Technically, yes — a few extra tokens. Practically, politeness isn't your problem. Unnecessary context is.
Well… technically, you're using a few extra tokens.
And if you're thinking "what the hell is a token?", you're not alone.
Put simply, tokens are the small pieces of text AI models use to process what you type and generate their response. The more text you send and receive, the more tokens are used.
So this:
Please can you review this code and tell me what's wrong? Thank you!
uses slightly more tokens than:
Review this code and tell me what's wrong.
But here's the important part: don't start being rude to ChatGPT to save money. The few tokens used by "please" and "thank you" aren't going to make a meaningful difference for most people.
The bigger lesson is about how you prompt AI
Instead of "can you please look at this and tell me what you think?", try "review this code, identify the problem and provide a fix".
And instead of dumping an entire document into AI, give it only the information it actually needs.
A good prompt usually contains four things
- Context — what does the AI need to know?
- Task — what exactly do you want it to do?
- Constraints — are there things it must or mustn't do?
- Output — what should the answer look like?
For example:
I'm running a Rails API on AWS. My EC2 instance is returning 200 when I curl the application directly, but the ALB health check is failing. Help me identify the likely causes. Give me 5 things to check, in order of likelihood.
That's a much better prompt than "my ALB isn't working, please help".
So, don't worry about removing every "please" and "thank you". Instead, focus on removing unnecessary information.
Don't optimise the politeness. Optimise the context.
And yes… you can still say thank you.