The price of AI has nothing to do with how hard the question is. A chatbot charges the same for a birthday poem as for a legal analysis. Difficulty is free. Volume is not.
AI is billed by the token — roughly three-quarters of a word — counted both going in and coming out. The bill tracks how much text moves through the machine, not how clever the work is.
That is why ordinary chat stays cheap. You are the brake. You type a question, read the answer, think, and type again. Maybe twenty exchanges in an hour. At the per-token rates the AI companies publish, an hour of conversation a day runs somewhere between $5 and $30 a month in raw computing. Which is why a $20 subscription works.
The expensive part of AI was never answering you in chat. It is most expensive when working without you. The next time you hear an AI expert talk about ‘agentic AI’ and setting up agents to work independently, that expert is advising you to spend 30 times more tokens than your chat window does. So before you ask AI to do the work for you, consider the possible cost.
AI remains inexpensive for ordinary chat, but the way you use chat matters. Keep chat conversations in each window dedicated to a single topic. When you change topics, change windows. AI reads the entire scope of the conversation in the window. Conversations that accumulate unrelated material dilute attention and degrade the quality of responses.
Briefs and quick reads selected and edited by The Smart Reader’s editorial team.
