What can personalised news miss?
Mainly the stories you didn’t know to ask for: a council decision, a health warning, a breakthrough in a neighbouring field. The bigger fear, that algorithms seal everyone inside a bubble, has weaker evidence behind it than you might expect. The real risk is quieter, and it is one you can manage.
The appeal—and the trade-off
Following your own interests is a reasonable way to tame a huge amount of news. A marine biologist may want new ocean research. Someone renovating a house may want planning rules and energy prices. A selection built around those interests can spare them a lot of material they would otherwise wade through.
But relevance is not the same as importance. Relevance means the story matches what you asked for. Importance means it affects you whether you asked or not. A change to local bus routes or a public-health warning can matter a great deal and still fall outside every topic you picked. Carrot is a personalised news app in development, so this is a trade-off we need to take seriously, not wave away.
None of this is unique to software. A newsletter, a newspaper chosen for its outlook and a list of favourite websites all have edges. Every selection has boundaries, whether a person, an editor or software made it.
The filter-bubble story is more complicated
The popular version says algorithms trap everyone in a bubble of agreeable news. The research is less tidy. A 2022 Reuters Institute literature review found that using search engines and social media tends to lead people to slightly more diverse news, not less. One reason is that ranking systems sometimes turn up sources people would not have picked themselves. At the time of their 2022 review, the authors said they knew of no comparable studies supporting the filter-bubble idea, and found that partisan echo chambers are real but generally small.
Two cautions apply. Much of that research comes from the United States, which the authors call an outlier, so the findings may not travel to other countries. The studies also looked at which sources people encountered, not whether their views changed, and most predate AI-written news summaries. They say little about how a brief produced that way might narrow or widen what you see.
That leaves the oldest filter of all: us. We decide what to follow, which link to open and which source to trust. These choices are useful shortcuts, and they can narrow attention as surely as code. Rather than asking whether a product creates a bubble, ask what this particular selection includes, and what it leaves out.
Personalisation means different things
Readers are split. In a 2025 Reuters Institute analysis covering 27 markets, close to half of respondents said they were comfortable with news personalisation. That was lower than their comfort with personalised weather, music or TV. In their own words, wary respondents worried that personalisation might block out important stories, replace a general overview with a narrow one, intrude on their privacy or take away their choice. Interest in specific AI-powered features was modest, below 30% for any single option. These are reported attitudes, not a test of any product.
It helps to separate what ‘personalised’ means. Choosing topics yourself is different from a service inferring your interests from your behaviour. Changing how a story is explained is different from deciding which stories appear at all. An earlier Carrot alpha included a way to describe a stream in your own words and separately adjust writing style and detail. That still decides which stories appear, so the blind-spot question applies to us too.
Three ways to keep a wider view
Keep a general source in the mix. A local newsroom, a public broadcaster or a specialist publication may cover developments that don’t fit your saved interests. No single source is complete; the point is to have a second route into the news.
Compare the evidence, not just the labels. Two outlets can repeat one announcement. When it matters, open the underlying material and ask who is affected, what is known and what remains uncertain. Our guide to checking an AI news summary sets out a short routine.
Review your own boundaries. An exclusion can cut noise and also hide a useful exception. If you follow a technical field, for instance, leaving out business news might also leave out a funding decision that determines which research continues.
Choose the size of the window
A focused brief is useful when it helps you follow a question over time. It is less useful if you expect it to give a complete account of everything important. A broad source and a focused brief do different jobs; keeping both in the mix can help.
That is how we think about Carrot. We are building Carrot to sit alongside a broader source, so you can keep a wider perspective. It is still in development, and we have not yet shown how well it surfaces important stories beyond the streams you choose. For why constant updates can feel so draining in the first place, read why the news can feel relentlessly negative. Neither piece offers a formula for the perfect news diet. The practical question is whether your mix lets you notice both what matters to you and what you might otherwise miss.