When Everything Starts Looking Like Claude

One of the things I genuinely admire about modern AI tools is how quickly they raise the quality of work that people can produce. Someone with no formal design training can now create a polished presentation, a credible landing page, or a well-structured document in minutes. That is an extraordinary step forward. For founders, consultants, charities, small businesses, and anyone who has ever stared at a blank page wondering where to begin, tools like Claude Design remove a huge amount of friction.

I use AI every day. I use it to explore ideas, challenge assumptions, improve drafts, and occasionally help with visual design. I don’t think that is that controversial anymore. Used well, these tools make me more productive and allow me to spend more time on the parts of the work that matter most.

But I’ve found myself noticing something recently. I can usually tell when something has been designed by Claude. Not because it looks bad (quite the opposite) - but because it looks like Claude.

Once you’ve seen enough Claude-generated work, patterns begin to emerge. The muted pastel palette. The generous spacing. The rounded cards. The understated icons. The typography hierarchy. The distinctive treatment of headings, complete with those familiar all-caps descriptors sitting above the main title. Even the layouts begin to feel familiar.

None of these individual choices are problematic. In fact, most of them represent good design practice. They’re clean, readable, accessible, and generally pleasant to look at. The issue is that they are becoming recognisable.

Design has always reflected the tools available at the time. You can often date a website within a few years simply by looking at its navigation, typography, or visual style. Early web pages, Flash websites, skeuomorphic mobile apps, Material Design, flat design - each era developed its own visual language before eventually giving way to something new.

I suspect we are witnessing the beginning of another such period. The difference is that this time the style isn’t emerging from thousands of designers influencing one another. It is emerging from millions of people drawing from the same statistical model.

Part of Claude Design’s appeal is that it has absorbed years of accumulated design wisdom. It understands hierarchy. It respects whitespace. It follows accessibility principles. It builds coherent layouts. It produces work that feels balanced and professional. That is precisely why it is so useful.

The danger comes when we mistake consistency for creativity. Design systems have transformed software development because they create coherence across products. They reduce unnecessary variation and make interfaces easier to understand. AI extends that principle further still, giving almost anyone access to good design patterns.

Yet great design has never been about simply applying the correct patterns. It has always involved judgement. Knowing when to follow convention, and knowing when to ignore it.

The world’s most memorable products rarely looked exactly like everything else when they launched. They introduced new ideas, challenged expectations, or deliberately broke prevailing conventions. Many of those decisions probably looked slightly wrong before they looked visionary.

AI, by its nature, is optimising towards what is most likely. Original design often begins by moving away from exactly that point.

I sometimes hear people describe AI-generated work as “80% there”. Usually that’s presented as a limitation. I’m beginning to think the real problem is that 80% is already more than good enough for most organisations.

Historically, that remaining 20% was where much of the value was created. Not because the basics were unfinished, but because that was where someone exercised taste. They chose an unexpected typeface. They simplified rather than adding. They removed a visual element everyone else would have kept. They introduced personality rather than polish.

That final layer is difficult to describe because it isn’t a checklist. It is judgement - and judgement has always been expensive because it comes from experience rather than process.

If AI makes the first 80% almost free, there is a risk that organisations conclude the final 20% no longer matters. I think they’ll regret that.

There is an argument that the last defensible advantage humans have over AI is taste. I’m not entirely convinced that’s the right way of thinking about it. AI doesn’t need to possess taste in order to imitate the outputs of people who do. The greater risk is that organisations become satisfied with imitation.

If every company can generate a polished website, an attractive slide deck, and a respectable annual report in a matter of minutes, visual quality stops being a differentiator. The average improves dramatically, which is unquestionably a good thing. But the ceiling doesn’t necessarily move with it… In fact, it may become harder to justify the investment required to reach that ceiling.

That would be a shame, because the businesses that genuinely stand out rarely do so through competence alone. They stand out because somebody made a series of subjective decisions that no statistical model would ever have considered the safest option.

This is why I don’t believe AI replaces creativity. It changes where creativity sits. The mechanical execution becomes increasingly automated. Producing layouts, refining copy, maintaining consistency, checking accessibility, generating alternatives - all of these become faster and cheaper.

What remains stubbornly human is deciding what should exist in the first place.

Those questions cannot be answered by prompting alone because they aren’t questions of execution. They are questions of identity.

The same applies to writing. I have no issue saying that AI forms part of my workflow. It helps me explore arguments, identify gaps, test structure, and improve clarity. But the direction, the opinion, the experiences, the judgement, the examples, and ultimately the point of view still have to come from somewhere. Otherwise, what exactly are we optimising?

The tool can help me express an idea - it cannot decide which ideas are worth expressing.

I suspect we’ll look back on this period much as we now look back on the first generation of stock photography. Suddenly everyone had access to professional-looking images. Marketing became cleaner, more polished, and dramatically cheaper to produce. It also became strangely familiar.

Eventually, brands rediscovered the value of commissioning original photography because everyone else was using the same stock image mess of smiling people shaking hands in anonymous meeting rooms. I wonder whether AI-generated design will follow a similar path.

The organisations that rely entirely on statistically average design will look increasingly similar to one another. Those that continue to invest in genuine creative direction, distinctive visual identity, and thoughtful design will become easier to recognise precisely because they are swimming against a rising tide of competent sameness.

Perhaps that’s the future? AI creates the average / humans create the memorable? As the average continues to improve, the value of being memorable may only become more important.

Before you publish that AI-generated presentation, landing page, proposal, or report, ask yourself five simple questions…