Article: Can Adobe Illustrator score an AI knockout?
In one corner, wearing the orange trunks, we have Adobe Illustrator's generative vector tool, still in beta but ready for a fight. In the other corner, in the crisp minimalist black trunks, the reigning heavyweight champion of AI imagery, the Sultan of Syntax, the Picasso of Prompts, putting the 'Art' in Artificial Intelligence, and sporting two thumbs on one hand, OpenAI's DALL-E 3.
Created with: DALL-E 3 & Adobe Photoshop
When Adobe announced vector based generative AI for Illustrator, my mind jumped to the possibilities. It felt promising, but as I shared in my earlier Daily Generations article, the results were mixed: while the vector lines were surprisingly well constructed—better than Illustrator's Image Trace, for sure—but compared to the richness of art from other generative AI tools, Illustrator's work felt a little flat.
You may remember my article testing if DALL-3 was a useful tool for icon creation. The answer being a resounding sort-of! Raster output being the big limitation. Curbing the use cases and making simple edits like color changes more difficult. However, if Illustrator could create the same set in vector lines with a similar amount of effort it could really find it's niche in the AI landscape.
Can Illustrator's unexpected left hook score a knockout? Will OpenAI's raw power and experience win out? And what timeline are we living in where Adobe is an underdog in imaging?
Before we even get to the icons, the user experience deserves a mention. DALL-E 3 operates through the familiar chatbox interface we all know, with some ability to ask for edits on specific areas. But specificity is always a bit of a gamble—exact colors are hard to nail down, and following instructions that include small but important details, frankly, not its strong suit.
Illustrator, on the other hand, comes loaded with settings and sliders, including the ability to specify a source style image. The ability to set the dimensions of the output is a win as well. Its biggest power move, the ability to define exact HEX values for the AI to work within—and, impressively, it mostly listens. For production level brand work, this puts Illustrator ahead at the end of round 1.
Created with: Adobe Illustrator
However, when the metaphorical punches start flying Illustrator falls behind. You'll see in the two sets of icons DALL-E's work is more appealing, it interprets the prompts better, and the overall set is a little more cohesive. Illustrator did complete the brief and in fairness stuck more closely to specific instructions. But its work is more basic, a little more "junior".
The "stack of stones" image in the top right of each is the strongest example of this difference. DALL-E's entry looks like a stack of rocks. Illustrator however gave us a column of colorful ovals.
When the final bell rings, and before the judges announce their decision, I have to acknowledge a few things. Strictly speaking these are not icons. They're mini-graphics. I wouldn't use them in an interface or a menu for example. Best used to illustrate blocks of text and to draw the eye. They get called icons a lot, but are inappropriate for navigational use.
We're creating mini-graphics instead of true icons because creating a set of interface or design system icons in DALL-E, as it currently stands, would be practically impossible. Which brings us to my next acknowledgment. In this test, DALL-E is being forced to do something it's really not built for. It does admirably considering this, but it's not the tool for the job. Use it to make mood boards, get inspiration, and ask it for ideas on what you could you possibly use to represent some deep esoteric idea inside of 200 x 200 pixels.
Illustrator on the other hand should be perfect for this. If adobe improves its generative model, in both output and prompt comprehension, it will be incredibly powerful tool.
Created with: DALL-E 3
When the final bell rings, and before the judges announce their decision, I have to acknowledge a few things. Strictly speaking these are not icons. They're mini-graphics. I wouldn't use them in an interface or a menu for example. Best used to illustrate blocks of text and to draw the eye. They get called icons a lot, but are inappropriate for navigational use.
We're creating mini-graphics instead of true icons because creating a set of interface or design system icons in DALL-E, as it currently stands, would be practically impossible. Which brings us to my next acknowledgment. In this test, DALL-E is being forced to do something it's really not built for. It does admirably considering this, but it's not the tool for the job. Use it to make mood boards, get inspiration, and ask it for ideas on what you could you possibly use to represent some deep esoteric idea inside of 200 x 200 pixels.
Illustrator on the other hand should be perfect for this. If adobe improves its generative model, in both output and prompt comprehension, it will be incredibly powerful tool.
Lastly how do the graphics compare? First up, Illustrator’s set: clean, scalable, and editable, but lacking sparkle. They do the job—imagine a competent junior designer meeting the brief but playing it safe. Shapes are logical, the images are recognisable, and the color palette is obediently on-brand. No real problems, but no standing ovation either.
Then there’s the DALL-E 3 set. These icons have personality. They take creative risks. There’s unexpected flair, charming quirks, and a sense that DALL-E, despite not understanding what an icon should be, decided to make something delightful anyway. The problem? They’re stuck as raster images. You can't scale them up, tweak them, or swap a color without some extra work.
If we’re talking raw artistic instinct, DALL-E 3 is the clear winner. But in terms of practical, usable output? Illustrator has some potential. Just… don’t ask it to surprise you.
