Instagram has launched a feature called 'Your Algorithm' that lets users inspect and adjust the topic signals driving their Reels and Explore recommendations — a meaningful expansion of user-facing controls over a system that, until now, operated largely as a black box from the viewer's perspective.

The feature was announced by Instagram in December 2025, and it surfaces, for the first time, the specific interest categories the platform's recommendation engine has associated with a given account. Users can review those topics and make direct modifications — pruning interests that no longer reflect their preferences, or amplifying ones that do — with the stated effect of reshaping the content mix served to them going forward.

What 'Your Algorithm' Actually Does

At its core, 'Your Algorithm' is a transparency and tuning layer sitting atop Instagram's existing content ranking infrastructure. The feature does not expose the model weights, engagement signals, or graph-traversal logic underneath — it surfaces a human-readable abstraction of inferred interest categories and hands the user a set of controls to modify them.

This is an important distinction. What users see is not the algorithm itself but a curated representation of the inputs the algorithm has derived from their behaviour. The actual ranking system — which factors in watch time, interaction velocity, content recency, creator affinity, and a range of other implicit signals — continues to run underneath. What changes is the weighting applied to explicit user-declared preferences versus the behavioural inference layer.

In practical terms: if Instagram's system has tagged an account as interested in, say, fitness content and home renovation, but the user's actual current interest has shifted, 'Your Algorithm' provides a surface to correct that drift without requiring the indirect, friction-heavy workaround of aggressively skipping or hiding content until the model recalibrates on its own.

The feature applies to both Reels — Instagram's short-form video surface, which has become the dominant discovery and engagement vector on the platform — and the Explore tab, which serves as the broader content discovery grid.

The Regulatory and Competitive Context

It would be a mistake to read 'Your Algorithm' as purely a product initiative divorced from the regulatory environment in which Meta operates. The EU's Digital Services Act (DSA), which came into full force for very large online platforms in 2023, explicitly requires that algorithmic recommendation systems offer users at least one option not based on profiling. Platforms must also provide meaningful information about the parameters driving personalised recommendations and allow users to modify or disable those parameters.

Instagram's new feature tracks closely with those DSA obligations, and while Meta has not framed this launch as a compliance measure, the timing and scope are consistent with the kind of user-facing controls the regulation demands.

Separately, TikTok — the competitive pressure that drove Instagram's Reels buildout in the first place — has offered its own version of interest-management controls for some time. YouTube has long provided explicit "Not interested" and "Don't recommend channel" mechanisms on Shorts and the main feed. Instagram arriving at a named, consolidated control surface in late 2025 means the major short-form video platforms now broadly converge on this UX pattern, even if the underlying implementations differ substantially.

Worth flagging here: the existence of a user-facing control surface does not automatically mean the controls are effective or that the feedback loop is tight. Platforms have an uneven track record on this. The gap between "you told us you don't want this topic" and "the recommendation engine has durably incorporated that signal" can be wide, and users have historically found that implicit behavioural signals — watch completion rates, re-watches, shares — tend to reassert themselves over explicit declared preferences within relatively short feedback cycles. Whether Instagram's implementation weights explicit corrections robustly enough to produce a noticeable and persistent shift in the feed is something that will only become clear through user experience over time, not from a product announcement.

Why This Matters for Platform Practitioners

For engineers, product managers, and data scientists working on recommendation systems or consumer platforms, 'Your Algorithm' is worth examining as an interface design problem as much as a machine learning one.

The challenge of reconciling revealed preference (what users actually engage with) with stated preference (what users say they want) is one of the genuinely hard problems in personalisation. Behavioural signals are noisy but abundant and temporally dense; explicit signals are sparse but semantically cleaner. Designing a system that lets explicit corrections propagate meaningfully without destabilising the recommendation quality that drives engagement metrics requires careful architecture — and it's an area where there is no settled best practice.

We have been around this loop before. When RSS readers gave way to algorithmic social feeds in the early 2010s, the promise was that machine-inferred relevance would outperform user-curated lists. It often did, in engagement terms — but it also progressively removed user agency. The decade that followed produced a sustained regulatory and public-interest backlash that has now, across multiple platforms, resulted in precisely the kind of re-exposed control layers that 'Your Algorithm' represents. The industry built the black box, discovered the consequences, and is now building the windowed panel on the side. That is not a cynical observation — it is, more or less, how consumer technology tends to self-correct, slowly, and usually with external pressure accelerating the timeline.

What Comes Next

Instagram has not disclosed whether 'Your Algorithm' will expand beyond topics to include creator-level or format-level controls — for instance, the ability to explicitly up-weight or down-weight specific accounts, or to prefer certain content lengths or formats within Reels. That would be a natural evolution of the feature set if the initial rollout demonstrates sufficient user engagement with the controls.

The more consequential open question, from a platform-accountability standpoint, is whether Meta will publish any data on how 'Your Algorithm' preferences are actually used — how many users engage with the feature, which categories are most commonly modified, and whether the corrections produce measurable divergence in subsequent content distributions. That kind of transparency would give researchers and regulators a clearer picture of whether user-facing controls are genuinely operative or are, in practice, marginal.

For now, the feature marks a legible step in the direction of user agency on a surface that reaches an enormous global audience. The gap between announcement and demonstrated impact remains to be measured.