AI Has Come for Serif Fonts

As public backlash to the seeming omnipresence of artificial intelligence intensifies, the collective quest to weed out—and reject—telltale signs of its use continues.

One of the first casualties, to my dismay, was em dashes—which are a great, and very human form of punctuation, by the way! There's also the “rule of threes,” which is meant to scan as rhythmic, but often comes across predictable, hackish, and stale. And, of course, there are the clunky grammatical constructions of the “not X, but Y” variety.

Now certain fonts and typefaces—specifically serifs—seem to be defining (and giving away) AI, both in actual software, and in vibe-coded design boilerplates. Some are calling it “tasteslop,” the results of the effort to make generative AI designs seem superficially sophisticated or distinguished.

The shift away from slicker, more conspicuously computerized typefaces is something the San Francisco Bay Area writer, designer, and type practitioner Keya Vadgama has termed “the serif renaissance.” In a recent newsletter, published on her Substack, Vadgama suggests the move is a bid for companies to project more “personality and warmth.”

“It’s not that difficult to discern why AI-native companies in particular are being drawn to serif fonts: AI is inherently cold and without opinion,” she writes. “[Using serifs] signals ‘We’re AI! But real humans use (and made) our product! We swear!’”

“Serifs have an origin in calligraphy,” Vadgama tells WIRED. “It connotes a very human, fluid way of making letterforms.” Vadgama has noticed that Anthropic’s Claude was defaulting to serifs. Other AI companies—Runway, Perplexity, Manus—had also adopted similar typefaces in their UX and branding.

Reached for comment, Perplexity chief communications officer Jesse Dwyer tells WIRED: “Why wouldn’t we have human design? Perplexity is for people.”

Vadgama believes the use of serifs is as much about aesthetics as building confidence between users and brands. Certain font choices signal, even at some preconscious psychological level, trust. Sans serifs (your Arials, Calibiris, Helviticas) are too clean, too computer-y. Good old Times New Roman, and similar typographic designs, can feel a bit more dignified. Recently, Vadgama was doing some branding work with a (since-shuttered) AI startup, which favored the serif text. “A big part of it,” she says, “is, ‘How do we position ourselves in a way that people are not afraid of us?’”

Serifs can help build that conviction, or at least the illusion of it. Times New Roman itself was commissioned in the 1930s by Britain’s Times newspaper. The typeface carries a certain authoritative heft. Books and newspapers are printed using it. It was all but standardized in the decades before screen reading. Perhaps most famously, the Encyclopedia Brittanica—arguably the authoritative compendium of human knowledge, at least pre-World Wide Web—was set in Times.

“In the broad public, a serif carries connotations of scholarship,” says Ali S. Qadeer, chair of graphic design at the Ontario College of Art and Design in Toronto. “Claude is interesting. It’s using this slightly brown background to mirror a book page. It’s sort of emulating the feeling of reading print. And print has deeper associations with trust.”

As reported by The New York Times, even the US State Department has returned to using Times New Roman after Secretary of State Marco Rubio decried Calibri as “informal,” pegging the department’s adoption of the sans serif typeface on some wider, Biden-era DEI initiative.

Both Qadeer and Vadgama see the trend toward serifs as a rejoinder to AI’s perceived (and, indeed, literal) lack of soul, and the wider public suspicion of the technology. They’re not the only ones. Alongside the “tasteslop” discourse, people online have criticized the serification of AI aesthetics as “generic” and “very ugly.”

“If half the internet starts vibe coding ‘unique and interesting’ fonts without understanding the basics of design, those fonts will become associated with slop,” wrote one X user in response to a post about AI web design.

Human Touch

Flip back just a few decades and computerized design seemed harsh and ugly: slime-green fonts glowing fluorescently from clunky computer terminals. Think of those old IBM terminal typefaces. The eyes practically strain just remembering them. For Qadeer, who teaches the history of typography, this change is a part of a deliberate rebranding effort. “I 100 percent believe it’s an effort to soften people up,” he says. “It is a response to large-scale social criticism. The sterile look of tech that has dominated for the past 20 years has increasingly negative connotations.”

Not everyone is so despairing. While designer and founder Yitong Zhang described the transition to serifs as “cursed” on X, he doesn’t think it portends anything especially sneaky or sinister. “Somebody at these labs is trying to get these models to be good at design,” he says.“It’s pretty pragmatic.”

Zhang compares AI, in its present state of aesthetic maturation, to a teenager downloading and experimenting with different fonts. (Which computer-native of a certain age didn’t download Iron Maiden or South Park TTF font packs to fiddle around with?)

He likens this emergent style to “premium mediocre,” an idea coined by the blogger Venkatesh Rao, to describe a kind of popular, faux luxury. (“Premium mediocre," Rao writes, “is the finest bottle of wine at Olive Garden.”) Of course, AI design choices are to a great extent self-replicating, as models train on outputs created by other models, serifs beget more serifs.

Good ole Claude itself pretty much confirms much of this. Prompted to account for its shift to serif typefaces, the chatbot cites issues of trust, authority, and “literary seriousness,” while also accounting for “heavy borrowing” and networked herd mentality. It also seems to acknowledge that slick, sophisticated, tastesloppy design choices operate as a kind of feint: “The clean lines, the fluid animations, the assured typography all communicate ‘This system knows what it's doing.’ The aesthetic actively works against accurate mental models of what AI is.”

Vadgama agrees. “I do think there’s something a little dishonest in trying to use serifs to signal ‘We’re not one of those scary AI companies,’ when in reality, you still are,” she says. “You can use Comic Sans, if you wanted. It doesn’t stop you from still being an AI company.”