The Recommendation Row Is Lying to You (Politely)

It is 8:47pm on a damp Tuesday, and the telly is doing that thing again. You open your streaming app of choice, scroll past the big banner for a film that looks like a perfume advert, and land on the row labelled ‘Because You Watched…’. What follows is either a show you finished three years ago, a documentary about competitive falconry, or yet another Scandinavian crime drama in which someone stares at a fjord.

This is not a new complaint, but by November 2026 the recommendation row has become one of the most quietly powerful bits of furniture in the British living room. It shapes what we sample, what we abandon and what we end up arguing about over the washing-up. And yet it frequently feels less like a clever assistant and more like a flatmate who once saw you eat a croissant and now buys you croissants every week.

The Flatmate Who Thinks It Knows You

The core problem is that streaming algorithms are brilliant at spotting patterns and terrible at understanding context. If you watch one episode of a cosy murder mystery while folding laundry, the system does not see a tired person who wanted background telly. It sees a viewer with a confirmed interest in British crime, village life and suspicious vicars.

From there, the row multiplies. You get more murders, more villages, more detectives with complicated home lives. The app has decided that this is your personality now, and it is going to keep serving it until you either submit or move to a different platform entirely.

Shared accounts make this worse. A household profile can contain a teenager’s anime phase, a parent’s obsession with property programmes and someone else’s late-night true-crime binge. The algorithm blends it all into a recommendation smoothie that tastes faintly of everything and satisfies no one. No wonder the row sometimes feels like it was assembled by a committee who have never met.

Cosy, Crime and the Great Genre Puddle

British telly has a particular knack for blurring genres. We have cosy crime, grim crime, crime with a nice kitchen, crime set on a windswept island, and crime where the detective is the only person in the village who owns a decent coat. To a human viewer, these are distinct moods. To a recommendation system, they are often just ‘crime’, ‘British’ and ‘popular’.

That is why you can finish a gentle Sunday-night mystery and be immediately offered something bleak enough to require a follow-up comedy. Or why watching a baking show can lead to a row of ‘feelgood’ programmes that includes a documentary about a disastrous hotel and a drama about a family falling apart in slow motion. The labels are technically accurate and emotionally useless.

The same happens with comedy. Watch one panel show and the app assumes you want every panel show ever made, plus a stand-up special recorded in a theatre with aggressive lighting. Watch one weepy drama and it decides you are ready for a six-part exploration of grief. There is little room for the very British mood of ‘something on in the background while I sort out the recycling’.

What the Data Actually Sees

Streamers do not watch over your shoulder. They see signals: what you started, what you finished, what you paused, what you skipped, what you searched for and what you left running while you answered a message. The trouble is that all these signals get treated as affection.

Autoplay is a particular offender. Leave a show running while you make tea, answer the door or fall asleep on the sofa, and the platform may log it as a completed episode. It does not know you were not really there. It certainly does not know you were mildly annoyed by the whole thing. It just knows the episode played, and so the row grows another branch.

There is also the small matter of what each platform wants. A recommendation engine is not a public service. It is there to keep you inside the app, which means it may favour a show it owns over a show you would actually love. That is not sinister, but it does explain why the row can feel less like a thoughtful bookseller and more like a shelf stacked by the shop’s marketing team.

The Loop of Three Shows

We have all been trapped in the loop. The row offers the same three shows in slightly different orders, like a shuffled playlist with only three songs. You can tap ‘not interested’ if you can find the button, but the effect is often temporary. The system nods politely and then suggests the same thing again next week, perhaps with a different thumbnail.

What is missing is memory with nuance. A useful system would remember that you abandoned a series after two episodes, not because you hated it, but because it was too long for the time you had. It would know that you like crime dramas but not ones with lingering shots of autopsies. It would understand that ‘cosy’ means different things on a Friday night and a Sunday afternoon.

What a Genuinely Useful Recommendation Engine Would Look Like

A better row would start with mood, not genre. It would ask — or infer — whether you want something comforting, something gripping, something funny or something you can half-watch while scrolling. It would factor in time: one episode before bed is a different request from a free Saturday afternoon. It would know whether the kids are in the room, whether you are eating, and whether you have the attention span for subtitles.

It would also be honest. A simple ‘why this?’ label would go a long way: ‘Because you finished three episodes of a British crime drama’ is at least a starting point. A proper ‘less like this’ option, rather than a buried thumbs-down, would let viewers steer the thing without writing a review. Separate profiles would help, but so would a shared profile that understands compromise — the streaming equivalent of one person wanting a thriller and the other wanting a nap.

For British viewers, the details matter. Availability changes constantly across BBC iPlayer, ITVX, Channel 4, My5, Netflix, Prime Video, Disney+ and NOW. A useful engine would know the difference between a live channel, a box set and a film leaving on Sunday. It would not recommend a six-part drama to someone who has twenty minutes and a mug going cold.

And it would leave room for surprise. One wildcard per row, clearly marked, would be enough. Not a random documentary about competitive falconry — unless, fine, you are into that — but something adjacent to your taste rather than identical to it.

Until then, the ‘Because You Watched’ row will keep doing its polite little guess. Sometimes it will be right. Often it will be baffling. And every so often it will offer up something genuinely brilliant, which is probably why we keep scrolling. The row is not evil. It is just overconfident. A little humility — and a decent ‘not tonight’ button — would make the whole sofa feel smarter.