Meta launches 'AI Mode' on Facebook using public data from its platforms

What happened?

On June 15, 2026, Meta announced the rollout of 'AI Mode' on Facebook, a new feature that integrates a conversational assistant capable of answering questions using public information from its platforms: Facebook, Instagram, Threads, and WhatsApp Business. According to TechCrunch, the feature is being gradually implemented in the United States and will be available in the Facebook search bar. This move comes amid intense competition in artificial intelligence, where Meta seeks to regain ground against OpenAI, Google, and Anthropic. The company has invested billions in AI infrastructure and has launched language models like Llama 3, but until now lacked a conversational assistant integrated into its main applications.

How does it work?

'AI Mode' allows users to ask about topics such as local events, restaurant recommendations, or fashion trends, and gets answers based on public posts, reviews, and business profiles within Meta's ecosystem. For example, a user might ask: 'What are the best Italian restaurants near me according to Instagram reviews?' The AI processes natural language and extracts information from indexed public content. Meta has confirmed that the feature uses its own language model, trained on public data from its platforms, and does not access private messages or non-public account content. According to leaked internal documents, the system employs retrieval-augmented generation (RAG) techniques to combine language generation with real-time search of posts and profiles. Additionally, the AI can generate responses with direct links to sources, which increases transparency but also opens the door to potential biases toward popular or sponsored content.

Why is it important?

This launch represents a strategic move by Meta to compete in the AI assistant market, dominated by OpenAI, Google, and Anthropic. By leveraging its vast repository of public social media data, Meta offers a differentiating value: contextualized answers with content generated by real users. However, it also intensifies the debate on privacy, as the feature uses data that many users consider semi-private, even if public. The European Commission has already expressed concern about using social media data to train AI models without explicit consent, and this launch is likely to trigger new regulatory investigations. In the United States, the FTC has shown interest in how platforms handle user data for AI purposes. Additionally, Meta has acknowledged that the feature may display sponsored ads in responses, raising questions about the impartiality of recommendations.

Potential consequences

For users, 'AI Mode' could improve the search experience within Facebook, but could also increase exposure to targeted advertising and data collection. For content creators and businesses, the feature could drive traffic to their profiles if the AI recommends them. However, there is a risk that the AI may bias responses toward popular or sponsored content, harming small businesses or emerging creators. European regulators have already expressed concern about using public data without explicit consent, which could lead to investigations in the EU. In the competitive landscape, this move could pressure Google and OpenAI to integrate social media data into their assistants, though both companies face their own privacy challenges. Additionally, Meta could use 'AI Mode' to increase dwell time on Facebook, which has declined among younger users in favor of TikTok and other platforms.

Comparison with other assistants

Unlike ChatGPT, which relies on general web data, or Google Gemini, which indexes the web, 'AI Mode' is limited to Meta's ecosystem. This gives it a more niche focus but also limits its usefulness for general queries. For example, asking about the capital of France would not make sense in 'AI Mode' unless there is relevant public content on Meta's platforms. In contrast, for questions about local trends or product reviews, Meta's AI could offer more up-to-date and contextualized answers than its competitors. Meta claims that the AI does not train on private data, only public information, but the line between public and private is blurry on social media. A 2025 Stanford University study showed that 40% of Facebook users believe their 'public' posts are only visible to friends, indicating a lack of understanding about data indexing. Additionally, unlike assistants like Siri or Alexa, 'AI Mode' cannot perform actions (such as sending messages or making purchases), limiting its functionality to information retrieval.

What readers should know

• The feature is currently available only in the US, with no date for other countries. Meta has indicated that international expansion will depend on local regulatory assessments.
• Responses may include links to profiles, posts, or ads. Meta has confirmed that advertisers can opt to have their ads appear in AI responses.
• Users can opt out of having their public content indexed by adjusting privacy settings. However, this adjustment does not affect already indexed content, according to Meta's policy.
• Meta has stated that the AI is not trained on private messages or non-public business account data. However, the company has not specified how it will audit compliance with this policy.
• Content creators can request that their posts not be included in AI responses through a new tool in page settings.

In conclusion, 'AI Mode' is another step in the AI race, but also a reminder that our public data is a valuable resource that companies are willing to exploit. The feature could redefine search on social media, but also poses ethical and regulatory challenges that should not be underestimated. Users should be aware that every public post can feed AI responses, and businesses must prepare for a new channel of visibility and competition.