For a long time, making something that looked genuinely good required a certain amount of knowledge, practice and time. You had to understand the tools, whether that meant Photoshop, Figma, a camera, a typeface or a programming language. You had to develop an eye for composition, typography, colour and proportion, and you had to make enough bad work before you could consistently make good work. That friction was frustrating, but it also served a purpose. It forced people to learn how things worked before they could reliably make them work. Generative AI is removing a lot of that friction. In 2026, you can generate an image, explore a visual direction, create a website concept, rewrite content, produce interface variations or even generate motion simply by describing what you want. Figma is now building AI directly into the design workflow, while Adobe has integrated generative and agentic tools across its creative ecosystem. This is no longer some distant possibility for designers. It is already part of the software they use every day. There is plenty to like about that. Adobe's 2026 Creators' Toolkit Report, based on more than 16,000 creators, found that 87% of creators using creative AI say it has accelerated the growth of their business or audience, while 75% describe it as integrated or essential to their workflow. At the same time, 85% say the final creative decision should always remain theirs. That last number is particularly interesting. It suggests that the real debate isn't whether AI belongs in creative work. It is about where we draw the line between assistance and authorship. Because AI can make good design. That's the problem.
Good-looking and good are not the same thing. A polished website can have terrible UX. A beautiful identity can be completely forgettable. An impressive image can communicate absolutely nothing. Design has never really been about producing attractive objects; it is about making decisions in response to a particular problem, audience and context. The visual result is what remains after those decisions have been made. Generative AI is extremely good at producing plausible answers, but it doesn't necessarily know whether the question was worth asking. Give it a description of a contemporary website and it can produce something that looks like a contemporary website. Ask for a luxury brand and it knows the visual language of luxury brands. Ask for an experimental editorial layout and it can remix the conventions associated with that aesthetic. The results can be genuinely impressive because these systems have absorbed an enormous amount of existing visual culture. The problem is that if millions of people use similar systems, trained on broadly similar material, the ability to produce convincing work can start pushing people towards similar answers. There is now growing research behind this concern. A 2025 meta-analysis covering 28 studies and more than 8,000 participants found that people collaborating with generative AI performed better on creative tasks, but their ideas were also significantly less diverse. Even more recently, research published in July 2026 found that AI-generated ideation could compress collective creative diversity, while using AI to refine people's own ideas was much less damaging to that diversity. The distinction is important: the problem may not be AI itself, but what part of the creative process we hand over to it. You can see how easily this happens in practice. When a tool can instantly generate twenty options, it becomes tempting to stop asking what should exist and start asking which option looks best. The first question requires a point of view. The second requires a preference. Those aren't the same thing. A designer's value is often hidden in decisions that don't appear in the final image: what was removed, why a particular typeface was chosen, why an interaction behaves in a certain way, why a layout is deliberately uncomfortable or why a seemingly obvious solution was rejected. The finished website may look simple, but the thinking behind that simplicity rarely is.
This matters for the creative industry as well as for clients. Designers learn through making things, including the boring, repetitive and imperfect work that gradually develops judgement. Automating unnecessary production is obviously useful, but if AI starts removing too much of the work through which junior creatives learn, the industry could end up with fewer opportunities to develop the experienced designers it will still need. At the same time, clients face a new problem: when almost anyone can generate something that looks professional, it becomes harder to distinguish visual polish from actual expertise. There is also a problem for everyone consuming this work. If production becomes dramatically cheaper, we will probably produce more of it: more ads, more landing pages, more product imagery, more social content and more variations of everything. Human attention does not scale at the same speed. The internet already has an almost infinite supply of competent visual material. It doesn't necessarily need another perfectly acceptable image. This is where "AI slop" becomes more interesting than simply a term for ugly generated content. Slop can be technically excellent. It can have perfect lighting, sophisticated typography and a convincing colour palette. What makes it slop is often the absence of a reason for it to exist. The real danger isn't necessarily that everything will start looking terrible. It might be that everything starts looking perfectly fine.