When Websites Design Themselves

Apple's 1984 Macintosh revolutionized graphic design—but that was nothing compared to the coming wave of websites that'll design themselves.
Image may contain Art Modern Art Clothing and Apparel
fungirlslim/iStock

Graphic design used to require physical work. To compose letterheads, business cards, brochures, magazines, books, and posters, you hunched over a desk or a light table. You cut and pasted paper or assembled metal type on a printing press. You processed 35mm film by hand, developing pictures in a darkroom with chemicals.

In 1984, Apple’s Macintosh arrived and changed everything. Layout software such as Aldus PageMaker and its successors enabled designers to make changes with a click. Graphic design transitioned from the workbench to the computer screen, in what we came to call the desktop publishing revolution. Design work moved from the laborious world of hands-on creativity to the freer but more abstract digital realm, where you can see the results of choices instantly—but each decision carries less weight, because you can undo it with a single command.

Today, we’re on the verge of another revolution, as artificial intelligence and machine learning turn the graphic design field on its head again. The vision is, to quote one project’s slogan, “websites that just make themselves.” Software will evaluate your text content, line of business, and imagery, and spit out finished pages without your having to lift a finger. These kinds of automated tools will arrive on the web first, but print design will change, too, as design-software makers inject machine learning into their layout tools and apps.

For all the noise about AI-driven graphic design, however, today’s reality lags stubbornly behind the grand vision. Many of the products now available will disappoint users expecting miraculous results from AI genies. That’s a letdown, for sure, but it also gives us some time to think about what kind of design work we want machines to do for us, and what roles we should be reserving for human beings.

The Grid promises to hand the design of your site over to an AI named Molly: “She’s quirky, but will never ghost you, never charge more, never miss a deadline, never cower to your demands for a bigger logo…Molly can apply a simple five-color palette to your site in more than 200,000 unique ways.”

One of the earliest entries into the artificial intelligence web design marketplace, The Grid has been promoting “AI websites that design themselves” since it launched a crowdfunding campaign in 2014. “Conceptually feels very next level, an obvious, natural progression just waiting to happen,” tweeted 37Signals/Basecamp founder Jason Fried when The Grid’s promotional video made the rounds of the web design community.

The Grid asked its “founding members” for $96 but then took years to deliver a product, and reviews have been decidedly mixed. If you watch the company’s video today, the next video that YouTube queues up for you will probably be “The Grid Sucks,” a rambling, hour-long complaint by early beta user DigitalDan. Molly might be an AI breakthrough, but at the moment, her role mostly involves generating color palettes and auto-cropping photos. Reddit is full of dissatisfied users who say that The Grid’s much-touted AI abilities produce look-alike websites that are difficult or impossible to customize. Other reviewers are similarly caustic (“shoddy and expensive,” says CMSWire).

The only way to reach the company is through a contact form, and its automated email replies—signed "Love, The Grid"—directed me to a demo video titled “The Five Minute Website on The Grid.” (The video itself clocks in at 56 minutes.) The software the demo walks through looks very similar to WordPress, Squarespace, Weebly, and Wix—content-management systems that took most of the code out of web design a good while ago.

Wix, another popular website builder, also offers an AI solution: Wix ADI (Artificial Design Intelligence). Nitzan Achsaf, head of Wix ADI, says it can create a website all by itself by using the content you provide to suggest “billions of beautiful design options.” You click the option you prefer and the program does all the reformatting. The company’s materials make it look quick and easy—but like an enhanced version of Wix, rather than a breakthrough tool for auto-generating websites.

Firedrop is yet another web design tool incorporating AI and ML. Firedrop’s AI takes the form of Sacha, a chatbot who walks you through the site-building process, asking you a series of questions and then offering suggestions and recommendations. Launched in March 2015 as a drag-and-drop website builder, Firedrop evolved into a design tool using AI at the end of 2015—around the same time that The Grid gained momentum. CEO Marc Crouch says Firedrop’s chatbot is intended to replicate the experience of working with a professional web designer.

These web design tools might offer assistance using artificial intelligence, machine learning, and algorithms, but on the whole, they still require hands-on use. You input data, see options and select what works, and choose from prescribed templates. That can be a great help, particularly for the novice designer or small business owner. According to Firedrop’s Crouch, small businesses are Firedrop’s target market. (A 2016 study by Clutch found that almost half of the small businesses surveyed did not have a website; the number dropped to 29 percent in a March 2017 followup.)

Template selectors and drag-and-drop interfaces have already lightened the burden on these businesses’ efforts to get online. The promise of new AI-based tools is not only to make life even easier for customers, but also to make sure that the resulting websites don’t feel like cookie-cutter lookalikes. Yet today, when these tools use algorithms and routines to generate a website from pre-formatted elements, the results too often still feel like they’ve been punched out of preset templates.

Doug Bartow, principal and design director at New York’s id29 studio, equates using templates in any capacity with going to the copy center. Templates work, he says, but the layouts become “fairly neutered” rather than distinct. Templates also make fine tuning difficult, according to Bartow, because you get what you get and for the most part can neither adjust nor customize the layout. Bartow does, however, appreciate the potential that the next wave of software might offer, be it for web or print design, and is open to using it—provided he can still do his best design work and be “original and distinctive.”

For professional designers and media makers, Adobe is the essential software toolmaker, and it has been making its own play in artificial intelligence and machine learning. Adobe CTO Abhay Parasnis says Adobe has collected its data and experience in the super high-end of the market over the years, and built the resulting expertise into Adobe Sensei—an artificial intelligence and machine-learning framework operating behind the scenes in Adobe’s tools. Currently, you can see it at work in the Face-Aware Liquify feature found in both Photoshop Creative Cloud and Photoshop Fix; the tool uses AI-driven face recognition to let users select and edit human faces in photos.

Courtesy of Adobe

Parasnis believes that these features and others on the horizon will empower both pro designers and non-designers by enabling “more productivity and creativity.” But today, Adobe’s offerings—like The Grid’s site builder—provide only a glimmer of what design’s AI revolution could look like. For a fuller view, you have to turn from commercially available tools to the research world.

DesignScape, a project out of the University of Toronto, has won a lot of attention since its YouTube debut in 2015 for the companionship it seems to offer non-designers. Unlike The Grid’s Molly or Firedrop’s Sacha, DesignScape doesn’t promise to do all the work for you, but it doesn’t leave you on your own, either. Instead, more like a teacher than an assistant, it nudges you toward alternate and better solutions. DesignScape’s developer, Peter O’Donovan, now works for Adobe, along with two other of the project’s co-authors, so its features could soon make their way into Adobe’s flagship products. As an Adobe spokesperson pointed out, Adobe Illustrator’s still-in-development Quick Layout, showcased at Adobe MAX 2016, automatically adapts to items you’ve placed in the composition—much like DesignScape.

As today’s rudimentary tools mature, AI experts see the field moving down DesignScape’s path. In the next desktop publishing revolution, users will step back from hands-on labor and let the software generate ideas and plans. In this world, design work will become more like curation and management. Our tools will propose designs, and we will decide what works. But how useful—and how good—will the resulting designs be?

Nina Stössinger, senior typeface designer at Frere-Jones Type, expects artificial intelligence to have an ability to “replicate patterns and logic” frequently used in design work. Some of the results might be predictable, and some might be layouts we could create on our own. If something like DesignScape could help get those warm-up and immediate ideas out of us faster, Stössinger sees that as a useful aid. But she also worries that such assistance could trap designers, preventing them from thinking outside the box or customizing their work.

AI and ML assistance won’t stop professional graphic designers from wanting and needing to “make aesthetic decisions about the retouching and the typography,” according to Paula Scher, who has been a partner in the New York office of Pentagram since 1991. Even if the future of graphic design software involves more curation than creation, Scher emphasizes that “fine tuning” will still fall on designers’ shoulders. But as the software becomes more widespread and more sophisticated, “entry level jobs may be lost,” says Scher.

Of course, professional graphic designers have fretted about being replaced since the first versions of PageMaker and Quark fired up. One veteran who doesn’t believe in worrying is John Maeda, currently the Head of Computational Design & Inclusion at Automattic, the parent company of WordPress. Maeda says he’s been preparing for this latest wave of change since the 1990s: “In each decade, I have looked to acquire the skills that can keep myself ahead of the machine. I know it to be an impossible task, but I also don't give up easily. I guess I’m more of a warrior than a worrier. And I’m excited about the challenges that are coming to design.”

If and when professional graphic designers hand over some of their responsibilities to machines—or get cut out of the creative process entirely—many might welcome the change, seeing it as a chance to step away from the computer, whether to work by hand or just take a break from the screen. If this second desktop revolution’s AI algorithms save human designers time and make more room in their lives for reflection and creativity, it will win cheers all around. But those who want the software to function exactly like a professional designer should be careful what they wish for. “If you can find a machine that thinks like me, don’t buy it,” says Paula Scher. “It would be incredibly annoying.”