Meta AI has landed in Threads direct messages, rolling out worldwide since Monday. For the first time, Threads users can chat one on one with Meta’s assistant, hand it a post, an image or a video, and dig into a topic without leaving the app. It is one more brick in the company’s push to keep its users inside its own walls.
Key Takeaways
- Meta AI is rolling out globally inside Threads DMs from July 27, a first for the network.
- Users can share posts, images, links and videos with the assistant and ask follow-up questions in private.
- It is the fourth Meta surface to host the assistant after Facebook, Instagram and WhatsApp, all in the name of retention.
Have an AI Sum Up This Article
ChatGPTMeta’s Assistant Moves Into the Threads Inbox
Since July 27, Meta has been rolling out Meta AI worldwide inside Threads direct messages. Until now, users could only summon the assistant by tagging it in a public post, a mechanic close to how Grok is called on X. Moving it into private messages changes the nature of the tool.
In practice, every user now has a one on one thread with the AI. They can forward it a post from their feed, an image, a link or a video, then keep asking questions to go deeper in the same conversation. Meta frames the private inbox as a way to give people more flexibility in how and where they interact with the assistant.
The company also planned for the reluctant. Anyone who does not want the assistant can mute the @meta.ai account or hide its replies. The rollout is global, which marks a clear step up from the handful of markets that had been testing the public version.
What Changes When an AI Sits in Your DMs
For users, the appeal is continuity. No more copying a post, opening a separate app and pasting the text to ask for a summary or an explanation, because it all happens in the inbox. That closeness lowers the barrier to using AI, especially for the curious who would never have installed a dedicated assistant app.
The everyday uses are immediate. A user can ask Meta AI to summarize a thread gone viral, translate an incoming message, explain an obscure reference or reword a draft before sending it, all without switching apps. For a newcomer who never dared open an assistant, meeting the AI inside a conversation they already know removes much of the intimidation. It is the same reason a familiar messaging window often keeps people engaged where a brand new app screen would lose them, and Meta knows that habit is its strongest asset.
This shift of AI toward direct action follows a deeper trend. When ChatGPT can now drive an entire desktop by voice, the assistant stops being a separate tab and becomes a layer inside the tool you already use. Meta applies the same logic to its messaging.
The flip side is privacy and attention. An AI parked in the inbox sees, by design, whatever the user chooses to share with it, and its constant presence nudges people to stay in the app rather than leave it. Convenience and attention capture move here hand in hand.
More articles on Horizon
- ChatGPT at Work Is Erasing the Lines Between Jobs
- Claude shared chats got indexed by Google
- Apple delays its AI glasses over privacy concerns
Why Meta Is Locking Down the Inbox Against ChatGPT
On the competitive side, the math is plain. By parking its AI inside Facebook, Instagram, WhatsApp and now Threads, Meta turns the inbox into a battleground. The stated goal is to short-circuit the reflex of opening a third-party app, and to hold the user inside its ecosystem at the exact moment a question comes up.
The target is clear. Meta wants to cut the trips to rival assistants, at a time when Gemini is closing in on a billion monthly users and ChatGPT remains the default reflex for millions. Meta’s lever is not raw model quality, it is distribution: billions of accounts already installed across its apps.
For rival labs, the fight moves to new ground. It is no longer only about benchmarks, but about where the assistant sits the moment a user needs it. A slightly better model that has to be opened in a separate app loses to a decent AI already present in the conversation. That is Meta’s bet, and it shifts the value of the model toward the point of contact.
One big unknown remains: perceived quality. If Meta AI disappoints on simple requests, convenience alone will not hold demanding users, who will keep the reflex of opening a stronger assistant. The distribution bet only works if the model stays at least decent, and that is exactly where the next round between Meta and its rivals will be decided.
Follow the story on Horizon.
