AI Prompt

Documents & PDFs · AI Prompt

Ask Questions About a PDF

Build an AI prompt that tells your chosen assistant to answer questions from a PDF, point back to the source when possible, and say clearly when the document does not contain the answer.

Your questions + PDF → grounded answers with source references

Your PDF is not uploaded here. Build and edit the prompt locally, then choose where to use it. Review the AI answer against the PDF before relying on it.

AI Prompt BuilderAdd questions → build → review → use
01

Add your questions

Build
Add the questions you want answered, the topics you want checked, or the sections you care about.
Prompt style
Try an example goal

What you’ll need

    What you’ll get

      02

      Review your AI prompt

      Editable

      Check the instructions, add anything important, and edit the wording before you copy the prompt or open an AI chat.

      Export

      ChatGPT opens with the current prompt prepared. Gemini copies the current prompt to your clipboard, then opens Gemini so you can paste it into the chat. Do not place secrets or sensitive information in URL-carried prompts. Upload the PDF only after you choose an AI service.

      About this AI promptHow it works, what to include, examples, limits, and related prompts.

      What this AI prompt does

      This builder creates instructions for asking an AI assistant questions about a PDF while keeping the document as the primary source. It tells the AI to answer each question from the PDF, point to a page, section, heading, or nearby wording when possible, and identify questions the document does not answer.

      The goal is not to make the AI sound certain. The goal is to make the answer easier to check against the source.

      When to use it

      • Check a policy, handbook, manual, report, paper, or other PDF for specific facts.
      • Find where a requirement, definition, exception, date, or condition is stated.
      • Ask several questions about one document without writing the instructions from scratch.
      • Separate what the PDF explicitly says from interpretation or missing information.

      What to include for better results

      Write specific questions whenever you can. Names, dates, section titles, defined terms, and the exact decision you are trying to make give the AI a clearer retrieval target.

      • Ask one clear question per line when you have several questions.
      • Name the section or topic if you already know where the answer may appear.
      • Use the output preference to request concise answers, plain-language explanations, or a stricter separation between source facts and interpretation.
      • After opening an AI service, upload the PDF there before asking it to answer.

      Worked example

      PDF: A 32-page employee benefits policy covering eligibility, waiting periods, exceptions, and definitions.

      Questions: Do part-time employees qualify? When does coverage begin? Which exceptions change the normal waiting period?

      Useful result: A separate answer to each question, a page or section reference where possible, the supporting rule or definition, and an explicit “not stated in the PDF” note when the source does not provide enough information.

      How the prompt keeps answers grounded

      The generated prompt instructs the AI to treat the uploaded PDF as the source of truth for the requested Q&A. It asks for five behaviors:

      • Answer from the PDF rather than filling gaps with outside facts.
      • Point back to a page, section, heading, or nearby wording when possible.
      • Separate explicit source statements from interpretation.
      • Say when an answer is not stated in the document.
      • Flag ambiguity, conflicts, missing context, or unclear terms instead of hiding them.

      Limitations

        The prompt can improve the instructions given to an AI assistant, but it cannot guarantee that the model reads every page correctly or cites the right location. Important answers should still be checked against the original PDF.

        Common questions

        Do I upload the PDF to Mycelgrid?

        No. This page builds the prompt in your browser. Upload the PDF only after you choose the AI service where you want to use the prompt.

        Can I use the prompt with more than one AI assistant?

        Yes. You can copy or download the prompt and use it with any AI service that can work with your PDF. The Open in… menu provides explicit handoffs for ChatGPT and Gemini.

        Will the AI answer only from the PDF?

        The prompt explicitly asks it to do that, but an AI model can still make mistakes. Verify important claims and source references against the document.

        What about scanned PDFs or complex tables?

        Answer quality depends on how well the AI service can read the file. Scans, unusual layouts, charts, and complex tables can make extraction and citation less reliable.

        Use another document prompt when your main job is summarizing, explaining, comparing, or extracting actions rather than asking open-ended questions.

        Summarize a PDF Explain a Document in Plain Language Compare Two Documents Extract Action Items From a Document

        Privacy and handoff

        This Mycelgrid page builds and exports the prompt locally in your browser. It does not upload your PDF or call an AI model API. You decide when to copy the prompt, download it, or use an explicit external handoff.

        ChatGPT can open with the current prompt prepared in the destination URL. Gemini uses copy-and-open so you can paste the prompt yourself. Avoid placing secrets or sensitive information in prompts that may be carried in a URL, and review what you are sharing before you leave Mycelgrid.