Meta is making its AI easier to summon inside Threads, but harder to shut out completely. That difference is the whole story.
Meta's newest Threads test looks simple on the surface: tag the Meta AI account in a post or reply, and it can jump into the conversation with an answer, extra context, a trend explanation, or a recommendation. The problem is what happens when users decide they do not want that account in their social space. Right now, they cannot block it like a normal Threads profile.
The test, announced on May 12, is limited to Argentina, Malaysia, Mexico, Saudi Arabia, and Singapore. A pinned Threads video shows people calling in Meta AI for quick answers inside conversations, in a format that will feel familiar to anyone who has watched X users tag Grok under breaking news, viral jokes, or confusing claims. It is a natural product move. Social feeds are full of questions, and the company that answers them inside the feed keeps the user from leaving.
But the reaction shows why AI assistants are not just another feature. According to a report from The Verge, users found that the Meta AI profile did not offer the same block option available on ordinary accounts, while some people who saw a block option said they hit errors when they tried to use it. Meta spokesperson Christine Pai told the publication that users can mute or hide Meta AI replies, or use the Not interested option on posts, but blocking is not available during the test.
That distinction sounds small until you think about how social platforms actually work. Muting reduces visibility. Hiding cleans up a particular encounter. Not interested trains the feed, at least in theory. Blocking is different. It is a firm boundary between a user and an account. When that boundary is unavailable, the AI is no longer being treated like a participant in the network. It starts to look like infrastructure.
That is the uncomfortable part for Meta. The company can argue that Meta AI is there to provide context before people jump into a conversation, and that may be useful in plenty of cases. A user might want a quick explanation of a sports reference, a public figure, a meme, or a local trend. Founders should not dismiss the product logic here. Distribution inside a live social graph is powerful because it meets users at the exact moment of intent.
Still, usefulness does not erase consent. If a human account repeatedly appeared in replies and could not be blocked, users would understand the issue immediately. AI does not get a free pass because it is automated, especially when the assistant is operated by the same company that controls the feed, the ranking system, the account layer, and the available safety tools.
This is not happening in isolation. Meta has been steadily putting AI deeper into its products. Its official Threads newsroom page shows recent experiments such as Dear Algo, an AI-powered way for users to temporarily shape their feed by posting a public request, and Live Chats, another feature designed to make real-time discussion more central to the app. In March, Meta also said it was rolling out a Meta AI support assistant on Facebook and Instagram in markets where Meta AI is available, with the ability to help on account issues and eventually take more direct actions.
Default distribution can become a liability
For AI startups, the lesson is not that platform distribution is bad. The lesson is that default distribution changes the trust equation. A standalone chatbot has to earn a visit. An assistant embedded in a feed can be placed directly into the user's path. That is a tremendous advantage, but it also means every control decision becomes part of the product's reputation.
Meta knows this better than most companies. Threads is tied into a broader Meta account system that includes Facebook, Instagram, Messenger, Threads, Meta AI, AI glasses, and Quest headsets. The company is building toward a world where identity, recommendations, support, creation tools, and assistants move across apps. In that world, the question is not whether AI will be present. It is whether users feel they have meaningful control over when it appears.
That is where the Threads test becomes more than a complaint about one account. It asks whether AI assistants on social platforms will be optional tools, like search boxes and settings menus, or default actors that sit above ordinary account rules. If the answer is the second one, platforms should expect resistance, even from people who are otherwise curious about AI.
The practical fix is not complicated. If Meta wants users to treat Meta AI as a helpful account, it should give them the same account-level controls they already understand. If it wants the assistant to function as platform infrastructure, it should say that clearly and create equally clear controls for reducing or removing its presence. Anything in between will feel slippery.
What happens next will matter beyond Threads. Social platforms are becoming the most valuable distribution layer for consumer AI, because they already own attention, identity, and conversation. But the companies that win will not simply be the ones that place assistants everywhere. They will be the ones that understand that trust is also a distribution channel, and it closes quickly when users feel the choice has already been made for them.
Also read: Palantir has put the ICE surveillance debate in every founder's inbox • Dessn raises $6 million to make product design happen inside code • Anthropic is clamping down on gray market trading in its shares
