How does the free version make money, and does it matter?
Why the business model behind a free chatbot matters.
Every answer runs on large, power-hungry computers. Training a top model costs a fortune. Yet you can use many of them for free.
Somebody is paying. It's worth knowing who, because it shapes what the product is built to do.
None of these is evil. But each one pulls the product in a slightly different direction.
In early 2026, OpenAI announced it was testing ads in ChatGPT on its free and cheapest plans, with the higher paid tiers kept ad-free.
What you tell a free chatbot may now help decide what it shows you.
On many free and personal plans, your conversations can be used to train future models, unless you switch that off.
Anthropic, for example, asked consumer Claude users in 2025 to choose. People who opt in have their data kept for up to five years. Work, education and developer accounts are excluded.
Free often means "your conversations help make the product".
Any product paid for by attention has a reason to keep you chatting. The OECD's AI literacy framework, adopted in 2026, specifically warns about features like streaks, characters with names, and "overly agreeable responses" designed to boost engagement.
That last one should sound familiar. In 2025, an update leaned too heavily on thumbs-up feedback and made ChatGPT too flattering (see "Why AI agrees with you so much").
Check your plan's data settings once. Is training on your chats on or off? How long are chats kept?
Use a work account for work. It usually comes with stronger data protections than a personal one.
Notice when it's trying to keep you there. A product built for engagement will happily keep talking long after the useful part is over.