Politeness aimed at a billing meter
Please, thank you, and "I would like you to" cost tokens on every request forever.
Courtesy phrases are the easiest tokens to find and the least interesting to lose. A model does not need to be asked nicely, and "please" in a system prompt is not politeness — it is a recurring line item.
The saving per instance is tiny. The reason it is worth fixing is multiplication: a system prompt is re-sent on every single request, so eight wasted tokens at ten thousand requests a day is eighty thousand tokens a day, every day, for the life of the feature.
This is also a reliable signal that a prompt has never been audited. Prompts that still say "thank you" usually have much more expensive problems further down.
Please could you kindly extract the invoice total. Thank you!Extract the invoice total.Check your own prompt
The analyser checks this pattern along with the other 25, prices each finding against your request volume, and hands back a rewritten prompt. It runs in your browser — nothing is uploaded.
Run the analyserMore on filler
- A role that describes nothing"You are a helpful AI assistant" tells the model nothing it did not already assume.
- Prompt folkloreTips, deep breaths, and other emotional-appeal phrasing, copied from 2023-era prompting advice and never removed.
- Chat pleasantries in a system promptSign-offs written for a human conversation, left in a machine instruction.
- Long ways of saying short things"Due to the fact that" is four tokens. "Because" is one.