Instagram's 'Your Algorithm' tweak is the wrong answer to a real question

On 27 June 2026, TechCrunch reported that Instagram is testing new ways for users to customise the 'Your Algorithm' control — letting people tune more aggressively what surfaces in the algorithmic feed that, for most users, has become the de facto front door of the platform. The framing is straightforward: more knobs, more agency, more "choice." Read past the announcement and the framing collapses. What Meta is offering is a settings screen bolted onto a system whose inner workings remain opaque to the people whose attention funds it.

That gap — between the language of personalisation and the reality of a proprietary ranking engine — is the actual story. Every platform in the attention economy now competes on the rhetoric of user control, while the underlying machinery that decides what gets seen, and in what order, gets more elaborate, more opaque, and more commercially consequential.

What Meta is actually shipping

The Indian Express's report on the test describes a more granular set of interest sliders and content-type toggles attached to the existing 'Your Algorithm' feature. According to the report, users will be able to push harder on the signals they already feed the system — more of a topic, less of another, finer-grained adjustments to what shows up in the main feed and in Reels. The pitch is that the algorithm becomes something a user actively steers, rather than something that simply observes them.

That is a meaningful change in UX terms. It is not a meaningful change in governance terms. The ranking model — the actual decision procedure that turns a user's stated preferences into an ordered feed — is still Meta's property. The user gets to nudge the inputs; the user does not get to see the weights, the training corpus, the advertiser overlays, or the A/B experiments running on top.

The counter-narrative: why "more controls" is not a bad-faith answer

It is worth taking the counter-case seriously before dismissing it. There is a real, well-argued position — held by product teams inside the major platforms and by some independent researchers — that users do not actually want inspectability. They want outcomes: a feed that feels relevant, that doesn't make them feel awful, that surfaces what they care about. A settings panel that exposes raw model parameters would confuse most users, create new attack surfaces, and hand useful information to the people gaming the system.

That argument has force. A car dashboard does not expose the fuel-injection map. A search engine does not publish its ranking weights. The product question is genuinely hard. But the analogy breaks at one specific point: cars and search engines are not the primary funding mechanism for a multi-hundred-billion-dollar advertising duopoly whose business model depends on harvesting attention at industrial scale and selling it. The asymmetry matters.

What "customisation" actually changes

Three things, in declining order of importance:

First, it shifts the locus of blame. When a user sees something objectionable, the platform can now point to the slider set to "more of this" and the consent that produced it. That is not nothing — it is the difference between a system that quietly shapes you and a system that claims to do what you asked. But the shift is rhetorical as much as substantive, and product lawyers understand that distinction very well.

Second, it generates training data. Every adjustment a user makes is a labelled preference signal feeding back into the same model. The customisation layer is, among other things, a cheaper annotation pipeline than the ones Meta currently pays contractors to produce. Users are, in effect, being asked to label the training set that will be used to shape the next generation of the algorithm.

Third, it addresses a political problem. Platforms have spent three years on the back foot over child-safety concerns, mental-health scrutiny, and legislative pressure in the EU, the UK, and several US states. A visible control panel is the kind of artefact a regulator can be shown. It is also the kind of artefact a regulator will, fairly quickly, learn to look through.

The structural picture, in plain terms

The deeper pattern is not unique to Meta. Every major platform has, over the last five years, added user-facing levers of some kind — Twitter's algorithmic feed toggle, TikTok's "for you" reset, YouTube's "take a break" reminders, Spotify's mood controls. The proliferation is real. The cumulative effect on how these systems actually rank content is, by every independent measurement we have, modest.

What is changing faster is the business model underneath. The algorithmic feed is not a product feature; it is the inventory. The ranking model is the mechanism that converts a user's time into a metric an advertiser will pay for. User-facing controls that do not threaten that conversion are welcomed. User-facing controls that would meaningfully reduce it — a hard cap on daily impressions, a real "do not profile me" switch, an exportable personalisation model — are nowhere on the roadmap.

That is the test. Anyone serious about platform governance should be asking not "what sliders did they add?" but "what is the maximum the user can do to reduce the platform's monetisation of their attention, and does the platform meet it?" On that question, the current crop of customisation features does not move the needle.

Stakes

If the current direction holds, the algorithmic feed becomes more personalised and more deeply entangled with commercial ranking — while appearing, to regulators and users, to be more under user control than at any point in the platform's history. That gap is the durable political asset of the major platforms. It lets them absorb each wave of scrutiny — from the EU's Digital Services Act to US state-level child-safety laws — by shipping a settings panel and pointing at it.

The cost is borne elsewhere. Independent creators, smaller publishers, and anyone whose content does not fit the model's preferences find the door narrowing further. The user's relationship to their own attention becomes harder to reason about, because the explanatory layer is now actively shaped by the same incentives as the feed itself.

The honest version of the story is that Instagram's test is a small, real product improvement wrapped in a much larger claim about user agency. The improvement is welcome. The claim deserves scepticism.

Monexus treats platform-governance stories as a beat of their own, separate from the consumer-tech cycle. Where wires frame this kind of feature as a user-empowerment story, this publication reads it as a governance story — and asks, every time, what the user can actually do that the platform would prefer they couldn't.