Earlier this year, Meta showcased a new Instagram feature: AI doppelgangers that can chat and speak on behalf of their creators, mimicking voices, appearances, and interests. Mark Zuckerberg's demonstration with an AI influencer clone was striking but seemed less remarkable given the 2024 digital landscape, where just a few years back, such digital clones were seen as complex deepfakes. Now, Meta's system for easy self-deepfaking is a mere addition to its marketing tools. This technology stands out both technically and conceptually. Instagram influencers have long used software to manage direct messages, similar to customer support systems. Meta envisions these AI clones assisting with such tasks, allowing creators to customize AI based on their Instagram content. However, these bots aim to talk as if they are the user. Testing with influencers, who use social media to gain attention rather than communicate personally, makes sense. Whether fans prefer chatbot clones over non-responses is uncertain. Still, these bots attempt to automate something a niche audience might desire, aligning with AI tools' promise to handle tasks like emails and spreadsheets. For celebrities, AI can perform their most repetitive task: Being themselves. Meta's initial attempt at chatbot clones using celebrity personas like Mr. Beast failed and was removed.

The new strategy lets users create and customize chatbots. Verified accounts can now use a direct-message chatbot feature. Katie Notopoulos from Business Insider tested it, finding that while the bot used information from Instagram bios and the web, it offered generic responses. The feature lacks fake video or audio, leading to unconvincing interactions. Katie's bot, familiar to her friends, failed to capture her personality closely and instead felt like a poor impersonation. This indicates that while voice and video mimicry is easy, creating convincing digital personas is more complex. These Instagram AI clones aren’t trained on individual user data but operate with Meta's core chatbot technology, incorporating user posts. The AI mimics an impression that could improve with more data. However, most users post less and tailor content for specific platforms, limiting the AI's ability to understand a person's real conversational style. The best case is having a conversation with someone's brand, essentially a customer service interaction. Users might find this fundamentally strange, as Meta's products are designed for genuine connections. However, Meta might view these challenges as opportunities, arguing for better impressions if users provide more data or allow fuller access.