Australia wants to put the social graph back at the heart of social media

The Australian government now wants to require social media platforms to offer users over the age of 16 a feed limited to accounts they have chosen to follow. Presented as a protective measure, the “My Feed, My Way” initiative reaches into the core of TikTok, Instagram and Facebook’s model: personalised recommendations that organise attention, distribute audiences and generate advertising revenue.

Following an account no longer means seeing what it publishes. Conversely, never having heard of a creator does not prevent someone from watching ten of their videos in a row. Social media still relies on follows, friends and contacts, but its feeds increasingly show users what platforms believe will hold their attention rather than what they have chosen to see.

The Australian government wants to make that shift reversible. Unveiled on 8 September by Prime Minister Anthony Albanese and Communications Minister Anika Wells, “My Feed, My Way” would require platforms to ask users to choose between a feed that includes personalised content recommended by an algorithm and, as in the early days of social media, one limited to friends and creators they have chosen to follow.

“This is not about giving government control, it is about giving people control,” Australian Prime Minister Anthony Albanese said. The formulation neatly places the debate on the terrain of individual freedom. Yet the proposal goes far beyond adding another setting to an app: it represents Australian intervention in the way platforms design their products, distribute audiences and create value.

The social graph has given way to the interest graph

The first major social networks were built around a promise: allowing people to find posts from their friends, professional contacts or accounts they had decided to follow.

Those feeds were not necessarily chronological, let alone free of algorithms. Facebook was already ranking posts, Twitter was ordering conversations and Instagram was deciding which photos appeared first. But the initial universe was still largely assembled by the user. People chose the accounts; the platform then organised their content.

TikTok popularised a different architecture. Its “For You” feed does not wait for a newcomer to patiently build a network. It observes interactions, measures the time spent on each video and gradually refines its recommendations.

The audience growth and engagement generated by this mechanism persuaded other platforms to change their display rules. Instagram, Facebook and YouTube progressively increased the share of posts coming from accounts users do not follow. The product no longer merely organises a social graph made up of explicit relationships. It builds an interest graph from observed behaviour: watching, stopping, sharing, commenting and returning.

This mechanism also enables social platforms to provide an almost inexhaustible stream. Once friends have finished posting their photos and followed media outlets have published their latest articles, the engine can still find another video, and then another. A social network ceases to be merely a place where relationships circulate. It becomes a personalised media service in which every user receives their own programming.

Two social networks will have to coexist inside the same app

“My Feed, My Way” seeks to restore a fork between these two experiences. On one side is a feed personalised by the platform, capable of including content from unknown authors. On the other is a feed confined to accounts the user has actively chosen to follow.

The originality of the Australian proposal does not lie entirely in the technical existence of these options, since several platforms already offer alternative feeds or personalisation settings. It comes from the obligation to present this choice explicitly to all affected users and to respect the outcome.

The shorthand of an “algorithm-free” feed should nevertheless be avoided. Even when limited to followed accounts, a feed may continue to use algorithms to rank posts, remove duplicates, filter spam, apply moderation rules or insert advertising.

The Australian proposal therefore does not provide a complete exit from algorithmic systems. It aims more specifically to let users refuse personalised content selected by a platform from accounts they have not chosen to follow.

Two words could considerably narrow the law’s reach

The Australian government specifies that the choice will apply to the “default feed”. The effectiveness of the measure will depend on how those two words are defined, because platforms now contain multiple overlapping surfaces: the home feed, short-form videos, discovery tabs, search engines, notifications, account suggestions and content displayed after a user opens a post. A platform could theoretically provide a compliant main feed while retaining personalised recommendations elsewhere in its app.

A user who has rejected recommendations in their feed could still encounter them in Reels, Shorts, Explore or their equivalents. They could also receive notifications designed to bring them back to content selected by the platform.

The role of advertising will also need to be clarified. Will a feed reserved for followed accounts still be allowed to include ads targeted using behavioural data? Will sponsored content count as a recommendation? Will platforms be able to suggest accounts between posts from chosen sources?

The proposal must therefore define not only what users can refuse, but also how far through the product that refusal applies.

The real choice will be made in the notification design

Another key issue is the selection process itself. If recommendations remain active until users take steps to disable them, inertia will favour the current model. If they can only begin after explicit consent, the effect on behaviour could be much greater. A third possibility would be to withhold either option until the user actively selects one.

The legislation will also need to govern the interface itself. Platforms know how to present two apparently equivalent choices while steering users strongly towards one of them: a large coloured button to retain “a personalised experience”, a discreet link to receive a “limited” feed, a few additional screens and a warning about the content the user risks missing.

Communications Minister Anika Wells says users will be able to change their minds whenever they wish and that platforms will have to respect their decision.

For its part, the regulator will need to verify that the choice persists, carries over between devices and is not undermined by repeated prompts encouraging users to abandon it. Without those safeguards, the reform risks creating another digital ritual of consent: plenty of windows, a few buttons and ultimately the same option as before.

Feed choice is only one part of a broader regulation of design

“My Feed, My Way” forms part of a wider Digital Duty of Care proposal. It would cover not only social media, but also search engines, online games, apps, messaging services and artificial intelligence chatbots.

For users under 18, companies would be expected to take measures against content promoting eating disorders, pornography, harassment, misogyny, the glorification of crime and certain dangerous behaviours. They would also have to address mechanisms that could encourage compulsive use or damage self-esteem.

This is where the proposal changes in nature. Platform regulation has long focused on individual posts: flag a piece of content, determine whether it is lawful, then potentially request its removal. The Duty of Care addresses the conditions in which risk is produced and amplified.

A video may not present the same danger when it appears once as it does when followed by 50 similar items. A reward mechanism may seem harmless in isolation but become problematic when it leads a child to return dozens of times a day. The government is therefore asking companies to examine the effects of the system as a whole: content, repetition, recommendation, interface and frequency of use.

Platforms will have to identify foreseeable risks, document the measures they take and verify their effectiveness over time. For children under 16, the proposal also provides for features such as algorithmic personalisation and infinite scrolling to be switched off, alongside Australia’s existing ban on access to social media for that age group.

Breaches could lead to substantial fines. The eSafety Commissioner would be responsible for enforcement and would also receive strengthened powers to act against certain nudification services and illegal content.

Recommendations also shape the creator economy

The measure does not concern only the relationship between platforms and their users. It could redistribute audiences among creators, media outlets and brands.

Accounts that have already built substantial communities could regain more regular access to their followers. The value of the “follow” button, progressively diluted by the arrival of recommended content, would be partly restored.

The outcome could be less favourable for new creators. Today, the interest graph enables a video published by an unknown account to reach a considerable audience very quickly. If a significant share of users chooses to remain within the perimeter of their subscriptions, that discovery capacity will decline.

The reform could therefore produce two opposing effects: reducing platforms’ discretionary power over distribution while strengthening the position of those who already possess an audience. New entrants could then have to devote more resources to advertising, partnerships or maintaining a presence across several services.

For Meta, TikTok and YouTube, the risk will depend less on whether the choice exists than on how many users exercise it. Limited uptake would change little. Widespread rejection of recommendations could reduce the amount of content consumed, time spent and some advertising impressions.

The Australian government is not explicitly seeking to reduce platform revenue. It is nevertheless introducing an option that conflicts directly with their economic interests by allowing users to choose an experience that may be less abundant than one optimised to prolong their presence.

Australia turns a technical requirement into a political choice

The European Union has already opened a comparable path. Since the Digital Services Act came into force, very large platforms have been required to offer at least one recommendation option that does not rely on profiling. Implementation has nevertheless proved difficult, with regulators notably arguing that some platforms had not made opting out sufficiently straightforward.

The Australian approach has the advantage of clarity. The difference between a system based on profiling and one that is not remains technical. The distinction between “recommended content” and “accounts you have chosen to follow” can be understood in seconds.

Australia is thus extending a doctrine built in stages: a social media ban for children under 16, powers entrusted to the eSafety Commissioner, a preventive duty to protect users and now a choice over recommendations. The country is no longer merely asking platforms to remove the worst material. It is beginning to regulate the mechanisms that determine what becomes visible.

From Canberra to London, Paris and Brussels

Australia did not invent platform regulation. The European Union and the United Kingdom already had the Digital Services Act and the Online Safety Act. It nevertheless opened a new chapter by becoming, in December 2025, the first country to require certain large platforms to prevent children under 16 from holding an account. That decision turned a proposal previously discussed with caution into enforceable public policy.

In June 2026, the British government announced that it would explicitly adopt the Australian model to ban social media for children under 16 from 2027. London nevertheless intends to go further on certain features, including live streaming, communication with strangers, mechanisms that prolong scrolling and default restrictions for 16- and 17-year-olds. The regime will have to fit within the Online Safety Act and be enforced by Ofcom, using age-verification mechanisms adapted to British law.

France moved in the same direction, setting the threshold at 15. But the circulation of a political idea does not guarantee that its legal text can travel with it. France’s law was struck down by the Constitutional Council in the summer of 2026, before it entered into force. Emmanuel Macron has since asked Ursula von der Leyen to support a harmonised ban at European level, while announcing a new version of the national measure.

That difficulty captures the next stage. The Australian principle can travel; its implementation must be rebuilt within each legal order. In the United Kingdom, it must be aligned with Ofcom’s powers and the existing obligations of the Online Safety Act. In France, a national measure must respect the Constitution while avoiding encroachment on the framework harmonised by the Digital Services Act. At European level, it must also satisfy requirements of necessity and proportionality, protect freedom of expression and children’s right to access information, and provide age verification compatible with privacy protections. The European Union Agency for Fundamental Rights has stressed that any restriction must balance child protection with freedom of expression, participation, non-discrimination and data protection.

Australia therefore offers less a ready-made law than a starting point. The United Kingdom is adopting its age threshold while broadening the range of restricted features. France is seeking to adapt the model to its constitutional law and is now pushing Brussels to produce a common response. The European Union, for its part, will have to prevent 27 competing national regimes from fragmenting the obligations imposed on platforms.

The international ripple effect of “My Feed, My Way” remains to be demonstrated, even if that of the social media ban for children is already established. If the choice between the social graph and personalised recommendations becomes a genuinely deployed feature in Australia, other governments will have the same argument at their disposal: what platforms have managed to build for Canberra can no longer be presented as technically impossible in London, Paris or Brussels.