Whether you use ChatGPT, Claude, Gemini or Copilot, these habits make answers faster, sharper and cheaper — and help you hit usage limits much later.
1. How you ask — write smarter prompts
- One clear ask per message. "Summarise this in 5 bullets" beats "Can you help me with this?"
- Be specific upfront. Add who it's for, format and length in the first message, so you don't need 3 follow-ups.
- Skip politeness filler. "Hi, hope you're doing well, I was wondering if you could please…" → just ask.
- Give context once, not every time. Explain your situation at the start, then ask short follow-ups.
- Use bullets, not long paragraphs when giving instructions.
- Ask for one answer, not options, unless you really want choices: "Give me the single best subject line."
- Group related questions into one message instead of sending five separate ones.
2. Control the reply length
- Say the size you want: "in 3 lines", "under 100 words", "just the code", "yes/no first".
- Say "no intro, no summary at the end" to remove filler.
- Ask for a table or bullets when you only need to scan.
- For code, ask for "only the changed lines", not the full file again.
- Ask for "short answer first, details only if I ask".
3. Manage chat history — the hidden cost
- The AI re-reads the whole chat every time you send a message. A long chat gets more expensive with every reply.
- Start a new chat for a new topic. Don't ask about travel plans in your coding chat.
- Edit your last message instead of sending "no, I meant…" corrections — the wrong answer drops out of the history.
- Regenerate instead of arguing with a bad answer.
- Long chat? Summarise and restart: ask "Summarise this chat in 10 lines", copy it into a fresh chat and continue.
4. Files, images and pasting
- Paste only the part that matters, not the whole document, email thread or code file.
- Remove junk before pasting: signatures, disclaimers, repeated headers, old email replies.
- Don't re-upload the same file in the same chat — the AI already has it.
- Crop screenshots to the area you need; big or many images cost a lot of tokens.
- Convert to text when you can. Copying the text of a PDF is usually lighter than uploading a scanned image of it.
- Big documents: ask about one section at a time instead of "read this 100-page PDF and tell me everything".
5. Settings and model choice inside the app
- Pick the lighter model ("mini", "flash", "haiku", "fast") for simple tasks like rewriting, translating or quick facts. Save the top model for hard thinking.
- Turn off extended thinking / reasoning mode for easy questions — thinking uses lots of extra tokens.
- Turn off web search, deep research and tools when you don't need fresh info; each search adds pages of text.
- Disable connectors and plugins you aren't using in that chat.
- Keep custom instructions and memory short — they're added to every chat you start.
6. Reuse instead of repeating
- Save your best prompts in Notes as templates and reuse them.
- Use Projects / Custom GPTs / Gems for repeated work, so you don't paste the same background every time.
- Keep a short "about me" paragraph to paste only when it matters, instead of re-explaining in every chat.
- Reuse earlier answers — copy them into your notes instead of asking the same question again.
7. Language
- English usually uses the fewest tokens; Hindi, Telugu and other native scripts often take 2–4× more for the same meaning.
- Hinglish or Teluglish in English letters is a good middle path when you're more comfortable that way.
- Ask for the answer in one language, not "English and also translate to Hindi" unless you need both.
8. If you use AI through an API or automation tool
- Set max output tokens so replies can't run long.
- Keep the system prompt short — it's sent with every request.
- Turn on prompt caching if your provider supports it, and keep the fixed part of the prompt at the top.
- Use a low temperature (0.2–0.5) for focused, shorter answers.
- Use small models by default, bigger ones only when needed.
- Send only the last few messages, not the full history.