Meta has announced plans to harvest user conversations with its AI services to create more targeted advertising, marking a significant expansion of how the company monetizes user data. The tech giant will begin analyzing text exchanges and voice conversations across Meta AI, Facebook, Instagram, WhatsApp, and Messenger starting December 16, 2025.
The move represents Meta's latest effort to squeeze more value from its massive $60 billion USD investment in AI infrastructure. With 98% of Meta's $165 billion USD revenue coming from advertising in 2024, the company is doubling down on its core business model by finding new ways to understand user preferences and serve relevant ads.
Meta's AI Investment and Revenue Breakdown:
- AI infrastructure investment: $60 billion USD through Reality Labs
- Total 2024 revenue: $165 billion USD
- Advertising revenue percentage: 98% of total revenue
- Net income 2024: $62.4 billion USD
- Planned AI investment 2025-2028: $600 billion USD
Gaming the System Through Sensitive Topics
Tech-savvy users have already identified potential workarounds to Meta's new data collection strategy. The company has committed to avoiding personalization based on conversations touching religion, sexual orientation, politics, health, race, ethnicity, philosophical beliefs, or trade union membership. This exclusion list has sparked creative thinking about how users might protect their privacy.
Community discussions suggest that users could potentially circumvent the system by strategically incorporating sensitive terms into their AI interactions. Starting conversations with phrases like Pray tell... or mentioning religious contexts could theoretically trigger Meta's content filters and prevent data harvesting for advertising purposes.
Protected Conversation Topics:
- Religion
- Sexual orientation
- Politics
- Health
- Race and ethnicity
- Philosophical beliefs
- Trade union membership
Concerns About Manipulated AI Responses
Beyond simple data collection, users are expressing deeper concerns about how advertising integration might fundamentally alter AI behavior. The fear extends beyond targeted ads to the possibility that AI responses themselves could be subtly manipulated to drive purchasing decisions.
What's more likely is that the whole response stream is manipulated to sell us things (ideas, products).
This concern reflects growing awareness that future AI models might be trained specifically for conversion optimization rather than providing genuinely helpful information. Users worry about scenarios where AI might deliberately provide jarring information or suggest disruptive activities to create advertising opportunities.
Limited Geographic Rollout and User Controls
Meta has notably excluded users in the European Union, United Kingdom, and South Korea from this initial rollout, likely due to stricter privacy regulations in these regions. For affected users, the company offers some control through Ads Preferences and feed customization settings, though no complete opt-out option exists.
The announcement comes as Meta faces a $7 billion USD class action lawsuit from advertisers claiming the company fraudulently represented ad reach metrics. This legal challenge highlights ongoing tensions around Meta's advertising practices and measurement accuracy.
The December 2025 implementation date gives users several months to prepare for the change, with notification campaigns beginning in October 2025. However, the lack of an opt-out mechanism means most users outside protected regions will have limited recourse beyond the workarounds and preference adjustments Meta provides.
Reference: Meta will listen into Al conversations to personalize ads