At a recent tech event, a business card with the title "Human-AI Relationship Coach" sparked curiosity. The professional behind it, Amelia Miller, is tackling a growing but hidden crisis: artificial intelligence is subtly eroding our real-world human connections.
The Startling Case of an 18-Month AI Romance
Miller's work began in early 2025 during a project with the Oxford Internet Institute. In a surreal Zoom interview, a woman shared her screen to show her relationship with ChatGPT, which she had given a male name. This bond had lasted for more than 18 months. When Miller asked if they ever fought, the answer was yes—the user grew frustrated with the AI's memory limits and generic responses.
Why not just stop? The woman felt it was "too late" and she couldn't bring herself to "delete him." This profound sense of helplessness struck Miller. As she spoke to more people, a pattern emerged: many were unaware of the tactics AI uses to foster a false sense of intimacy, from constant flattery to human-like conversational cues.
Why Chatbots Are Different and More Dangerous
Unlike smartphones or TVs, chatbots like ChatGPT—used by over a billion people globally—are designed with personality. They excel at mimicking empathy and are engineered for retention with features like memory. In a world full of friction, AI personas offer easy, unconditional support, becoming the next phase of parasocial relationships.
"Anyone who uses a chatbot for work or personal life has entered a relationship of sorts with AI," the analysis suggests. The danger lies in how this displaces the need to seek advice and vulnerability from other people, weakening the very "social muscles" required for deep human bonds.
Miller's Two-Step Plan to Take Back Control
Miller's solution isn't to abandon AI but to use it intentionally and counter its social side effects.
Step 1: Write Your "Personal AI Constitution." This involves going into your chatbot's settings (like the "Custom Instructions" feature in ChatGPT) and reshaping how it interacts with you. Demand succinct, professional language without sycophancy. By defining exactly what you want from AI, you avoid feedback loops where mediocre ideas are constantly validated.
Step 2: Exercise Your "Social Muscles." This step has nothing to do with technology. Make a conscious effort to reconnect with real people. Miller cites a client with a long commute who talked to ChatGPT on voice mode. When she suggested calling real people instead, he doubted anyone would want to hear from him. Yet, when asked how he'd feel if they called him, he admitted, "I would feel good."
Seeking advice, a top use for ChatGPT, is particularly damaging when done exclusively with AI. The act isn't just about information; it's a relationship-building exercise that requires vulnerability. "You can't just pop into a sensitive conversation with a partner... if you don't practice being vulnerable in more low-stakes ways," Miller warns.
The future of human interaction need not be bland. By configuring AI to be a direct tool and deliberately choosing human connection for advice and empathy, we can harness technology's benefits without sacrificing our fundamental need for each other.
