TL;DR

The Information Commissioner’s Office published a 51-page report on 13 November on how large language models and agentic AI are reshaping online advertising. It sets no new legal requirements. Instead, it explains how PECR, the UK GDPR and the Data Protection Act 2018 apply when firms let AI plan campaigns, pick audiences or serve ads inside chatbots. Among its warnings is what a conversation can reveal about the person having it.

What the ICO looked at

The regulator calls the current phase “AdTech 3.0”. By that it means a direction of travel, not a product: people talking to advertising systems in natural language, and firms handing tasks to software that can act with some independence. The report covers five settings: ads shown beside chatbot answers, industry protocols such as the Ad Context Protocol that let AI agents talk to ad platforms, walled gardens, retail product discovery, and streaming to connected TV sets.

Its chatbot example is simple. Someone asks for holiday advice, the model answers, and a separate ad system uses the question’s travel theme to pick an ad. The ICO says it examined cases where the model passed such signals, like the apparent intent of a request, to the advertising system.

The research behind the report was commissioned externally and supplemented by talks with advertisers, publishers, agencies and technology suppliers. Shopping agents that buy on a consumer’s behalf are included, though the ICO says they were not yet an established part of UK advertising when the research was done.

Where the risk sits

Profiling is not new to adtech; the ICO’s 2019 real-time bidding report already raised it. What changes, the report argues, is depth. A single exchange with a chatbot can expose someone’s immediate circumstances or intentions, and a prompt might mention a health condition, a religious belief or a political view outright. The system may also infer such traits when the person never states them.

The report sets out seven themes. Firms must be able to show accountability, give people real control, handle sensitive inferences with care and keep advertising transparent and fair. They should also check suppliers along the chain, collect no more data than they need, and assess the risks.

Looking Forward

The findings will feed planned ICO guidance covering agentic AI and automated decision-making and its AI and ADM code of practice. The ICO also plans to keep tracking adoption, to talk to industry, and to coordinate with bodies whose remits overlap, from the ad standards regime to competition and consumer law. Firms building these products can apply to its Regulatory Sandbox. In our view, the message to UK marketers is that a new interface does not reset the compliance clock: consent and transparency questions from the real-time bidding era carry straight into conversational ads.