Meta entered the dedicated AI assistant market tosay, releasing its standalone Meta AI application for iOS and Android users. This app directly challenges established players like ChatGPT, leveraging Meta’s own Llama 4 large language models announced earlier in April.
Beyond standard chatbot functions, Meta introduces a distinct social feature called the Discover feed and positions the app as the central control hub for its Ray-Ban Meta smart glasses, replacing the previous Meta View software.
The company frames the release as a “first step toward building a more personal AI.” Core interactions happen via text or voice, with the AI capable of pulling real-time web results and generating images.
Meta highlights voice as a key input, claiming Llama 4 enables more conversational and personalized responses. Users in the US, Canada, Australia, and New Zealand can also test an opt-in beta “full-duplex” voice mode, designed for more natural conversational flow with potential for overlapping speech, although this demo version currently operates without web connectivity and may exhibit glitches. A toggle allows users to set voice as the default interaction method upon opening the app.
A Social Twist on AI Interaction
Perhaps the most unique element is the app’s Discover feed, which integrated stream allows users to browse AI interactions—prompts and generated results—that other people, potentially including friends from linked Facebook and Instagram accounts, have explicitly chosen to share publicly.
These shared interactions can then be liked, commented on, or remixed for personal use. According to Meta’s VP of product, Connor Hayes, the feed intends to “demystify AI” and show “people what they can do with it.”
This social integration reflects a broader industry pattern where chatbot functionalities are merging with social networking paradigms, seen with X’s integration of Grok and reported ambitions from OpenAI. Meta stresses that sharing content to the Discover feed requires explicit user action.
The application leverages Meta’s existing ecosystem for personalization, a capability currently active in the US and Canada. The assistant can utilize information from Facebook and Instagram profiles connected via the Meta Accounts Center, alongside user engagement data like ‘likes’, to tailor its responses.
This aligns with Meta’s previously stated goal of “building toward a smarter, more personalized assistant”. Users retain the ability to tell the AI specific details to remember. While Meta already integrated Meta AI into search bars across its platforms, reportedly reaching “almost” one billion users that way, Hayes conceded that a standalone app offers the “most intuitive way to interact with an AI assistant.”
Connecting Hardware and Software Under Llama 4
The Meta AI app now serves as the primary interface for the company’s Ray-Ban Meta smart glasses, absorbing the functions of the retired Meta View app. Settings and media for existing glasses users are automatically carried over to a new “Devices” tab.
This integration facilitates a degree of cross-device continuity: conversations started via voice on the glasses can be viewed in the history tab of the app or the meta.ai web interface. Interaction can also flow between the app and the web version. However, Meta notes a current limitation: “[you] cannot start in the app or on the web and pick up where you left off on your glasses.”
This hardware link follows an April 24th update that rolled out expanded AI features for the glasses, including global availability for live translation and wider access to the “Look and ask” visual AI, all now managed through the central Meta AI app.
The Llama 4 models underpinning the app utilize techniques like Mixture-of-Experts (MoE) architecture—activating only necessary model parts per task—for efficiency, and feature native handling of multiple data types like text and images. Developing these models has proven resource-intensive; reports surfaced that Meta had approached Microsoft and Amazon seeking co-funding for Llama development.
Lingering questions also surround Llama’s training data, which is the subject of ongoing lawsuits alleging infringement based on the use of copyrighted books. Court documents revealed internal messages, with one engineer reportedly writing, “Torrenting from a [Meta-owned] corporate laptop doesn’t feel right.”
Navigating Market Dynamics and Policy Debates
Meta’s app launch comes as the company actively shapes its AI’s behavior and market position. It has stated a goal for Llama 4 involves countering perceived political leanings, saying “It’s well-known that all leading LLMs have had issues with bias—specifically, they historically have leaned left when it comes to debated political and social topics… This is due to the types of training data available on the internet.”
This technical tuning occurred alongside significant platform policy shifts, including the January 2025 termination of Meta’s US third-party fact-checking program. Global Policy Chief Joel Kaplan attributed that change partly to moderation errors, noting internal reviews suggested “One to two out of every 10 of these actions may have been mistakes.”
Competitively, Meta has acted to keep users within its AI ecosystem, notably by blocking Apple’s integrated Apple Intelligence features from functioning inside Meta’s iOS apps since late 2024. Furthermore, global feature deployment remains uneven; the integrated chatbot’s European debut in March 2025 came with restrictions on image generation and personalization due to GDPR privacy rules, suggesting potential regional variations for the new standalone app’s capabilities as well.
