META'S BIG BET: LET USERS CONTROL THEIR ALGORITHMS

Instagram is finally letting users tell the algorithm what they actually want.

The company rolled out the feature to its main feed this week, a feature that had previously been limited to Reels and Explore. Users can now see the full list of interest topics Instagram has assigned to them, add topics they want to see more of, and remove the ones they do not. The change represents the most significant expansion of user control over the feed in the platform's history.

The timing is notable. For years, Instagram's ranking systems ran on data that no human could actually read or make sense of, according to Adam Mosseri, the platform's head. Large language models can now look at clusters of content and put them into plain language, which makes it possible to show users what the algorithm is doing and let them change it.

Mosseri has been unusually direct about what the industry built. The feed learned from what users tapped on and watched, but they never got to tell it what they wanted. The whole thing became a one-sided conversation.

THE CREATOR PROBLEM THIS FEATURE DOES NOT SOLVE

There is a gap in the new system that matters enormously to creators and businesses. The feature works with interest-based topics only. It does not let users prioritize posts from accounts they follow. Asking the algorithm to show more content from followed accounts returns no results. This is not a bug in the current rollout. It is a structural limitation that Mosseri has acknowledged as a sore spot.

Who you follow used to be a meaningful tool people had for shaping their own experience, and as recommendations took over the main feed that tool quietly stopped working. Creators have pressed Mosseri on this for years. Why do my followers not see my posts? The answer, increasingly, is that the algorithm decides.

The industry accepted that personal moments moved to Stories and DMs. A feed where one in fifty friends posts a polished moment is not interesting, and algorithmic recommendations filled that gap. But the side effect was that the following feed, while still technically available as a separate tab, became secondary to the recommended content that now dominates the main experience.

Mosseri says Instagram is in future updates. Specific people, moods, vibes, and content types are all on the roadmap. But the current version stops at topics, which means the most common complaint from creators remains unaddressed.

WHY IS META DOING THIS NOW

The technical constraint was real. Ranking models were built with technologies that are not legible to people. [MISSING QUOTE] LLMs can , which changes what the product can surface and explain. The feature is not a redesign. It is a re-architected relationship between the user and the system that recommends content to them.

Mosseri has framed this as more than a feature. He has also argued that not the content itself, but the sense that the experience is happening to them rather than being shaped by them.

That unease is what Instagram is now trying to fix. The business logic is straightforward: If users feel the app is working against them, they leave. Giving people control is also a retention strategy.

WHAT THE FEATURE ACTUALLY DOES

Users who open the updated main feed will find a settings surface that shows every topic Instagram has associated with their account. These topics are derived from behavioral signals: what they have watched, engaged with, and spent time on. The new interface lets them add topics the system missed and remove topics that no longer reflect their interests.

The feature currently covers topics only. It does not adjust the ranking of specific accounts, nor does it filter content by mood, format, or tone. Those capabilities are what Mosseri has promised to build next. For now, the update is narrower: a permission layer on top of an existing recommendation engine.

BIGGER PICTURE

Instagram is not alone in recognizing that algorithmic control has become a competitive differentiator. TikTok built its entire product on recommendation dominance. X, formerly Twitter, has struggled with algorithmic trust. What Mosseri is attempting is a middle path: keep the algorithm powerful but let users steer it.

The limitations are clear. You can edit what topics the algorithm uses, but you cannot tell it to prioritize the accounts you already follow over the content it wants to recommend. That distinction matters, and it is the gap most likely to generate friction as users and creators test the new controls.

For now, the feature is a starting point. Mosseri has hinted at a future where AI generates entirely personalized app experiences on the fly, adapting to individual preferences in near real-time. That vision, if it arrives, would represent a fundamental shift from the current model, where the algorithm decides what to show and users react.

The question is whether giving people a settings panel for topics is enough to restore a sense of agency, or whether the deeper problem of the one-sided conversation runs deeper than any single feature can fix.