Features

Can AI design a typeface?

Re:Vision, Monotype’s latest Type Trends report, explores typography's role in cultural conversations. Transform editor Jack Cousins sits down with Charles Nix, senior executive creative director at Monotype, to discuss the report’s findings on the complicated relationship between typography and AI.

Tell me about the 2025 Type Trends report and what this year’s edition does differently.

Charles: Trends reports, and type trends reports specifically, have become commonplace and very formulaic. There are a lot of them, and they all follow the same basic idea: you collect a lot of interesting typographic work from the past year, analyse it, find the through lines and then group similar typefaces and type treatments together. You then give each group a cool name and present them.

We've done that for probably half a decade now, but this year we decided to do a dramatic pivot. My co-authors, Phil Garnham and Tom Foley, and I have grown wary of the superficial nature of type trends reports. There’s a danger of courting typographic sameness or stylistic sameness that is inherent in that process of presenting 'type top 10' lists or 'best of' lists. So, instead, we wanted to explore the underlying forces shaping typography – and being shaped by typography – as a broader cultural force. We identified a whole host of themes, but then honed in on six: multimedia, AI, law and order, ageing, conflict and peace, and the environment.

Focusing on that AI theme you mentioned, we have started to see a push towards using AI to design typography over the past few years. But how commonplace really is AI-designed type in 2025?

Charles: I've been a typographer since I was a little boy, so I get hyper-specific about the terminology around typography. There's a fundamental difference between what is a font and what is a piece of lettering. Logotypes, for instance, could generously be lumped under the umbrella category of typography, but they're actually pieces of lettering. A lot of what we're seeing in AI typography right now is, technically, lettering – letters created for a specific purpose, or letters created by an AI that can't be taken apart and recombined. We could, after the fact, take apart what the AI has produced and turn it into typefaces, but strictly speaking, it's creating static presentations of letters.

All of that specificity aside, some of the letterforms being created are truly spectacular. It is possible, and we've seen it very recently, for people to take AI lettering and turn it into a font, but that's still a really niche area. It's not widespread in the brand ecosphere; it's something that's happening more in the experimental corners of the design world, which means we don’t see it much in commercial use. Instead, what we see are a lot of remarkable, clickable and sometimes terrible things on social media. I think the lack of commercial application stems really from a range of factors.

What are some of those pitfalls?

The first is the issue of ownership and indemnity. Anything directly made with a large language model (LLM) isn’t strictly ownable – it’s not created from a human and it’s created from many inputs, which means there’s a general sense that nobody owns anything that comes out of LLMs.

There's a second problem, that's purely technical, which is resolution and scalability. The quality isn’t there yet to reliably take something generated by GenAI and scale it up for use across commercial touchpoints.

Third, there's the inability to iterate. While some of the latest LLMs are starting to show iterative capabilities, they’re essentially redrawing designs from scratch rather than refining a previous version.

And finally, there's the problem that AI is unpredictable and occasionally hallucinates. I personally love that weird side of generative AI, but it also makes it completely unusable because those hallucinated letterforms don’t make for clear communication.

[Image credit: Monotype and Robert Connelly]

Co-writing the report brought up a lot of tricky questions, such as ‘What is it that we do that a machine will never be able to?’ In the context of type design, what do you think is the answer to that?

Charles: Well, I think generative AI will become a tool for making typefaces. It sits between two crucial, distinctly human processes: the prompt and the product. Designers, by our very nature, are prompt engines. We've always been tasked with framing and reframing problems over and over again, asking whether this problem we’re being asked to solve is the right one. When faced with the infinite possibilities of what to ask AI, I think designers are uniquely well-positioned to ask better, more interesting questions. Crafting the right prompt is already central to what we do, and that will continue to be our specialty and differentiator.

We also are output specialists, and if you look at the general studio models over the last 100 years – which run from entry-level designers to VPs of creative – as you move up that ladder, you become more and more a curator of output. The more experience you have, the better you become at judging the efficacy, beauty and quality of what is made. I think that still remains one of the designer’s superpowers: the ability to look at something and say, “Yes, that is new and interesting.” I think prompting and curating will remain essentially human, and vital to the relationship between design and AI.

What are some of your favourite examples of those AI letterforms you’ve come across?

Some of the letterforms being created are truly spectacular. They do exactly what you want: they're engaging, unique and a real testimony to the power of letters to communicate via their form. I'm thinking here of Ale Paul's experiments with generative AI that have this incredible rendered surface to them. They stick in my mind because they’re three-dimensional and shiny, almost with a Jeff Koons quality to them.

There's something about it that's really cool, but I think I'm still most drawn to the hallucinations from the early days of AI. The first couple of months of Midjourney – around fall of 2022 – when it really had no concept of what type was. I found those results really shocking and unsettling because pieces of letterforms would be grafted onto other ones in odd, nonsensical places. You would be left with something that was symbol-like, but incapable of communicating beyond its form. And yet, in that moment, you could sense right then that something had fundamentally changed.

How prepared are Monotype for this brave new world?

Charles: Well, it's moving so fast that I don't think any of us can be quite prepared for the pace at which the models are learning. That said, Monotype is very much in the mix and creating AI solutions to typographic problems along three really fundamental lines.

The first is, of course, the generation of type using AI. We've been experimenting with that for some time and have actually used machine learning to analyse typefaces long before 2022. By 2016, we had the sort of underlying technology that informs our 'WhatTheFont' font identification service.

The second is an outgrowth of that, which is using the ability of an LLM or machine learning to understand typographic relationships. So, what makes Bodoni not Didot, but adjacent to Didot? And what makes Helvetica off in its own corner compared to those two? That may sound like an interesting conceptual exercise, but being able to do that then allows you to search for and discover typefaces more effectively.

And then that third thing is the intercession moment, like finding ways to insert high-quality typography into generative AI processes.

Those are the ways we're approaching it at Monotype. But personally, I'm incredibly excited about the prospect that a brand-new typographic technology is emerging, and that Monotype is embracing this opportunity to innovate within our industry. I understand the trepidation many feel at the power of it. But these revolutions are fundamental to the way that type is made and used. They are exciting because they involve people coming to grips with the idea that there's a better way to do something. Change can be challenging, but we should face that challenge, head-on and with intention.