I don’t think Groove Armada was singing about AI in their 1999 dancefloor banger, unless they knew something the rest of the world didn’t, but have you noticed that AI is making blandness a trend? 

To be fair, Ui design was noticeably heading in the same design direction before AI, but now it’s jumped on a motorbike wearing a helmet with blinkers and is going full throttle into the sunset.

If you don’t believe me, let’s try a quick experiment; Step one, jump on Google; Step two, search something like ‘cybersecurity’; Step three, have a look at dozen sites or so. The findings? Navy blue or dark grey backgrounds, check; Electric blue accents, check; A shield shaped logo with a padlock, check; Stock photography image of abstract networks in electric blue with glowing halos on a dark blue background, possibly with some code overlaid for effect [optional extra], check. Summary; 12 different sites, 1 near identical Ui aesthetic.

The test works in most sectors, not just cybersecurity, where you can’t tell the brands apart: 

Fintech: The new entrants have worked hard to portray a combination of new and funky but techie and smart, to differentiate themselves from the stale high street banks. There were bright colour palettes, gradients suggesting movement, abstract geometric marks hinting at currency without committing to it, all chosen to make them feel modern not institutional. But what used to be different has curdled into a kind of uniform.

Wellness: Earthy tones, lowercase lettering, lots of white space and elegant serif typefaces. No idea whether it’s a spa, supplement, sour dough bread or oat milk but it’s definitely going to make you feel healthier, calmer and younger.

SaaS: Hero section with a floating dashboard, tech sans serif typeface, headlines about workflows and teams all finished off with one of 6 accent colours.

AI: Not sure what the collective noun for AI companies is, but, referring back to the lyricists above, it could well be an 'armada of AI' as there seems to be millions of them. We’re seeing lots of orbital shapes, connection icons and of course names that end in AI. Given that the technology is supposed to be reshaping human creativity, why do they all look like they’ve been designed by committee?

See previous blog posts:

https://bd2.com/insights/itll-do/

https://bd2.com/insights/serif-or-sans-serif-that-is-the-slightly-interesting-question-not-just-for-typographers/.

If AI needed a defence, and it probably doesn’t care anyway, things were already heading in the direction of uniformity before it properly joined the party. We blogged about the simplification of brand design some time ago, see: 

https://bd2.com/insights/the-simplification-of-brands/

This blog looked at the many brands who have stripped back their logos to make simpler icons partly because they need to work as a centimetre square for apps, and partly to follow the trend. It’s really evident in the car industry where 3D effects and embellishments have disappeared in favour of ‘flat’ graphics, and especially in fashion where everyone has gone for just black text in caps on white. 

Inevitably, because AI references what’s gone before, it has accelerated this descent into sameness. It’s the copy of a copy problem: 

“AI does not invent. It identifies the most statistically probable output based on what it has seen. What it has seen is the internet, already narrowed by cautious brands, already trending toward the same palettes with the same typefaces and layout conventions. Research published in 2026 confirmed that AI tools used to build websites reproduce dominant style conventions from their training data, accelerating convergence toward a visual mean.”

Vanessa Porter, artificial intelligence and machine learning expert.

The problem lies in the feedback loop. AI-generated output gets its data from the web, the next model trains on it, narrowing the range of what it sees as ‘good design’. AI company logos are themselves the most obvious example: the technology supposedly reshaping creativity has produced some of the most derivative visual identities in recent memory, built largely on orbital shapes and the same shade of corporate blue.

“Think of photocopying a document and then photocopying the copy. The fine detail disappears, and what remains is a high-contrast, simplified version of the original.”

One of the most referred to and criticised examples is Coca Cola’s 2025 Christmas ad. It's based on the 1995 ‘Holidays are Coming’ ad with lorries heading up winding roads, but in the first version the wheels on the truck didn’t go round and round properly and the lighting was all over the place. The furore led them to remaster it, but it still felt like an imitation of nostalgia rather than the ‘real thing’ to coin a somewhat appropriate phrase, and when a brand’s value relies on emotional warmth built up over decades which isn’t in the training data, something gets lost. 

Sadly, as our ‘it’ll do’ blog debates, there is an increasing acceptance of poorer, but obviously cheaper, AI generated output that’s deemed good enough. However, there is a cost to this sameness and it’s not just boredom, although there is plenty of that. Just think about all those dreadful illustrations and cartoons suddenly appearing on Linkedin with their yellow cast which just lead to a collective rolling of eyes and scrolling past as you can’t be bothered to read all the crammed in text - 'Ai slop'. When everything is the same, users default to familiarity and price and brand equity, which has been built up over years or decades, is eroded and ultimately becomes worthless. That's not to say there wasn't plenty of soulless design work before AI came along, the difference now is that AI can make that soulless work in seconds.

The answer is not to refuse to use AI, that’s like being King Canute on the seashore with the water already waist high, the answer is to understand what AI does and to prompt and interrupt effectively.

Start with references from outside the algorithms. If you’re using sites like Behance and Pinterest you are drawing from a pool already shaped by what worked well on those platforms. So films, books, photography, architecture and cultural references will provide inputs that haven’t been overused. Use AI for the ideation phase not the end result.

AI is great rapid exploration and visualisation of ideas, but not so good at identifying what makes something distinctive because originality is, by definition, statistically unusual. So, use AI for the 70- 80 per cent it generates quickly as a starting point, and then add the 20-30 per cent that makes it interesting and memorable which still requires a human eye. It’s the human element which AI can’t generate that makes it valuable and, because AI has that inherent and generic slickness which makes it look cheap, it leads to design decisions that are deliberately disruptive, that embrace friction and even imperfections, that use old-school skills like hand drawn typography and unique illustration styles to create something original that audiences will notice, not ignore.