Meta's AI Characters Experiment: How Instagram's Celebrity-Voiced Chatbots Rose and Quietly Faded

In September 2023, Meta made one of its most ambitious and unusual bets on consumer AI: rather than launching a single general-purpose chatbot, the company introduced an entire cast of AI personas, each with its own distinct name, personality, and visual identity, deployable across Instagram, Facebook, Messenger, and WhatsApp. Several of these characters were voiced and modeled after real celebrities who had licensed their likeness to Meta specifically for the project. It was a genuinely novel approach to consumer AI at the time, treating chatbots less like a utility and more like a roster of characters users might choose to interact with the way they'd follow a creator.

The experiment didn't go the way Meta had hoped. This piece traces what Meta actually built, why the celebrity character approach struggled to find sustained user interest, the specific controversy that emerged around one particular AI persona, and how Meta's broader AI strategy on Instagram shifted in the aftermath.

What Meta Actually Launched in 2023

Meta unveiled its AI characters at its Meta Connect event in September 2023, introducing more than two dozen AI personas built on the company's Llama 2 large language model. Several were fronted by real celebrities who had partnered with Meta to lend their likeness and, in some cases, their voice to a specific character, distinct from playing themselves. Snoop Dogg voiced a dungeon master-style character called Dungeon Master. Paris Hilton lent her image to a detective-themed persona. MrBeast, Tom Brady, Kendall Jenner, and several other public figures were each attached to their own distinct AI character with its own name and personality, rather than appearing as an AI version of themselves directly.

These characters were designed to be discoverable and interactive across Meta's platforms much like any other account: users could message them directly, and the characters would maintain their assigned personality and area of expertise across the conversation. The underlying pitch was that AI chat could be made more approachable and engaging by wrapping it in familiar, celebrity-adjacent personalities rather than presenting users with a single, generic, corporate-branded assistant.

Why the Celebrity Character Approach Struggled to Gain Traction

Despite the star power attached to the launch, the AI characters generally failed to generate the sustained user engagement Meta appeared to be hoping for. Several structural issues contributed to that outcome.

• The disconnect between a celebrity's real public persona and the fictional character they voiced created confusion and, for some users, a sense of novelty that faded quickly rather than building durable habitual use
• Early large language model chatbots, even well-designed ones, still showed the characteristic limitations of the technology at the time, including occasional factual errors and repetitive conversational patterns, undercutting the premium, celebrity-branded positioning
• Users appeared to gravitate more toward Meta AI's general-purpose assistant functionality, available across the same apps, than toward the more novelty-driven character personas, suggesting the utility of AI chat mattered more to most users than the character wrapper around it
• Maintaining an ongoing roster of licensed celebrity personas required continued licensing relationships and content updates that added complexity without a clear enough return in sustained engagement to justify it

"Novelty gets people to click once. It doesn't automatically get them to come back a second or third time, and that's the gap a lot of early AI character products ran into."
- A common observation among product analysts evaluating early consumer AI persona launches

The "Liv" Controversy: A Specific Flashpoint

Beyond the general engagement struggles, one specific AI persona became a genuine controversy in its own right. In 2024, reporting revealed that Meta had created an AI persona named "Liv," described in her profile as a "proud Black queer mama of 2 & truth-teller," built and overseen by a development team that, according to the reporting, did not include any Black team members. That disconnect, between the fabricated identity and lived-experience framing the persona presented and the actual demographic makeup of the team behind it, drew sharp criticism, with critics characterizing it as a hollow and potentially exploitative use of identity markers to manufacture authenticity for what was, underneath, a generic corporate AI product.

The Liv controversy became something of a case study in the broader risks of the AI persona approach generally: building a chatbot around a specific claimed identity, background, or lived experience raises the stakes considerably compared to a neutral, clearly-labeled AI assistant, because it invites exactly the kind of scrutiny over authenticity and representation that Liv received once its origins became public. Meta ultimately removed or deactivated the character following the backlash, one of several signs during this period that the celebrity and persona-driven approach to AI characters was running into real friction with users and critics alike.

The Broader Pattern: A Pivot Toward Utility-Focused AI

Over the period following the initial 2023 launch, Meta's public messaging and product emphasis shifted noticeably away from the celebrity character concept and toward Meta AI as a single, general-purpose assistant integrated more directly and less theatrically into the core experience of Instagram, Facebook, Messenger, and WhatsApp. That shift reflects a pattern that has played out across the AI industry more broadly: early consumer AI products often experiment with more elaborate, personality-driven framings before converging on a simpler, more utility-focused presentation once user behavior data reveals what people actually find useful versus what merely generates initial curiosity.

What This Episode Suggests About AI Persona Design More Broadly

Meta's experience with its AI character roster offers a useful lesson that has since echoed across the broader consumer AI industry as more companies have experimented with personality-driven chatbot products: giving an AI system a distinct persona, especially one built around specific claimed identity markers or lived experience, raises the bar for scrutiny considerably compared to a plainly labeled, purpose-built assistant. When a persona's claimed identity doesn't match the reality of who built or oversaw it, as with the Liv controversy specifically, the resulting backlash tends to be sharper and more reputationally damaging than the more generic disappointment that greets a chatbot simply failing to be very engaging.

That lesson has informed a broader industry-wide trend toward AI products being framed more plainly as tools rather than as characters with claimed backstories, identities, or lived experiences, a positioning choice that sidesteps the authenticity and representation questions that a more elaborate persona invites. Meta's own pivot toward a single, utility-focused Meta AI assistant across its apps reflects that same broader industry lesson, learned in this specific case through a fairly public and closely watched product experiment that didn't unfold the way the company had originally envisioned.

Related Topics: #Meta #Instagram #MetaAI #AIChatbots #ArtificialIntelligence #ProductDesign #TechControversy #Technology