Five years ago, discussions among internet denizens in obscure online spaces sparked a debate about the authenticity of the web. It was suggested that since 2016, following Google’s introduction of its artificial neural network for machine translation, the internet had shifted dramatically. Proponents argued that bots began to take over online communications, leading to a web that felt increasingly desolate and unpopulated. A scholar noted that websites appeared “empty and devoid of people.”
While this claim lacks substantiation, it taps into a genuine concern. As early as 2021, social media platforms faced a deluge of fabricated views, likes, traffic, and engagement. Some speculated that automation could eventually dominate the internet and, by extension, our lives, but these fears were often dismissed as overblown. Yet, today, they seem eerily prescient.
The “Dead Internet” theory is regaining traction in mainstream discourse, as many now believe AI could overshadow human interactions, with conversations becoming nothing more than bots communicating with other bots. Each week brings further advancements in this direction. A report from the Pew Research Center in June showed that nearly a quarter of American adults engage with chatbots daily. By August, OpenAI unveiled a ChatGPT integration for iMessage, raising questions about how common AI-generated text will become in personal exchanges.
The integration for ChatGPT is designed to function on desktops rather than phones, requiring user approval before sending any messages. A similar feature has been available for Anthropic’s Claude, and various messaging apps already suggest AI-generated replies. OpenAI’s recent offering essentially formalizes practices that users have been implementing independently, either by copying text manually or through custom software solutions.
Michael Pfister, a Colorado-based software developer, pointed out that “low-effort communication has always existed.” He created his own version of the iMessage tool and receives prompts from ChatGPT daily to ensure he hasn’t overlooked important questions in his group chats. On occasion, he experimented with AI-generated messages, only to have friends quickly identify the text as being produced by an algorithm. He emphasized his preference for maintaining personal connections, stating, “I think relationships are experiential.”
Max Weinbach, a market analyst from Connecticut, similarly uses ChatGPT to communicate with his bosses, finding it simplifies his life. He likens it to Google’s suggested email replies, which have been around for years, though he acknowledges some might perceive this as disrespectful. “It’s hardly accepted or polite,” he remarked, noting that it can come off as rude.
While these individuals represent a cutting-edge mindset, many users on platforms like Reddit express concerns over AI’s intrusion into personal exchanges. Posts emerge regularly, seeking help to determine whether content—such as apartment listings or dating profiles—has been AI-generated. The anxiety intensifies when it comes to personal messages, prompting queries like “Did my partner use ChatGPT for that apology text?” or “Was that supportive message from my mom crafted by AI?” The underlying feeling is often one of betrayal.
Melissa Crowder, a 47-year-old from Alberta, Canada, shared her own encounter with AI in a dating app conversation. Initially enjoying an engaging dialogue with her match, she sensed something was off when he mistakenly copy-pasted her own message back to her. This revelation led her to scrutinize his previous responses, which exhibited template-like structures and unusual language patterns. When she confronted him about potential AI use, his acknowledgment—laced with emojis—felt disingenuous. Crowder stated that she wouldn’t have minded as much if he had simply been honest. Although she also resorts to AI tools for clarity in her messaging, she draws the line when the humanity is stripped from the interaction: “If I’m vibing with a chatbot and then I meet the person in real life, what is that conversation going to look like?”
Research conducted by teams from the University of Washington and Arizona State University delved into discussions within Reddit forums focusing on distinguishing AI-generated content. Their findings highlighted that users frequently assess the authenticity of images or videos while also sharing signs of AI-generated text—like excessive ornamentation or a repetitive vocabulary. The researchers expressed concern that users’ mental frameworks for identifying AI content may soon be obsolete as AI continues to evolve and improve.
The paper concludes on a cautionary note, indicating that as societal attitudes toward AI persist, the current stigma surrounding AI in personal communications might one day dissipate. In the future, we might find our phones less filled with human interactions or, paradoxically, that the "Dead Internet" theory remains as misguided as it has always seemed.
