SourceTechCrunch AI•Stalecollected in 60m
Meta rolls out AI Mode across Facebook platforms
Learn how Meta is leveraging its massive social data graph to power new AI features and boost platform engagement.
30-Second TL;DR
What Changed
Integrates public information across Meta's various platforms
Why It Matters
This feature signals Meta's aggressive push to leverage its massive social graph for AI training and personalized experiences. It may set a new standard for how social platforms utilize public user data for generative AI.
What To Do Next
Review Meta's updated data usage policies to understand how your public content might be ingested into their AI models.
Who should care:Developers & AI Engineers
Key Points
- •Integrates public information across Meta's various platforms
- •Designed to increase user engagement on Facebook
- •Part of Meta's broader strategy to catch up in the AI race
Key numbers$10 billion10%
Deep Insight
Background and context from public sources — not the original article. 13 sources cited.
Enhanced Key Takeaways
- •Meta's 'AI Mode' transforms Facebook's traditional search into a conversational experience, providing answers drawn from public content across Groups and Reels, rather than a generic list of results.
- •The new 'AI Mode' is powered by Muse Spark, an AI model that Meta debuted in April and was developed by Meta Superintelligence Labs, which is led by former Scale AI CEO Alexandr Wang.
- •Beyond search, the rollout introduces new AI-assisted photo and video editing capabilities, including collage cutout templates, video transition effects, and photo presets that allow users to change clothing, hair, and accessories.
- •Analysts project that if the new AI search tool retains 1 billion users and monetizes just 10% of daily queries, it could generate over $10 billion in annual revenue for Meta.
- •Meta AI, which underpins 'AI Mode', is integrated across Meta's ecosystem, including Facebook, Instagram, WhatsApp, Messenger, Meta smart glasses, and a standalone app, and runs on the Llama model family.
Competitor Analysis
Meta AI
- Key Features
- Integrated across Meta apps (FB, IG, WhatsApp, Messenger, smart glasses), conversational Q&A, content creation, image generation/editing, personalization, voice interaction.
- Pricing
- Free (paid tier for premium features exists but not aggressively marketed).
- Benchmarks/Focus
- Casual use, quick answers, basic creative tasks, scale, openness.
ChatGPT
- Key Features
- In-depth writing, coding, complex data analysis, versatile, strong ecosystem.
- Pricing
- Free tier (with ads); Plus: $20/month.
- Benchmarks/Focus
- Versatility, ecosystem, broad reasoning.
Claude by Anthropic
- Key Features
- Advanced professional reasoning, long-document analysis, coding, safety, long-term memory, tool use.
- Pricing
- Free; Pro: $17/month; Max: from $100/month.
- Benchmarks/Focus
- Advanced reasoning, coding, safety, nuanced tasks.
Google Gemini
- Key Features
- Real-time research, deep integration with Google apps (Workspace, Android, Search), native multimodal reasoning.
- Pricing
- Free (open-source); infrastructure hosting costs vary.
- Benchmarks/Focus
- Real-time research, ecosystem integration, native multimodal reasoning.
Grok by xAI
- Key Features
- Real-time social data analysis (X firehose), rapid reasoning, unfiltered analysis.
- Pricing
- Integrated with X Premium ($8-$16/month); API usage varies.
- Benchmarks/Focus
- Real-time social data, breaking news, sentiment tracking.
Microsoft Copilot
- Key Features
- Enterprise integration, embedded in Windows and Office, automates tasks within Microsoft 365.
- Pricing
- Around $30/user/month for Office integration.
- Benchmarks/Focus
- Enterprise integration, Office/Windows workflows.
Apple Intelligence
- Key Features
- On-device processing, deep integration with Apple devices (iPhone, iPad, Mac, Vision Pro), enhanced Siri.
- Pricing
- Free with recent Apple devices.
- Benchmarks/Focus
- Privacy, on-device processing.
| Competitor | Key Features | Pricing | Benchmarks/Focus |
|---|---|---|---|
| Meta AI | Integrated across Meta apps (FB, IG, WhatsApp, Messenger, smart glasses), conversational Q&A, content creation, image generation/editing, personalization, voice interaction. | Free (paid tier for premium features exists but not aggressively marketed). | Casual use, quick answers, basic creative tasks, scale, openness. |
| ChatGPT | In-depth writing, coding, complex data analysis, versatile, strong ecosystem. | Free tier (with ads); Plus: $20/month. | Versatility, ecosystem, broad reasoning. |
| Claude by Anthropic | Advanced professional reasoning, long-document analysis, coding, safety, long-term memory, tool use. | Free; Pro: $17/month; Max: from $100/month. | Advanced reasoning, coding, safety, nuanced tasks. |
| Google Gemini | Real-time research, deep integration with Google apps (Workspace, Android, Search), native multimodal reasoning. | Free (open-source); infrastructure hosting costs vary. | Real-time research, ecosystem integration, native multimodal reasoning. |
| Grok by xAI | Real-time social data analysis (X firehose), rapid reasoning, unfiltered analysis. | Integrated with X Premium ($8-$16/month); API usage varies. | Real-time social data, breaking news, sentiment tracking. |
| Microsoft Copilot | Enterprise integration, embedded in Windows and Office, automates tasks within Microsoft 365. | Around $30/user/month for Office integration. | Enterprise integration, Office/Windows workflows. |
| Apple Intelligence | On-device processing, deep integration with Apple devices (iPhone, iPad, Mac, Vision Pro), enhanced Siri. | Free with recent Apple devices. | Privacy, on-device processing. |
Technical Deep Dive
- 'AI Mode' is powered by Muse Spark, an AI model developed by Meta Superintelligence Labs.
- Meta AI runs on the Llama model family, including Llama 2, Llama 3, and newer Llama 4 based systems.
- Llama 4 models are natively multimodal, designed for both text and image understanding, and utilize a mixture-of-experts (MoE) architecture.
- The Llama 4 series includes models like Llama 4 Scout (17 billion parameters with 16 experts) and Llama 4 Maverick (17 billion parameters with 128 experts).
- Meta has confirmed that publicly shared posts, comments, and images from Facebook and Instagram were used in the training of Llama 3 and 4, while private user data is explicitly excluded.
- The underlying Meta AI platform features a Natural Language Processing (NLP) engine for text analysis, generation, and comprehension, built on advanced neural network architectures.
- It also incorporates computer vision capabilities for tasks such as image recognition, object detection, and segmentation.
- Meta AI employs continuous machine learning, adapting and improving through user interactions and data, and utilizes meta-learning to quickly adjust to new scenarios.
Future ImplicationsAI analysis grounded in cited sources
Meta's 'AI Mode' will intensify competition in the search engine market.
By positioning AI as a direct replacement for Facebook's search bar, Meta directly challenges established search providers like Google.
User trust and the spread of misinformation will become critical challenges for Meta's AI Mode.
The AI draws conclusions from public user opinions and posts, raising concerns about the accuracy and reliability of information, similar to issues faced by other AI search systems.
Meta will likely introduce paid subscription tiers for advanced AI capabilities.
The company has already announced global subscription plans for its apps, and future plans are expected to include separate paid tiers for more complex AI features.
Timeline
2013
Facebook Artificial Intelligence Research (FAIR) founded.
2017
FAIR released PyTorch, an open-source machine learning framework.
2021
Facebook, Inc. rebranded to Meta Platforms Inc., and FAIR renamed Meta AI.
2023-02
Meta released the LLaMA language model.
2025-06
Mark Zuckerberg initiated the formation of Meta Superintelligence Labs.
2026-06-15
Meta rolls out 'AI Mode' across Facebook platforms, powered by Muse Spark.
- 2013Facebook Artificial Intelligence Research (FAIR) founded.
- 2017FAIR released PyTorch, an open-source machine learning framework.
- 2021Facebook, Inc. rebranded to Meta Platforms Inc., and FAIR renamed Meta AI.
- 2023-02Meta released the LLaMA language model.
- 2025-06Mark Zuckerberg initiated the formation of Meta Superintelligence Labs.
- 2026-06-15Meta rolls out 'AI Mode' across Facebook platforms, powered by Muse Spark.
Sources (13)
Factual claims are grounded in the sources below. Forward-looking analysis is AI-generated interpretation.
1forbes.comvertexaisearch.cloud.google.com2zamin.uzvertexaisearch.cloud.google.com3rollingout.comvertexaisearch.cloud.google.com4blockchain-council.orgvertexaisearch.cloud.google.com5igmguru.comvertexaisearch.cloud.google.com6builtin.comvertexaisearch.cloud.google.com7goodday.workvertexaisearch.cloud.google.com8dupple.comvertexaisearch.cloud.google.com9youtube.comvertexaisearch.cloud.google.com10huggingface.covertexaisearch.cloud.google.com11medium.comvertexaisearch.cloud.google.com12easyweb-agency.frvertexaisearch.cloud.google.com13adsmurai.comvertexaisearch.cloud.google.com
Weekly AI Recap
Read this week's curated digest of top AI events →
Related Updates
AI-curated news aggregator. All content rights belong to original publishers.
Original source: TechCrunch AI ↗
This is a summary, not the original. Read the source, or get the weekly briefing.
The weekly digest
One email a week. Unsubscribe anytime.