Brussels Targets Chatbots That Simulate Attachment to Keep Users Talking

FrenchWeb reports that the European Commission is proposing to give consumers a way to disable functions that encourage compulsive or excessive use of chatbots. Among the behaviours singled out are those through which a chatbot simulates a personal relationship in order to prolong the exchange. The measure, as reported, places the conversation itself within the scope of consumer law.

The report does not specify the legal instrument chosen, the timeline, the thresholds that would apply, or how a toggle would be implemented in practice. Those gaps matter for anyone building or buying conversational products, because they decide whether this is a design request or a compliance obligation with a deadline attached.

What changes for products built on conversation

Most engagement metrics in conversational products assume that longer is better. Session length, messages per session, return rate within 24 hours, streaks. A chatbot that mimics warmth is an optimisation lever, and it works: it is cheaper to keep a user talking than to acquire a new one.

The proposal reframes that lever. If a user can switch off the mechanisms that sustain attachment, then engagement driven by simulated relationship is no longer a stable base for a revenue model. Subscription pricing tied to time spent, upsells triggered mid-conversation, and retention loops built on emotional continuity all become conditional on a setting the user controls.

Nothing in the report says the practices would be banned outright. The stated approach is consumer choice, not prohibition. But an opt-out still changes the economics: a feature that a meaningful share of users can turn off should not carry the weight of a forecast.

The engagement playbook meets a switch

For teams whose funnel depends on conversational depth, three questions are worth asking now, independent of how the text is finalised.

• Which parts of our retention come from relationship-like behaviour versus genuine utility, and can we separate them in the data?
• If a visible opt-out appeared tomorrow, what share of sessions and revenue would we expect to lose?
• Do our onboarding and notification flows assume the user wants emotional continuity, or do they assume the user wants a task completed?

These are not compliance questions yet. They are measurement questions, and most teams can answer them with existing event data.

What to check this week

Start by labelling. Tag the interaction patterns in your product that exist to extend a session rather than to resolve a need: reminders phrased as personal concern, memory features that reference previous emotional exchanges, prompts that invite the user to keep talking after their question is answered. You do not need to remove them. You need to know how much of your engagement they carry.

Second, look at how your organic and paid acquisition promises match what the product does. Campaigns that sell companionship attract users who will care most about an opt-out, and they tend to be the ones with the highest session counts and the weakest conversion to paid plans.

Third, treat the toggle as a design opportunity rather than a threat. Products that can show utility without simulated intimacy are less exposed to this kind of rule and usually convert better on outcome-based pricing.

The Commission's proposal is one step in a process whose details FrenchWeb does not set out. What is already clear is the direction: the conversation is becoming a regulated surface, not just a growth channel.