How to Photograph Marked-Up AI Policy Printouts and Turn Them Into a Clean Revision Log With Gemini

Many nonprofits are already using AI in small, unofficial ways, but the rules around donor data, confidential notes, public statements, and staff review often lag behind. This workflow helps you turn real internal material into a usable governance artifact instead of starting from a generic template. It is for nonprofit executives, operations leads, development leaders, board administrators, and policy owners who need something practical that staff can actually follow.

Editorial guardrail: Use AI to extract and organize handwritten edits. A human reviewer should compare every logged change against the marked paper before any revision is accepted into the official policy.

What you need

• Printed policy pages with handwritten edits from a committee, leadership, or board review
• A phone with a camera and Gemini access
• Good lighting and enough table space to photograph pages clearly
• A place to save the final revision log, such as a Google Doc or shared draft file

How to capture or gather the source material

• Lay the pages flat in good light. Avoid shadows across the margins where handwritten edits often sit.
• If there are many pages, photograph them in order or use your phone scanner to create one PDF first. Keep page numbers visible.
• Crop only if needed. Do not cut off handwritten arrows, side notes, or strike-through text.
• If the handwriting is dense, take a second close photo of any page with hard-to-read notes before you move on.

The fastest workflow

1. Upload the page photos or scan to Gemini from your phone.
2. Ask Gemini to create a clean revision log with page number, original section, proposed change, reason if visible, and open questions.
3. Ask for a second pass that separates clear edits from uncertain handwriting or ambiguous comments.
4. Move the cleaned log into your master policy draft and assign each item as accept, reject, clarify, or legal review.
5. Store the photos until the final policy is approved so you can verify any disputed edit later.

Tool-specific instructions

Primary path: Gemini

• Gemini is a strong fit because the workflow begins with photos and needs quick mobile capture rather than a longer desktop setup.
• Ask for a revision log, not a rewritten policy. Extraction comes first. Drafting comes second.
• If a page is hard to read, ask Gemini to list uncertain words or phrases instead of guessing them.
• For long meetings, process the packet in batches of 5 to 10 pages so it stays easier to review.

Fallback options

ChatGPT fallback

• Upload the page photos to ChatGPT and ask for the same structured revision log. Mark up any hard-to-read area before upload if you want the model to focus on it.
• This works well when one or two pages contain the most important edits.

Claude fallback

• Upload the scanned PDF or images to Claude and ask it to extract revisions into a table with page numbers and uncertainty flags.
• Claude is a good fallback when the meeting packet becomes a multi-page PDF instead of loose photos.

Copy and paste prompt blocks tailored to the workflow

Primary prompt

{
  "task": "Extract handwritten policy edits from uploaded page photos and turn them into a revision log.",
  "required_columns": [
    "Page",
    "Section or heading",
    "What was changed",
    "Suggested clean wording",
    "Reason stated or implied",
    "Unclear handwriting or open question"
  ],
  "instructions": [
    "Preserve page order.",
    "Do not guess unclear handwriting.",
    "Flag anything that needs manual confirmation.",
    "Keep the output focused on revision tracking, not full policy rewriting."
  ]
}

Fallback prompt

{
  "task": "Convert these marked-up policy pages into a committee revision checklist.",
  "required_headings": [
    "Clear edits",
    "Needs clarification",
    "Likely duplicate comments",
    "Items needing legal or leadership review"
  ],
  "instructions": [
    "Reference page numbers whenever possible.",
    "Do not smooth over uncertainty.",
    "Use plain English."
  ]
}

Quality checks

• Verify that every change in the log points back to the correct page and section.
• Make sure uncertain handwriting is labeled uncertain rather than silently normalized.
• Check that the log separates simple wording edits from policy decisions that still need leadership approval.

Common failure modes and fixes

• The model misses margin notes: Retake the image with more space around the page edges or use a scanner mode that keeps the full page visible.
• Handwriting gets normalized into the wrong meaning: Ask for uncertainty flags and compare the log against the original photo before updating the master policy.
• The output becomes a rewritten policy: Tell the model to extract edits into a revision table only.
• Pages get mixed up: Rename or upload them in order and ask the tool to preserve that order in the final log.

Sources Checked

• Google Gemini Apps Help, Upload & analyze files in Gemini Apps - Android. https://support.google.com/gemini/answer/14903178?co=GENIE.Platform%3DAndroid&hl=en. Accessed 2026-03-27.
• OpenAI Help Center, ChatGPT Image Inputs FAQ. https://help.openai.com/en/articles/8400551-chatgpt-image-inputs-faq. Accessed 2026-03-27.
• Anthropic Help Center, Uploading files to Claude. https://support.claude.com/en/articles/8241126-uploading-files-to-claude. Accessed 2026-03-27.
• Candid, Getting started on a responsible AI use policy for nonprofits. https://candid.org/blogs/how-to-create-responsible-ai-use-policy-for-nonprofits/. Accessed 2026-03-27.
• BoardEffect, Nonprofit leaders share their thoughts on AI. https://www.boardeffect.com/blog/leaders-thoughts-ai/. Accessed 2026-03-27.

Quarterly Refresh Flag

Review this article by 2026-06-25. Re-check product features, upload flows, and nonprofit workflow references before updating or republishing.