The Spotify Effect: How AI Is Impacting UX

The Spotify-ification of UX  #

AI has become a significant part of the UX practice; notably, when Claude Design was released in April 2026, it felt like an important inflection point in how AI was beginning to change the UX workflow. 

Like many designers, I've used it to synthesise research, structure workshops, explore user journeys, critique concepts and accelerate everything from discovery outputs to prototype content. The productivity gains are difficult to ignore; tasks that used to take hours can be completed in minutes, early concepts come together faster, and information is easier to interrogate, organise and summarise. The blank page feels a little less intimidating when you have something to react to. 

What started a conversational chatbot has expanded into a suite of AI features embedded in research, design, prototyping and development tools, and increasingly, we are asking AI not just to help us do something but to suggest what we should do. 

That shift is where I think things become interesting. 

For the purposes of this article, we will look at AI through a slightly different lens. 

Rather than focusing on productivity, tools or prompts, I want to think about AI in the same way we might think about recommendation engines such as Spotify. 

At first glance, they solve different problems - one recommends music, the other can generate or evaluate ideas. But underneath, they share a similar principle; both help us navigate overwhelming amounts of information by predicting what is most likely to be relevant, useful or successful. 

This raises an interesting question for UX, where some of the most valuable outcomes emerge not from certainty, but from exploring the unknown. 

Why AI recommendations feel so compelling #

Part of the reason Spotify works so well is that it removes friction from discovery. There is an almost ridiculous amount of music available to us; millions of songs, thousands of genres, endless artists, albums and playlists. We could spend hours searching for something to listen to but, instead, Spotify gives us Discover Weekly, Release Radar, personalised playlists and recommendations based on what we've listened to before. 

It makes a prediction: You might like this. And more often than not, it's pretty good. 

There is something psychologically appealing about this loop. Crucially, we haven't removed the decision completely - but we've reduced it to choosing between a handful of possibilities the system has already narrowed down for us. 

In Kahneman's terms, a potentially effortful decision becomes something closer to pattern recognition: Kylie or Gaga, rather than what music should I listen to? 

AI tools offer a similar experience for designers facing the uncertainty of a blank page; they immediately create structure and offer direction. 

Ask an AI model to generate an onboarding flow, and you'll probably receive recommendations aligned with established UX principles: progressive disclosure, logical hierarchy, familiar navigation patterns and carefully chunked information. If you ask it to critique a design, it will often identify usability considerations that an experienced UX practitioner would recognise. 

In many ways, that's remarkable and importantly, this isn't an argument that AI produces bad UX. My experience has largely been the opposite. 

AI is often very good at producing competent, well-reasoned solutions because it has access to an enormous body of collective knowledge. The challenge is that competence and innovation aren't always the same thing. 

UX is often a process of becoming comfortable with the fact that complexity and disorganisation is to be expected. During exploration and early synthesis, we learn to sit in the discomfort of the unknown. 

At the beginning of most projects, there is uncertainty everywhere: research findings can point in different directions, stakeholders have competing priorities, business goals aren't always aligned with user needs, and the problem itself is often still being defined. 

We conduct discovery because understanding the problem is often more valuable than rushing towards a solution. We test ideas because our assumptions might be wrong; we iterate because the first attempt rarely reveals the full picture. 

From the outside, this process can look inefficient.  

Why spend time talking to users when AI can summarise research in seconds?   

Why explore five different concepts when AI can recommend the most appropriate one?  

But is something inefficient if it helps us discover a flawed idea while it's still a wireframe, rather than after it has consumed months of engineering effort and hundreds of thousands of pounds in development? 

The discomfort that comes with uncertainty isn't necessarily a flaw in the process  it is often where the value is created. 

From assistant to recommender  #

I think we're beginning to see an important shift in how we use AI. 

The first phase was largely about execution: 

• Help me write this. 
• Summarise this. 
• Turn these notes into something useful. 

Then it became more collaborative: 

• What do you think of this? 
• What am I missing? 
• Give me some alternatives. 

Now we're increasingly asking: 

• What should I do? 

There is a big difference between AI helping us execute a decision and AI influencing which decision we make in the first place. 

Spotify doesn't simply give us access to music. It decides what is worth our attention, and AI is increasingly doing something similar with design. 

AI’s UX playlist: recommending design patterns #

Imagine asking an AI to design a better onboarding experience for a financial services product. 

It might suggest: 

•  Progressive disclosure 
•  Personalisation 
•  Smart defaults 
•  Clear hierarchy 
•  Contextual help 
•  Short forms 
•  Familiar navigation  

It has effectively created a playlist of established UX patterns that are statistically likely to work. 

This is where the concept of "best practice" becomes interesting.

Jakob's Law tells us that users spend most of their time using other products, and therefore expect ours to work in similar ways. Familiarity reduces cognitive load, and consistency improves learnability. Established conventions exist because they solve recurring problems - from that perspective, AI should be excellent at UX. 

After all, it has absorbed an enormous amount of accumulated design knowledge. 

But there is a subtle difference between best practice and better practice. 

Best practice tells us what has worked. Better practice sometimes begins with asking whether something else might work better. 

The most significant advancements in digital experiences began as a departure from accepted convention. 

The original iPhone interaction model, pull to refresh, and infinite scrolling were all once not “best practice”. At the point those ideas emerged, they didn't have a safety blanket of evidence and data. They became conventions because somebody was willing to explore beyond what was already known. 

So, can AI tooling reliably recognise which unconventional idea is worth pursuing before there is evidence that it works?  

The psychology of the probable  #

Humans like things that feel easy, familiar and certain. We're susceptible to cognitive shortcuts; we experience choice overload, and we look for signals of authority. AI is remarkably good at exploiting these preferences, but it also creates an interesting psychological trap. 

The more polished the recommendation, the easier it is to mistake confidence for correctness. An AI-generated UX recommendation can arrive wrapped in the language of established principles, patterns and evidence. It sounds considered and visually it looks comprehensive because the system has processed vastly more information than an individual designer could, so as a result, its recommendation can feel authoritative. 

We might therefore stop asking:  What else could we do?  and start asking: Which of these recommendations should we choose? 

That's a subtle but important change in the role of the designer.  

Could AI create a culture of UX convergence?  #

It's looking increasingly as if the future could be AI embedded throughout the UX process, and the systems providing recommendations have learned from broadly the same ecosystem of existing products, design principles, research and patterns. 

Perhaps nothing dramatic happens but, there is a possibility that we begin converging towards similar conclusions, similar interaction patterns and similar solutions to similar problems because they are likely to work. 

This is the paradox: the better AI becomes at recommending competent design, the more tempting it becomes to stop looking for something different – and potentially better. 

The Spotify effect: how AI shapes design preferences  #

Spotify has solved an enormous problem for music discovery, but recommendation comes with a paradox of its own. 

The system learns what we like, recommends more of it, we listen to those recommendations, and our listening behaviour gives the system more information. 

Our preferences are, in part, shaped by what we're exposed to. The recommendation system isn't simply responding to our taste; it is actively participating in the formation of it. 

AI could potentially do something similar to design. If designers increasingly use AI to determine what a good interface looks like and what good content looks like, then AI isn't simply reflecting design culture; it is actively participating in shaping it. 

That is what I mean by the Spotify-ification of UX. 

AI doesn’t just reflect our preferences; through what it recommends, it helps shape them. 

So where do trends come from next in an AI-driven world?  #

Historically, design has progressed through a combination of imitation and divergence. 

Someone creates something, others copy it and it starts to become familiar, and eventually even best practice.   

Then someone challenges it, and the cycle starts again. 

But if AI becomes increasingly good at identifying what already works and recommending it back to us, where do the next patterns come from? 

If every designer has access to an incredibly capable system that can instantly generate the established answer, what happens to the weird answer or the idea that doesn't have enough evidence to be recommended? 

The ability to interrogate vast amounts of information, generate alternatives and rapidly test ideas is incredibly powerful. But perhaps we need to become more deliberate about where we use AI to reduce uncertainty and where we deliberately preserve it. 

Designer as the 'skip button'  #

Spotify gives us a Skip button. It is a tiny but important part of the experience. 

The recommendation is there; we can listen or we can decide we don't like the song. 

This is one way of thinking about the future role of the designer. UX designers still need to decide what is worth optimising for. 

The valuable skill may therefore shift away from simply producing design and towards developing the judgement to recognise when the recommendation is wrong, incomplete or simply too obvious. 

Perhaps the future of UX isn't about resisting the recommendation engine. It's about knowing when to press Skip. 

UX isn't really about finding answers as quickly as possible; it's about understanding problems deeply in order to ask the right questions. 

If AI is going to curate the future of design, our job is not to reject the playlist - it's to keep making new music. 

Key takeaways: AI’s impact on UX design #

• Like Spotify, AI helps reduce friction by narrowing possibilities. This can accelerate decision-making and improve efficiency, but it also changes how we explore ideas.
• Competent design and innovative design are not always the same thing. AI is exceptionally good at generating solutions grounded in established patterns, principles and conventions. 
•  The uncertainty we try to eliminate is often where UX creates value. Discovery can feel inefficient, but it helps to uncover flawed assumptions before they become expensive realities.
• Best practice reflects what has worked before. Better practice often begins by questioning it.
• AI recommendations carry an inherent authority bias. The more polished and evidence-based a recommendation appears, the easier it becomes to confuse probability with certainty.
• There is a risk of design convergence. If designers increasingly rely on systems trained on the same body of knowledge, we may gradually arrive at similar solutions to similar problems.
• The designer's role is shifting from creator to curator of possibilities. One of the most valuable skills may be recognising when the recommendation is incomplete, inappropriate or simply too obvious.
• AI can help us find answers faster, but UX remains about asking better questions. Preserving curiosity, exploration and healthy uncertainty may become more important, not less, in an AI-assisted design process.