Open any app on your phone right now. Chances are it has an AI feature that didn’t exist six months ago. Your email client wants to write your replies. Your photo editor wants to generate backgrounds. Your note-taking app wants to summarise your own thoughts back to you.

Some of these features are actually helpful. But a distressing number of them feel like they were added because an executive saw a competitor’s press release and panicked.

The AI Feature Arms Race

Since late 2023, there’s been an unspoken rule in software: if your product doesn’t have an AI feature, investors will ask why. This pressure has created a gold rush of AI integrations, many of which solve problems that don’t exist.

Take AI-generated email replies. Gmail, Outlook, and a dozen other email clients now offer to draft responses for you. In theory, this saves time. In practice, the suggestions are so generic that you spend as much time editing them as you would have spent writing from scratch. And there’s something deeply weird about two people having a conversation where both sides are written by AI.

Or consider AI-powered playlist generators in music apps. Spotify’s algorithm was already doing a perfectly good job of recommending music based on listening history. The addition of a chatbot interface to do the same thing doesn’t make the recommendations better — it just adds an extra step.

The Verge has been tracking this trend, and their coverage paints a clear picture: companies are shipping AI features at a pace that outstrips user demand.

When AI Features Actually Help

To be fair, some AI integrations are genuinely brilliant. Code completion tools like GitHub Copilot have measurably improved developer productivity. AI-powered transcription in meeting tools like Otter.ai has saved countless hours of manual note-taking. Smart photo search in Google Photos — being able to type “dog at the beach” and find the right image — is the kind of feature that feels like it should have always existed.

The pattern is clear: AI works best when it handles tedious, repetitive tasks that humans don’t want to do. It works worst when it tries to replicate tasks that require nuance, creativity, or personal judgement.

Writing a thoughtful email to a colleague? That requires knowing your relationship, the context of the conversation, and the right tone. No AI can do that for you. Transcribing an hour-long meeting into searchable text? That’s pure drudgery, and AI handles it perfectly.

The Cost of Useless Features

There’s a real business cost to cramming AI into everything. Development resources that could be spent on fixing bugs, improving performance, or building features users actually want are instead diverted to AI projects that look good in marketing materials but add little practical value.

Team400, an AI consultancy, has written about this problem: the companies getting the most value from AI are the ones that start with a specific business problem and work backwards to a solution. Companies that start with “we need to add AI” and then look for places to put it almost always end up with features that feel bolted on.

Users notice. App store reviews are increasingly full of complaints about AI features that get in the way. “How do I turn off the AI?” is becoming one of the most common support questions across platforms.

Why Companies Keep Doing It Anyway

Three reasons.

Investor pressure. If you’re a publicly traded tech company or a startup seeking funding, AI is expected. Not having an AI story in 2026 is like not having a mobile strategy in 2012. It doesn’t matter if the feature is useful — it matters that it exists.

Competitive anxiety. When your competitor announces an AI feature, the temptation to match it is enormous. Nobody wants to be seen as falling behind, even if “behind” just means “not adding unnecessary features as fast.”

Genuine uncertainty. AI is moving so quickly that even the people building these products aren’t always sure what will be useful and what won’t. Shipping a feature and seeing if users adopt it is faster than spending months on market research. Some of today’s throwaway features might evolve into genuinely useful tools. Most won’t.

What Users Can Do

The best thing you can do is vote with your usage. If an AI feature is helpful, use it. If it isn’t, ignore it — and tell the company why. Most software companies track feature adoption closely. If an AI feature has low engagement, it’s more likely to be improved or removed.

Also, be wary of paying extra for AI features you don’t need. Many SaaS companies have introduced premium “AI tiers” that bundle AI features with their subscription plans. Before upgrading, ask yourself honestly: will I actually use this? Or am I paying for a feature I’ll try once and forget about?

The Pendulum Will Swing

Every technology cycle goes through this. Remember when every app needed a social feature? Or when blockchain was going to be in everything? The initial hype phase is always messy. Features get shipped that shouldn’t. Money gets spent that shouldn’t.

Eventually, the market figures out where AI genuinely adds value and where it doesn’t. We’re still in the messy phase. But the correction is coming, and the companies that survive it will be the ones that built AI features people actually wanted — not the ones that built them because they felt they had to.