ChatGPT has quietly turned into the “busiest hangout” spot on the internet. What began as a productivity tool now draws crowds like a digital piazza with more than 800 million registered users and around 125 million of them showing up daily. Average session time? Roughly 14 minutes. That’s the kind of attention Facebook hasn’t seen since your aunt’s Candy Crush phase.
That’s why Meta’s nervous
When Sam Altman recently told an interviewer that Meta sees OpenAI as its “deepest competitor,” he wasn’t talking about model parameters or synthetic benchmarks. He meant time. The scarce attention currency that fuels Meta’s $160 billion ad empire.
Each minute spent chatting with a large language model is a minute not spent scrolling past ads, hearting vacation photos, or feeding Meta’s recommendation machine.
ChatGPT’s interface now mirrors the addictive pull of a social feed; not through drama, but through dialogue. According to AI analyst Rohan Paul, the numbers show the shift. Users send 2.5 billion prompts every day, each one a micro-post between human and bot.
OpenAI traffic 2 billion visits in May 2025, according to Similarweb data. Inside those sessions, users do what they used to do on Facebook: seek validation, share ideas, and vent. Only this time, the party is private, the algorithm is personal, and nobody’s selling you sneakers.
ChatGPT: the threat beneath the feed
Meta’s business works only if users linger in human spaces, because ads need eyeballs. Instagram alone is projected to generate over $32 billion in U.S. ad revenue in 2025, more than half of Meta’s domestic take. Every minute that vanishes into ChatGPT’s prompt box is one less exposure to an ad impression and one more dent in that flywheel.
Adding insult to injury, ChatGPT isn’t just answering questions; it’s fulfilling social needs. LLMs can hold context, generate voice and images, and mimic empathy on demand. Research on AI companion apps like Replika shows users forming deep para-social bonds, in some cases, even romantic ones. Whether comforting, flirtatious, or purely informational, these bots meet people where they are; instantly and tirelessly.
What we used to outsource to timelines (affirmation, debate, connection) is starting to flow into 1:1 chat. Meta’s feed shows you others so you can see yourself. AI companionship removes the middleman entirely.
Two graphs, two futures
Facebook’s power came from a single vast social graph. You and your friends in perpetual comparison. ChatGPT’s power is atomized. Each user has a private relationship with a model that remembers their tone, preferences, and style of humor.
Under the hood, that intimacy runs on three building blocks: memory retrieval for continuity, persona prompts for tone, and a fast vector cache to simulate long-term memory. It’s a quiet revolution, and millions of personal mirrors are replacing one global stage.
Altman understands that Meta isn’t competing with OpenAI on data centers or diffusion models. They’re competing for human moments: the idle 14 minutes before bed, the “quick check” that turns into an emotional exchange.
If even a small fraction of Meta’s 3.4 billion daily users shift that engagement to chatbots, the financial impact cascades fast. Because every missing minute is a measurable loss on ad inventory.
And that’s why, inside Meta’s glass towers, the quiet hum of ChatGPT might sound less like innovation and more like a ticking clock.