Surveillance Capitalism: Data as the New Oil

The product was never just the product

Surveillance capitalism is often reduced to targeted advertising, but that framing is too small. The system this file examines works by capturing lived behaviour, converting it into data, and using that data to predict and influence what people do next. The business value comes from the loop: collect, infer, rank, steer, repeat.

Capture Clicks, pauses, routes, habits, and location traces become measurable signals. Predict Behaviour becomes more valuable when it can be forecast, clustered, and scored. Steer Ranking, prompts, defaults, and notifications turn prediction into intervention.

Core point: surveillance capitalism is not defined by collection alone. It is defined by the conversion of behaviour into prediction products and intervention capacity.

How digital convenience became behavioural infrastructure

Modern platforms observe users at a granularity older media never could. Search reveals intent, social feeds generate reaction signals, mobile devices add location and routine, and retail systems expose hesitation and repeat behaviour. The interface becomes a sensor.

That data is not merely stored. It is processed into commercial foresight, targeted delivery, and product tuning. The result is a business model that benefits from turning ordinary participation into a behavioural exhaust stream.

The recent FTC staff report on social media and video streaming services reinforces this structure. It says the companies engaged in vast surveillance to monetize personal information, retained large amounts of data, and fed users’ and non-users’ information into algorithms, analytics, and AI with little meaningful opt-out.

Behaviour becomes the hidden raw material

The first move is extraction. A service appears free or convenient, but the hidden transaction is behavioural collection. The FTC’s 2024 report is important because it confirms, in official language, that this is not an accidental byproduct. It is built into the platform economy’s core incentives.

The same report says many firms relied on targeted advertising, could retain troves of data, and often had weak approaches to minimization and deletion. That makes convenience strategically useful: the smoother the experience, the more time and behaviour the system can measure.

Operational reading: the user thinks they are consuming a service, but the system is also consuming the user.

Raw traces become commercial foresight

Collection alone is not enough. The next step is interpretation: clustering, ranking, recommendation, and ad delivery. Alphabet’s 2024 annual report says more than 75% of its total revenues came from online advertising, and Meta’s 2024 results emphasize reliance on advertising revenue and data signals. Those filings show how central predictive advertising remains to platform economics.

In this model, a click is not just a click. It can indicate preference, timing, and susceptibility. A pause can indicate uncertainty or interest. A route can reveal routine or vulnerability. Behaviour becomes valuable because it can be modelled probabilistically and sold as foresight.

Feedback-loop research supports that dynamic. Recommender systems do not merely observe preference; they can influence the behaviour that later retrains them, which makes the line between measurement and making harder to draw.

Recommendation turns into soft behavioural control

Once a platform can estimate what someone is likely to do, the next incentive is to shape what happens next. That can happen through feed ranking, notifications, defaults, prompts, friction design, and visibility management rather than overt command.

The OECD describes dark commercial patterns as digital practices that steer, deceive, coerce, or manipulate consumers through the way choices are presented. That gives a policy-language name to a familiar interface tactic: shape the field, then claim the user freely chose inside it.

This does not mean every recommendation is sinister. It means the system is rewarded for producing outcomes, not neutrality.

• What changed since the original assessment? The FTC published its 2024 staff report and updated press release, but they reinforce rather than overturn the original thesis.
• What stayed current? Alphabet still reports online advertising as its dominant revenue source, and Meta still flags reliance on advertising and data signals.
• What is still debated? The evidence supports extraction and steering incentives, but not a universal claim of total social control.

The system improves by watching how people react to it

One of the most powerful features of these systems is that they improve through use. Each intervention generates new data, and each response becomes training material. Recommenders and ranking engines therefore do not simply measure users; they participate in shaping the environment in which user preferences form.

This creates a feedback loop: the system pushes content, users adapt to what is surfaced, and the resulting behaviour trains the next version of the system. The result is a model that can become more capable at predicting and influencing behaviour without ever needing explicit coercion.

Commercial logic rarely stays in the commercial box

If surveillance capitalism were only about ads, it would still be significant. But the logic spills outward into retail, workplace analytics, insurance, education, connected devices, and public-sector tools. The broader pattern is the same: rank people, predict response, and optimise the desired outcome.

That spillover matters because it turns a market model into a governance substrate. Systems built to optimise response in commerce can be repurposed for institutional management, whether through contracts, partnerships, procurement, or regulation.

The exchange is not as clean as platforms imply

Supporters of the current model often point to consent: users agree to terms and receive value in return. But the FTC report undercuts that framing by noting little or no meaningful opt-out in how data was used by algorithms, analytics, or AI, and by describing weak minimization and retention practices.

That matters because consent loses force when the user cannot realistically understand the scale or downstream uses of the data being collected. In a world where everyday participation depends on data-intensive systems, the click-to-agree model tells only part of the story.