Let people fix their recommendations in real time, not hours later
For years, the invisible hand of algorithmic recommendation has quietly reshaped what millions of people see, often based on a single distracted glance. Instagram is now testing a more transparent compact with its users — placing the tools to redirect that invisible hand directly into the stream of everyday browsing, rather than hiding them in the architecture of settings menus. The move reflects a broader reckoning in platform design: that control offered in theory but buried in practice is not really control at all.
- A three-second pause on the wrong video can silently redirect an entire feed for days, and most users never find their way back.
- Instagram's existing algorithm controls have sat largely unused, tucked inside settings few people think to visit in the moment they're frustrated.
- New gestures — a pull-down on the feed, a swipe up from Reels — would surface algorithm controls exactly when and where users notice something has gone wrong.
- A conversational chatbot interface is also being tested, letting users type specific content desires rather than selecting from preset categories.
- The core wager is behavioral: frictionless access will turn a feature people ignore into one they actually use.
We've all experienced it — a brief pause on one video and suddenly the entire feed pivots toward that single moment of curiosity. Instagram has spent the past year building tools to let users reclaim their recommendations, but those tools have remained buried in settings, easy to forget and hard to find.
Now the company is testing a fundamentally different philosophy. Rather than asking users to seek out controls, Instagram wants to surface them in the exact moments browsing happens. Adam Mosseri has previewed the approach: pull down on your main feed and algorithm controls appear; swipe up from a Reel and the same options surface. The existing Your Algorithm feature — which lets users view and edit the topics Instagram believes they care about — would become ambient rather than hidden, present wherever recommendations actually unfold.
Instagram is also experimenting with a more conversational layer: a chatbot-like interface where users can type specific requests — workout videos, recipe ideas, positive content — and receive suggested topics to add with a single tap. This shifts Your Algorithm from a static list into an active, expressive tool.
The underlying problem is algorithmic drift: accidental clicks quietly train the system in the wrong direction, and by the time the damage is visible, it's already compounded. These new features aim to enable immediate course correction, before the algorithm has time to reinforce a mistake.
What Instagram is really betting on is friction as the enemy of agency. Controls that require three taps and a settings detour go unused. Controls that appear in the moment, where the frustration is already felt, might actually change behavior. Whether users will engage more than they have with buried versions remains an open question — but the logic of meeting people where they are, rather than where the app has decided help should live, is difficult to argue with.
We've all been there: you pause on a single Reel for three seconds, and suddenly your entire feed transforms into an endless scroll of the same content. Instagram knows this happens. The company has spent the last year building tools to let users take back control of what they see, but those tools have lived buried in settings, accessible only if you remember they exist and know where to find them.
Now Instagram is testing a different approach. Instead of making users hunt through menus to adjust their recommendations, the company wants to put algorithm controls directly into the moments when you're actually browsing—right there in your feed, right there in your Reels. Adam Mosseri, who leads Instagram, has shown previews of what these shortcuts might look like: pull down on your main feed and Your Algorithm appears. Swipe up from a Reel and the same controls surface. The idea is simple but significant: let people fix their recommendations in real time, not hours or days later when they finally remember to dig into settings.
The feature Instagram calls Your Algorithm already exists. It lets users see which topics the app thinks they care about and edit that list directly. But it's been confined to a specific section of the app, one more thing buried in the interface. The new tests suggest Instagram wants to make this tool ambient—present wherever recommendations actually happen. Users would see buttons beneath individual Reels asking whether they want to see more videos like the one they're watching. A simple yes or no, right there in the moment, could reshape what comes next.
But Instagram is also experimenting with something more conversational. The company is testing a chatbot-like interface within Your Algorithm that lets users type what they want to see more of. Not just broad categories like "fitness" or "travel," but specific requests: positive content, recipe ideas, workout videos. Instagram would then suggest related topics the user can add to their feed with a tap. This transforms Your Algorithm from a passive list you edit into an active tool for reshaping your entire experience.
The problem Instagram is trying to solve is real. Algorithmic drift happens constantly—an accidental click, a moment of curiosity, a video that caught your eye for the wrong reasons—and the algorithm learns from it. By the time you notice your feed has gone off track, the damage is done. You've trained the system to show you things you don't actually want. These new features would let you correct course immediately, before the algorithm has time to lock in the wrong direction.
What's notable is that Instagram isn't just adding controls; it's making them frictionless. The company understands that most people won't bother adjusting their recommendations if it requires three taps and navigating to settings. But if the option appears right where you're already looking, in the moment when you realize something is wrong, you might actually use it. That's the bet Instagram is making with these tests. Whether users will engage with these tools more often than they do with the buried version remains to be seen, but the logic is sound: meet people where they are, not where you've decided they should go to find help.
Notable Quotes
Instagram wants algorithm controls closer to the places where recommendations actually appear— Adam Mosseri, Instagram head
The Hearth Conversation Another angle on the story
Why does Instagram need to make these controls easier to find? Doesn't the feature already exist?
It exists, but almost nobody uses it. It's like having a fire extinguisher in the basement—technically available, but by the time you remember it's there, the house is already burning. Instagram wants to move it to the kitchen.
So this is about friction. Making it one tap instead of five.
Exactly. The algorithm learns from everything you do—every pause, every click, even the accidental ones. By the time you realize your feed has drifted, you've already trained it wrong. These tests let you correct it while you're still in the moment.
What's the chatbot-like feature actually doing differently?
Instead of just showing you a list of topics you can add or remove, it lets you describe what you want in your own words. You type "positive content" or "fitness clips" and it suggests related topics. It's more active, more conversational. You're not just editing a list; you're having a conversation with the algorithm about what you want to see.
Does this solve the core problem, or does it just make the symptom easier to treat?
It treats the symptom faster, which matters. But the real problem is that the algorithm is too sensitive to noise—one accidental click shouldn't reshape your entire feed. These controls don't fix that. They just let you undo the damage more quickly.
Will people actually use these if they're easier to access?
That's the question Instagram is testing. Probably some will. But most people won't think about their algorithm until they're frustrated. The hope is that by making the controls visible and immediate, frustration becomes actionable instead of just annoying.