Findings from an ATypI Panel
Ann Bessemans & Sofie Beier, September 2026
Ask four type experts where Calibri belongs in the Vox–ATypI classification system and you will get four different answers. Thomas Phinney tried it. His own answer didn’t match the one given by Calibri’s designer, Lucas de Groot.
That disagreement raises a question about the purpose of classification. What was this professional vocabulary designed to do, and for whom?
In 2021, ATypI de-adopted the Vox–ATypI typeface classification system it had endorsed since 1962 because it was heavily Eurocentric. It covered a narrow slice of the Latin typographic tradition, leaving most of the world’s typefaces and writing systems unrepresented. Some styles were elaborately subdivided; others were absent altogether. With that standard withdrawn, we need to consider what a classification system should be for, who should build it, and whether a single shared standard is the right goal.
This article reports on the first in a series of panel events organised by ATypI’s Education and Research Working Group, held at ATypI Stanford in May 2026. Educators and industry practitioners took part in two sessions: one on how classification is taught, the other on how the industry organises type and helps users find it. We asked how the field currently categorises typefaces and whether teaching aligns with the way type is organised for sale and search.
No new classification system emerged from the discussion. The participants described different needs that any proposed system would have to address.
Same word, different jobs
Teachers use classification to help students understand letterforms and their history. Font platforms use it to help users find typefaces.
In the classroom, students learn why letterforms look the way they do, which tools produced them, and what historical conditions made each style possible. They also learn what a designer was trying to achieve or reacting against. A student learning what a Garalde is encounters the humanist scribe and the broad-nib pen, including the diagonal stress that the tool produces. The category connects the form of the letters to how and when they were made.
On a commercial platform, users may have no vocabulary beyond serif and sans-serif, or may be searching by feel rather than form. Categories help them filter a library to find a suitable typeface or discover one they had not considered. Knowing why Garamond has diagonal stress may offer little help with a request such as “I need something warm and legible for long-form editorial.”
What the platforms see
Tom Rickner, senior director of Monotype, presented twelve months of filtering data from the Monotype Fonts platform. Roughly half of all visits involving active search were what Rickner called “intent-led”: users typed in a font name they already knew. In the remaining, discovery-led sessions, users rarely selected formal classification subcategories. The most searched subcategory, Slab Serif, accounted for fourteen hundredths of a percent of filter clicks. Grotesque, Humanist, Didone, Transitional and Old Style were used very little, despite their place in typographic curricula and style manuals.
An audience member questioned whether survivorship bias affected the findings: the data might describe those who managed to use the existing system while missing those who left without finding a font. The interface could also affect which filters users selected. Someone unfamiliar with “Grotesque” might choose “Sans-Serif” and stop there if no image or definition explained the narrower category. Rickner acknowledged the point. When another audience member asked whether subcategories had visual examples, he said they did not. “Are we as the industry teaching our users?” Rickner said. “Not in the best ways, no.” Low use of the subcategories may partly reflect how they are presented.
Users selected tags in roughly equal measure to formal categories. Rickner’s platform includes 61 manually selected tags drawn directly from search terms, including “retro”, “whimsical”, “editorial” and “tattoo”. Dave Crossland, programme manager at Google Fonts, described a similar process. UX research conducted around 2011 by Dawn Shaikh informed the original five Google Fonts categories (Serif, Sans Serif, Display, Monospace, Handwriting), based on what users selected. A more recent expansion, led by psychologist Hillary Palmen, added expressive and mood-based categories drawn from user sorting studies to the existing categories based on letterform structure. Both platforms used observations of users to develop their categories.
Rickner commented: “I think this is going to change over time as we add more tags that we can train AI to properly categorize consistently, and frankly, as fewer and fewer people are taught these categorization systems to begin with.”
Rickner’s comment raises a concern for educators: users may encounter less of the vocabulary taught in type design courses when they search for fonts.
The limits of tags
Tags can help users search by mood. Several people questioned how reliably those terms could be used across languages and contexts.
One audience member described the problem: “I personally hate tags because I find that ‘retro’ and ‘modern’ mean the same thing depending on your context. You cannot translate a tag. A tag can’t translate from a French font to an English font to an Indian font. It’s completely unscalable as a system, and it’s so biased.” Tags drawn from English-language searches reflect the associations of those users. Unless developers account for other associations, they risk making fonts harder to find for people who describe them differently.
Another audience member observed that tagging typefaces “Valentine” and “Halloween” risks treating type like clip art. If a tag repeatedly directs users to the same few typefaces for an occasion, much of a catalogue containing tens of thousands of fonts may go unseen.
Gabbi Soong, whose platform Arcotype uses visual similarity to help users discover fonts, said: “We’ve all kind of been hopping around the fact that words are limiting our font exploration.” She questioned how well keyword search, increasingly mediated by AI, helps people discover fonts when they do not yet know what to ask for. Arcotype arranges fonts spatially by visual similarity so that designers can explore a map. Rickner described a customer who missed browsing the Monotype catalogue as he had once browsed a printed specimen book: moving through it methodically, marking where he stopped, and returning the next day to continue. The digital platform did not support that way of browsing. Choosing a classification system also involves deciding how people will be able to explore a collection.
Description, not classification
Thomas Phinney, type designer and forensic font consultant, questioned whether fixed categories could accommodate the ways designers combine letterforms. The traditional academic model, he argued, organises fonts into branches like a biological taxonomy. He explained where the comparison fails: “Your biological tree of life would be pretty useless if you could easily have a creature that was literally one-third frog, one-third monkey, and one-third oak tree. With fonts, a designer can do exactly that. That means any classification system based on a tree structure and rigid categories is fundamentally doomed to failure.”
Typefaces designed in the last half-century often draw on several historical traditions. A designer might combine humanist features with geometric construction, or Didone contrast with contemporary proportions and spacing. A fixed category can obscure those choices. Phinney favours describing attributes along independent axes instead of assigning typefaces to nested categories. Formality, contrast, weight and rhythm can each vary across a range. Such a system could describe new combinations without requiring a new category for each.
Phinney noted that his legal work requires this kind of description. When font identity is at stake in court, he needs to identify specific, verifiable features that distinguish one font from another and can be examined in cross-examination.
What classification does in the classroom
The education panellists described how formal classification helps students notice and discuss differences between letterforms.
Borys Kosmynka, who teaches type history and design, values the historical and contextual understanding students gain through classification. When students understand the conditions in which a typeface style developed, they can discuss it more precisely with peers and clients, and judge their own choices. Whether they use Vox categories or describe the forms directly, they gain terms for explaining what letters communicate. Maurice Meilleur argued that without this vocabulary, students must rely on subjective feeling, which is difficult to defend when a choice is challenged.
Yves Peters teaches the history and function of type alongside one another. He begins with letterform analysis, asking how a shape is constructed, where its contrast comes from and which tool produced it. He then introduces category names as terms students can search for. One class is devoted to emotional response: what does a typeface feel like, and why? He wants students to be able to discuss type with both peers and clients, and to judge which terms will help in each conversation.
Milda Kuraitytė argued that students also need to learn how to build and critique classification systems. Examining what a system includes and the assumptions on which it rests gives them practice in questioning categories beyond typography.
Meilleur reported that systematic typography teaching is becoming less common in American design education, except in the most research-focused programmes. In his experience, the profession’s growing practical demands leave less room for it. AI tools increasingly answer classification questions on demand, while students have fewer opportunities to learn how to formulate those questions themselves. “I want more room to have critical, morphological, and historical conversations with my students”, Meilleur said.
Connecting type education and industry
The difference between classroom and platform vocabulary leaves some users needing help to apply what they have learned.
Crossland described how Google Fonts came to produce basic typographic teaching resources. As the platform published more typefaces, especially those using new font technologies, staff found that many users lacked the knowledge to use them effectively. Google Fonts responded with educational content to help users search more effectively. An audience member recalled learning about typography by asking questions and consulting books and online forums. He argued that these routes into the subject were no longer reliably available, and that industry needed to help provide resources for students beginning to take an interest in type.
David Berlow, co-founder of Font Bureau and a central figure in variable font development, asked educators to expose students to more designs and their uses. He described tags as shorthand for examples of work already made. Designers familiar with more of that work have less need for such shorthand. Education can give them the knowledge to decide what to search for.
The AFII precedent
When Rickner asked who had heard of the AFII classification system, almost no one in the audience had. The Association for Font Information Interchange developed it in the early 1990s with the aim of covering type beyond the Latin script. It included non-Latin correlations to standard Latin categories, explicit Japanese subcategories, and Arabic equivalents to Neo-Grotesque. It appeared as an appendix in Peter Karow’s 1994 book Font Technology: Methods and Tools, but did not gain widespread use.
The reasons for AFII’s limited adoption are not fully documented. Its history gives anyone proposing a new system an earlier attempt to examine, including how it addressed writing systems beyond Latin and why so few people came to use it.
What ATypI could do next
The Stanford discussion gives us reason to question whether replacing Vox with another single standard would meet the needs described by the panellists. Teachers and platforms use classification for different purposes, and even within each setting, users need different kinds of information.
ATypI could help connect the vocabulary used in education with the terms used on commercial platforms. A graduate who knows Garamond as a Garalde may encounter “classic”, “editorial” and “timeless” when searching for it. Knowing how these descriptions relate, and where they differ, takes work for which the graduate may receive little support.
ATypI could document the purposes of existing systems, the settings in which they work and the difficulties users encounter. This would give teachers, designers and platform developers a basis for choosing a system suited to their needs without requiring ATypI to mandate one.
Examining classification beyond Latin should also help test the assumptions behind those systems. An audience member used a screwdriver-and-hammer analogy: “classification is a tool, not a rule; use the right tool for the job”.
The ATypI Classification Workshop is a multi-session programme organised by ATypI’s Education and Research Working Group. At ATypI Stanford in May 2026, Ann Bessemans and Sofie Beier moderated the afternoon sessions with educators Milda Kuraitytė, Yves Peters, Maurice Meilleur, Borys Kosmynka and Thomas Phinney, and industry practitioners Dave Crossland, Gabbi Soong, Tom Rickner and David Berlow.