists the blade. A roller distributes ink unevenly. Acrylic paint records the movement of the brush, while spray paint behaves differently around the edges of a stencil. I could imitate some of those effects digitally, but the material would no longer be helping to shape the letters.
Sometimes the starting point is a visual idea. What would happen if the letters looked as though paint were dripping down the paper? How far could I take that effect before the words became difficult to read?
Occasionally, the idea is much simpler: I'd like to make a horror font.
You don't necessarily know what the finished alphabet will look like when you begin. The material, the technique and the accidents encountered along the way all push it somewhere slightly unexpected.
AI can imitate the finished appearance. The harder thing to reproduce is the physical process—and the human decision to begin with lino, a print roller, acrylic paint or a can of spray paint in the first place.
Four fonts that began away from the screen
These aren't textures added after the letters were drawn. The materials helped create the letterforms from the beginning.
Higgs Boson Blues: embracing the mess
I love a good old-fashioned lino print. It's relatively easy to do. All you need is some lino, a cutting tool, ink and something to print onto.
Once you've carved the letters, you can take as many prints as you want. Give the lino more ink and you get one result. Deliberately starve it of ink and you get another. Change the pressure slightly and a texture appears that you couldn't have planned.
Marvellous.
Higgs Boson Blues came from embracing that mess. Instead of designing every rough edge on a screen, I let the printing process create them. Some impressions worked better than others, but that's where the interesting stuff happened.
I love happy accidents. With lino, you get plenty of them.
Choose Love: made with a pound-shop roller
Choose Love was created with cheap rollers from a pound shop. They came in a children's painting set. I bought several packs, then cut some of the rollers down to make different widths.
I rolled them directly onto paper to create the letterforms and pick up the marks left by the ink. It was fun and messy, and each roll came out slightly different.
For some letters, I cut paper masks to contain the ink and create perfectly straight edges. For others, I didn't bother. I just put the roller down and let it travel across the paper.
That gave me a useful mixture of control and chance. Some parts looked sharp and deliberate; others carried all the uneven texture, excess ink and stray marks produced by the roller. The material wasn't added as an effect afterwards. It was there from the first mark.
Wild Irish Rose: acrylic on acetate
I chose acrylic paint for Wild Irish Rose because it dries quickly. You can apply it thick or thin, and it works well on transparency film.
I printed a baseline onto paper and placed it underneath the acetate as a rough guide. Then I painted the letters with a medium-sized brush. The baseline gave me enough control to keep everything heading in roughly the same direction without making it feel too rigid.
Once the paint had dried, I scanned the letters and started cleaning them up. That part takes forever, but it's worth it. Acrylic holds the movement of the brush, the changes in pressure and the thicker patches of paint. Those are the details I wanted the finished font to retain.
Autobahn: embracing my inner Banksy
Cutting stencils by hand takes ages. It also hurts your hands and fingers. That wasn't the interesting part for me.
The spray paint was the cool part.
I embraced my inner Banksy, put the stencils down and hit them with the spray can. I sprayed each one several times and got a different result with every pass. Some were heavier, some had softer edges and others caught more overspray.
Love it.
That variation became part of Autobahn. I wasn't trying to make one flawless master version of every letter. I wanted the spray paint to do what spray paint does and give me results I wouldn't get by carefully drawing the same effect on a screen.
Most people don't care. Brands have to.
Here's another paradox. Blimey, there are a lot of those today.
The average person probably doesn't care whether a font was painted by hand, drawn by AI or discovered inside an old filing cabinet. They see something they like and think, “That's neat. I want it.”
As a type designer, that's slightly annoying. You can spend weeks experimenting with paint, lino or hand-cut stencils, only for the customer to judge the result in roughly the time it takes me to eat a biscuit.
Bigger brands and astute designers have a different problem. They need to know where a typeface came from and whether they can use it safely.
The risk no major brand needs
Imagine launching a worldwide campaign with an AI-generated font, only to face accusations that it closely copies somebody else's work. Whether the resemblance was deliberate or produced by the software almost becomes irrelevant. The campaign is already public, the money has been spent and the brand is now answering questions it never wanted anyone to ask.
No major brand needs that risk. No design agency worth its salt wants to explain why its supposedly original campaign uses lettering that looks suspiciously like somebody else's typeface.
A handmade font has a traceable origin. There are sketches, paintings, lino prints, stencils and physical marks connecting the finished typeface to the person who made it. That doesn't automatically make the font good, but it does give the work a history that can be shown and explained.
For an individual buyer, that story may not matter at all. They like the font, buy it and get on with their day. For a designer or brand putting serious money behind a campaign, knowing who made the typeface—and how—could make it the safer choice.
So where does that leave us type designers?
Poorer, mostly.
Jokes aside, my instinct is to double down on physical materials and keep asking one question: could AI do this?
That doesn't mean adding a rough texture to something drawn on a screen and declaring it handmade. It means choosing materials that influence the letters: cutting lino, rolling ink, painting on acetate, making stencils or scratching into a surface with whatever happens to be nearby.
Start with a better idea
More importantly, AI makes me want to begin with a better idea. That sounds obvious, but finding an original idea in an oversaturated font market isn't easy. There are already thousands of script fonts, display fonts, horror fonts and hand-painted alphabets. Using a different tool isn't enough if the idea behind it feels familiar.
Document everything
I also need to document the process properly. That's something I've been terrible at in the past.
Documenting a font means keeping the early sketches instead of binning them, photographing the materials, recording video and showing the experiments that failed. Then there's editing the footage, writing about it and turning everything into social content. It takes time away from making the font itself, which is probably why I've avoided doing it.
But if the process is part of what separates handmade type from generated type, hiding that process no longer makes much sense.
For my next projects, I want the record of how I made the font to become part of the project itself—not something I try to reconstruct afterwards from three blurry photographs and a vague memory.
AI will catch up eventually
It would be useful if the Western world decided that AI-generated content had to reveal where it came from and what helped create it. At least then we could see the trail leading back to the original work.
The race nobody wants to slow down
The problem is that AI isn't being developed under one set of rules. China, Russia, America and everyone else are building their own models while trying not to fall behind. If America slows down and China keeps going, the fear is that China ends up controlling a much larger part of what we see and use online.
That's fucking scary.
It also makes meaningful restraint difficult. Nobody wants to limit their own technology while another country races ahead without the same restrictions. Commercial pressure pushes it forward from one side; political pressure shoves it from the other.
So it's probably only a matter of time before AI can reproduce the physical qualities I'm talking about. Paint, damaged ink, roughly cut lino, overspray from a stencil—it will study enough examples to generate convincing versions of all of them.
I don't think type designers can outrun it. But while I've got my trainers on, I might as well jog alongside the AI gods and admire their shoes, which look suspiciously like something they copied from Adidas and Nike.
For now, I'll keep making things with my hands, starting with better ideas and recording how I made them. AI will catch up eventually. I'd still rather give it something worth chasing.
Written by Simon Stratford, an independent type designer and graphic designer based in London. These are my opinions, reached after too little sleep and too much coffee.
You can also follow my antics on @itsmesimonok, or leave a comment below. Despite all the opinions, I'm actually quite nice.