I’ve written dozens of stories about Adobe over the years, and not to brag, but I’ve never needed an AI to edit any of them.
But today, the company is debuting its latest Sneak, or a tease of a new tool to come. It’s called Project Catchy Content, and it’s an AI that can analyze online content, from the photography to the written tone, and tell you if people will engage with it or not—and why.
Project Catchy Content promises to analyze blog posts and store listings, suggesting everything from better colors to tweaked copywriting, in order to get the best possible response from your audience. Put differently, Adobe isn’t content with helping you make your website pretty anymore. It wants to help ensure that the design is effective too.
As Steve Hammond, a VP of Adobe Experience Cloud who leads the Sneaks program, explains, the work stems from years of Adobe’s AI research. The company has already developed some powerful AI-based tools for creatives, such as Content-Aware Fill, which uses AI to analyze a scene and fill in plausible objects such as grass or water over a blemish that you’d like to cover up.
“That ability gives Photoshop an understanding between pixels, colors, and patterns,” says Hammond. But he explains that Catchy Content pushes Adobe’s image prowess farther, from editing to a deep evaluation. The AI classifies photos with all sorts of keywords (such as “swimming woman”), deconstructs their color palettes, and analyzes the accompanying text. It can then cross-reference all of this information against what people engage with—highly specific, demographic data—to develop a scorecard for your content.
The system is one giant AI analyzer that correlates what’s on your page to what people read, click, or buy.
To demonstrate the idea, the Adobe team ran me through a demo of a site they call WKND. WKND is built in Adobe Experience Manager, the company’s content-publishing software, which is used by a small slice of the internet, including big commercial clients such as Chase, Salesforce, and FedEx. (But Adobe says there’s no reason this AI couldn’t operate through the company’s entire Creative Cloud ecosystem). WKND has a pretty typical modern look, but with a button press, you can see what the AI sees. It’s code that labels the amount of “joy” or “anger” in the copy and adds tags such as “beach” or the “beautiful” based on qualities it finds in the scene.
The user doesn’t need to parse all this code, however, because Adobe provides an easy-to-read dashboard. The dashboard pulls the color palettes from your page and flags those that attract your audience and those that don’t. A list of Adobe Stock images that might work better is suggested. It also charts your tone and the reading level of the words. One example I see actually scores that content, on a 1-100 scale, in terms of its affinity for your intended audience. (Adobe’s own internal example only scored a “fair” 53 with “female surfers,” who are the target demographic.)
Exactly what photos and text are considered good depends entirely on your audience’s preferences and what you’re trying to get them to do. Adobe imagines this platform less for journalism, and more for corporate blogs, B2B sites, or e-commerce. And it can be tuned to different needs, such as increasing engagement times or increasing sales.
There are no two ways about it: This is a cold and calculating way to view creative work. But it’s also an analytic tool that reaches a lot deeper than existing options such as Google Analytics or Parse.ly, which can often track how well a piece of content on the internet is doing but can offer very little in terms of actionable advice to improve it.
“The content author can make a more informed decision about what’s effective for this audience,” says Hyman Chung, the product manager at Adobe who developed Catchy Content. The entire system, viewed from 1,000 feet away, reminds me a lot of how Hollywood has traditionally focus-tested films, showing them to an intended audience, getting feedback, and tweaking scenes before a nationwide release. Except, in this case, Adobe is automating the process with trends and big data, before quantifying far more granular decisions that you made in content and interface.
Now, allow me to disclose, I have no idea if Catchy Content actually works. I did not test-drive it. The demo I received didn’t prove itself over months of page views or product purchases. Most Adobe Sneaks will drop your jaw instantly because they are visual stunts that anyone can recognize as incredible. Catchy Content is an interesting idea, but I’m more than a little skeptical that it would be refined enough to tangibly improve engagement on my stories day-to-day—although journalists are admittedly not its target market.
My other concern is, assuming Catchy Content does work, is it possible that the system figured out a single successful formula that would make all websites look more or less the same, rather than dozens of formulas that would work in different contexts? That’s something Adobe is wondering too.
“What we’re looking at here does show a few examples to your point, and a lot starts to look similar. That will be true as long as brands have similar content,” says Hammond. But as brands and writers create more types of stories inside Catchy Content, Hammond imagines we’ll see optimal designs diversify.
In other words, Adobe is making a typical, and halfway convincing argument, which many AI developers make, saying they just need to feed the machine more data to make it smarter. But Chung also points to some big differences within beta testing between different audiences in the New Zealand and Australian markets, where customers seem to be responding to different, regional preferences. So in other words, more differing content, and more differing audiences, could reveal preferences that Adobe hasn’t considered yet.
As of today, Catchy Content is being tested in beta with a few of Adobe’s customers, including “large household names in consumer packaged goods and travel/hospitality,” along with a nonprofit and an enterprise tech company. As with any Sneak, there is no guarantee that Catchy Content will come to market, and if/when it does, the tool might look completely different. Hyman himself seems to be mulling whether Catchy Content is giving too much feedback and if they might pare back suggestions to only the most actionable items.
So I have no clue if Catchy Content will have much, if anything, to do with the future of publishing and e-commerce. That said, I must admit: Its purported powers are enticing. If Adobe doesn’t figure this one out, inevitably, some other company will.