The 'AI companion market' is worth $24 billion. Or $500 billion. Nobody actually knows.

If you write about this category, you'll constantly meet a sentence like "the AI companion market is worth $X billion." It's worth knowing, before you cite one, that you can pick almost any number you want.

For 2026 alone, the published estimates don't disagree at the margins — they disagree by more than twentyfold. Research and Markets puts the market at about $24 billion in 2026. Business Research Insights puts it at about $501 billion the same year. Precedence Research lands in between, around $37 billion for 2025, while other trackers cite figures as small as low single-digit billions. These cannot all be measuring the same thing, because they aren't.

The gap is definition, not data

The disagreement comes almost entirely from where each report draws the boundary of "AI companion." Define it narrowly — romantic and social chatbots people form relationships with, the subject of this whole blog — and you get a market in the single-digit-to-low-tens of billions. Define it broadly — any emotionally aware AI assistant, folding in mental-health tools, education aids, elder-care, and general productivity assistants — and you're suddenly counting a slice of the entire AI economy, which is how you reach figures in the hundreds of billions.

Two more multipliers make the spread worse. First, almost every number is a forecast: compound a 30–42% annual growth rate over a decade and a small difference in the base year explodes into a wildly different headline. Second, the source matters. Many of these figures come from firms whose product is the report itself, complete with a "download free sample" button — and a bigger addressable market sells better. None of that makes them fraudulent. It makes them estimates built on assumptions you can't see, presented with a precision they don't have.

What you can actually say

The honest move isn't to throw the numbers out; it's to demote them. A single dollar figure for "the AI companion market" is close to meaningless without three things attached: whose definition, which base year, and whether it was measured or modeled. Ask those, and most headline numbers quietly fall apart.

What's left is firmer and more useful. Adoption is genuinely large and rising — Common Sense Media found that nearly three in four US teens have already used an AI companion — and the leading apps report user and subscriber bases in the millions. Those are countable facts. The trillion-dollar trajectories are not; they're bets dressed as measurements.

This is a slightly different kind of dispatch from the rest of the series — less an event, more a reading habit. But it's the habit the whole project runs on: when a number is doing a lot of persuasive work, the first question is always how it was made. For the grounded version of where the category actually stands, the 2026 overview sticks to what can be sourced.

Sources