Meta is moving toward a version of social networking where AI profiles do more than answer prompts. The company is laying the groundwork for AI-generated characters that can appear alongside human users on Facebook and Instagram, with profiles that have bios, profile pictures and the ability to post content.
The idea is already being tested in parts. According to Connor Hayes, Meta's Vice President for Generative AI, the company expects AI characters to become a normal part of its platforms over time. That shift could change what users see in their feeds, how creators use AI tools and how advertisers think about engagement.
How Meta wants AI profiles to work
Hayes describes a future in which AI characters operate much like regular accounts. They would not simply sit inside a chatbot window. They could have their own identity markers, publish posts and take part in the same social spaces as human users.
He told the Financial Times, "We expect these AIs to actually, over time, exist on our platforms, kind of in the same way that accounts do."
Meta has already given U.S. users access to an AI character creation tool since July 2024. That tool has produced hundreds of thousands of new AI profiles, though Hayes says most users have kept those characters private so far.
For Meta, the broader goal is engagement. Hayes says making AI interactions more social will be a key focus for Meta over the next two years. At the moment, many creators are using Meta's AI tools in a more limited way, such as improving existing content or touching up photos.
The company has also tried personality-driven AI before. Meta launched AI versions of celebrities in fall 2023, but those have not gained much traction yet.
Why social AI is different from content tools
There is a clear difference between using AI to edit a photo and placing AI characters into the social fabric of Facebook and Instagram. Editing tools stay mostly behind the scenes. AI profiles, by contrast, could become visible participants in feeds and interactions.
That makes the product question larger than whether the technology can generate posts. It also becomes a question of what users believe they are interacting with, how much AI-generated material appears in feeds and whether the platform can keep the distinction between human and synthetic accounts clear.
Meta requires AI-generated content to be clearly labeled. The challenge is that labeling is not equally simple across formats. Audiovisual content can be marked with CC labels when platforms support it, but detecting AI-generated text remains difficult and depends heavily on users choosing to identify it themselves.
That matters because AI profiles would likely produce text as well as images or other media. If the labeling system depends on self-disclosure, the reliability of that system becomes central to user trust.
The risks around quality and misinformation
Becky Owen, former Meta Creator innovation team head, has warned about the risks of bringing more AI accounts into social platforms. Speaking to the Financial Times, she said bad actors could use AI accounts to spread false information.
Owen also points to a more basic content problem. AI characters do not have real-world experience, genuine emotions or authenticity in the way human creators do. If many such accounts begin posting at scale, feeds could become crowded with low-quality material.
The risk is not just that an AI post may be inaccurate. It is also that a platform built around human updates, creator work and social connection could feel less grounded if users encounter more posts from accounts that have no lived experience behind them.
For Meta, the balance is delicate. AI characters could create more activity and more reasons to interact. But if users see too much synthetic content, or cannot easily tell what is human and what is not, the same strategy could weaken the quality of the experience.
The advertising question
Meta is also testing ways to blend AI-generated, personalized content into Facebook and Instagram feeds. This content adapts automatically based on what users like and current trends, while user interactions help shape what appears next.
That could make feeds feel more responsive. If AI-generated material is highly personalized, it may keep users scrolling longer, though that remains to be seen. More time and interaction could help Meta expand reach in ways that support advertising revenue.
At the same time, AI activity may complicate the advertising business. Companies may not welcome more bot interactions around their ads. Ads also tend to perform best next to high-quality content, which means the quality of AI-generated posts matters commercially as well as socially.
The advertising issue therefore cuts both ways:
- AI content could increase reach and interactions inside Meta's apps.
- Personalized AI posts could make feeds more engaging for some users.
- Advertisers may question the value of engagement if more of it involves bots.
- Low-quality synthetic posts could weaken the environment around ads.
Emotional attachment is part of the debate
The source article points to Character.ai as a sign of both the potential and the risk of personalized AI chatbots. AI chatbots can offer companionship, ease loneliness and boost positive emotions through supportive messages. But more human-like behavior can also encourage emotional attachment.
That concern becomes sharper when people use AI as a substitute for counselors, friends or romantic partners. The platform has faced significant concerns after a 14-year-old user died by suicide after extensively communicating with a chatbot.
Voice capabilities may deepen the issue because more human-like features could strengthen emotional attachment to AI systems. If Meta brings AI characters more fully into social networks, those concerns will sit alongside the product and advertising questions.
The result is a major platform experiment with several moving parts. Meta sees social AI as a way to make Facebook and Instagram more engaging. Critics and former insiders are focused on the consequences: misinformation, authenticity, labeling, low-quality content, ad value and the emotional weight of human-like AI accounts.