A worker with a marker pen hides private lines on a printed report before they summarise a long report with AI. Illustration made with AI.

Summarise a Long Report With AI: 6 Safe Steps

ByOlaolu

Home › Blog › Summarise a Long Report With AI: 6 Safe Steps

Last checked: 4 November 2026

You can summarise a long report with AI, but the report may contain details that are not yours to share. Hide the private details first, check what your chat tool does with your text, and read the summary against the original before you use it.

How to summarise a long report with AI, step by step

  1. Decide whether it should go in at all. Reports that hold client records, health details, contracts or anything marked confidential may not be yours to paste into a chat. If in doubt, ask the owner of the report first, or summarise it by hand.
  2. Find the data setting before you paste. Each tool has its own. Based on Anthropic’s help page (dated 3 August 2026), Claude has a “Help improve our AI models” switch under Settings, then Privacy. Turning it off stops future chats being used in future training, but anything already used stays included. Based on Google’s Gemini privacy page (updated 26 October 2026), Gemini has “Keep Activity”. With it on, chats may be used for training and some are read by human reviewers. With it off, future chats are not used for training unless you send feedback, though some data is still used for safety. Menus change, so look for the nearest matching option. Neither page states the default, so check yours.
  3. Swap private details for labels. Replace names, company names, phone numbers, addresses and amounts. Made-up example: “Amara Stores, owner Kofi Mensah, phone 0000” becomes “Client A, owner Person 1, [phone]”. Keep a private list of what each label means.
  4. Paste one section at a time and give clear instructions. Try: “Summarise this section in five bullet points for a busy reader. Use only what is written here. If something is not stated, write ‘not stated’. List any dates and labels you saw.”
  5. Check the summary against the report. Compare every date, name and number with the original. Read the original sections that matter most. AI can drop a condition or invent a link between two points.
  6. Put the real names back yourself. Do this in your own document, not in the chat.

We have not tested this prompt on a particular assistant. Other assistants may answer differently, so check the result yourself.

What can go wrong

  • Private details hide in headers, footers, file names and image text. Look there too.
  • Labels can still point to a person if the story is unusual. Remove the odd details as well.
  • A work or shared account may follow rules set by someone else.
  • A neat summary can still be wrong. See six easy checks for an AI answer.

A habit worth keeping

Before you summarise a long report, ask whether someone expects it to stay private. A contract, a staff review or a client file usually carries that duty. When you do summarise a long report, share only the labelled version, and delete the chat afterwards if your tool lets you.

What to do this week

  1. Open the privacy setting of the AI tool you use and note what it is set to.
  2. Pick one low-risk report and try the steps above.
  3. Write your own list of details you never paste, such as names, phone numbers and account details.

For notes rather than reports, see turning messy notes into a clear summary. For file types a tool can read, see getting your files ready. For deciding what AI should and should not touch, see auditing your own tasks.

Sources: Anthropic, How do I change my model improvement privacy settings? (where the setting is and what turning it off does); Anthropic, Is my data used for model training? (when chats may be used); Google, Gemini Apps Privacy Hub (Keep Activity, training and human review). Based on each vendor’s help page, checked 4 November 2026. This is general guidance, not legal advice.

Similar Posts