Interactive Reading Apps
Guides
Checked
30 October 2026
By
The Companion Desk
Reading time
4 min
Sources
6

We did not run tests for this page. It explains how general chatbots work, using their makers' own documentation, and the points apply to ChatGPT, Claude and Gemini alike.

Can a general chatbot like ChatGPT recap a book without spoilers?

Sometimes, but it cannot guarantee it. A general chatbot does not know your page, what it learned about a popular book usually includes the ending, and with web search on it may also read wikis and reviews, so later events can slip into a recap or be mixed up with another book in the series. You lower the risk by naming the exact book and chapter, telling it to stop there and to leave out anything it is unsure about; when the twist matters, a page-locked tool is the safer choice.

Checked 30 October 2026 · The Companion Desk · Desk review

How does a chatbot answer "recap up to chapter 12"?

A general chatbot writes its answer from patterns it learned during training, plus whatever is in the conversation, plus (when search is on) web pages it looks up. It does not open your copy of the book, and it keeps no record of your page unless you type it. So "only up to chapter 12" is an instruction it tries to follow, not a filter that hides the rest of the story.

For a well-known book, what it learned likely includes plot summaries, reviews, wiki pages and fan threads covering the whole book and its sequels. The model has to separate "before chapter 12" from "after chapter 12" in its own memory, and that line is blurry. OpenAI's help center adds that ChatGPT's search tools are "enabled for all models by default", so a recap may also be built from web pages written by people who finished the book.

How much of the book each kind of tool can draw onFour rows, each with a book bar. No lock (general chatbot): the whole book and more. Prompt limit (you ask a chatbot to stop): the whole book, asked to stop. Chapter lock (Thus Far, Fable club rooms): up to the end of your chapter. Page lock (Magic Bookmark, Kindle Story So Far): up to your page. A dashed line marks your page across all rows.Strictness of the lockWhat the tool can see of the bookNo lock (general chatbot)whole book and moremorePrompt limit (you ask a chatbot to stop)whole book, asked to stopChapter lock (Thus Far, Fable club rooms)up to the end of your chapterPage lock (Magic Bookmark, Kindle Story So Far)up to your pageYour pageHow much of the book each kind of tool can draw onFour rows, each with a book bar. No lock (general chatbot): the whole book and more. Prompt limit (you ask a chatbot to stop): the whole book, asked to stop. Chapter lock (Thus Far, Fable club rooms): up to the end of your chapter. Page lock (Magic Bookmark, Kindle Story So Far): up to your page. A dashed line marks your page across all rows.No lock (general chatbot)whole book and morePrompt limit (you ask achatbot to stop)whole book, asked to stopChapter lock (Thus Far,Fable club rooms)up to the end of your chapterPage lock (Magic Bookmark,Kindle Story So Far)up to your pageYour page
Fig. 1How much of the book each kind of tool can draw on

What are the risks?

  1. Spoiler leakage. A later event, a character's fate or a reveal can appear in the recap, sometimes only as a hint ("little does she know...").
  2. The wrong place. Chapter numbers can differ between editions, parts can restart their numbering, and audiobook tracks do not match chapters. The chatbot may picture a different point than you are at.
  3. Invented or blended details. OpenAI says ChatGPT "can produce incorrect or misleading outputs", and Anthropic says Claude "can write things that might look correct but are very mistaken." In a recap that can mean a scene that never happened, or one borrowed from a sequel or a screen adaptation.
  4. Search results. Pages found on the web are written for people who have finished the book.
  5. Confident tone. A wrong recap reads exactly like a right one. OpenAI notes ChatGPT might sound confident even when it is wrong.

How do you ask more safely?

  1. Name the book precisely. Title, author, and the format or edition you are reading.
  2. Give your place two ways. The chapter number and the chapter title, or the last scene you remember.
  3. Draw the line out loud. Ask it to cover only events up to the end of that chapter and not to mention, hint at or foreshadow anything later.
  4. Give it permission not to know. Tell it to leave out anything it is unsure about and to say so.
  5. Turn off web search for this question if your app lets you, so it does not pull in reviews and wikis.
  6. Check against your book. If a name or scene looks unfamiliar, treat it as a possible error or spoiler and stop reading the answer.

A prompt you can copy:

I'm reading [title] by [author], [format or edition]. I have just finished chapter [number], "[chapter title]". Recap only what happens up to the end of that chapter. Do not mention, hint at or foreshadow anything after it, including twists or what happens to any character. Do not search the web. If you are not sure whether something happens before or after that point, leave it out and tell me. Keep it under 200 words, then list the main characters I have met so far, one line each.

More habits that help:

  • Avoid questions whose answer is a spoiler. "Does she survive?" gives the game away whether the answer is yes or no.
  • Ask it not to confirm or deny. "If this is answered later in the book, just say it hasn't been answered yet."
  • Use a fresh conversation per book. Earlier chats about the sequel can bleed in.
  • For less-known books, paste your own notes. Then it recaps from what you give it, not from memory. Whatever you paste goes to the chatbot's maker.

The spoiler-safe question bank has more wording you can reuse.

When is a page-locked tool the better choice?

When the book is built on twists, when it is part of a long series, when you are coming back after a long break, or when you read in print or audio. Page-locked tools limit what they use to your place in the software, not just in the prompt:

  • Kindle Story So Far recaps supported Kindle ebooks up to where you stopped, for US readers. See the Kindle profile.
  • Magic Bookmark recaps supported titles up to your exact page from a page photo, page number or audiobook phrase, on iPhone. See the Magic Bookmark profile.
  • Thus Far keeps discussion locked to your chapter, if you would rather talk with readers than read a recap. See the Thus Far profile.

Even a page-locked AI recap can contain mistakes. The lock stops it from using later pages; it does not make every sentence right. How the lock works is drawn in How page-locked recaps work. For more on avoiding spoilers across reading apps generally, see Best Book Tracker Apps.

Questions

  1. Q1

    Will ChatGPT spoil the ending if I ask for a recap?

    It might. It is not locked to your page, and what it knows about a popular book usually includes the ending. A precise prompt lowers the risk but does not remove it.

  2. Q2

    Can I make ChatGPT use only the chapters I have read?

    You can ask it to, and naming the chapter number and title helps. It is still an instruction, not a filter, so check the answer against your book.

  3. Q3

    Is Claude or Gemini safer than ChatGPT for this?

    They work the same basic way: no page lock, and answers that can be wrong. The prompting advice on this page applies to all three.

  4. Q4

    Does pasting the text help?

    It helps the chatbot stay within what you give it, because it can work from that text instead of memory. Whatever you paste is sent to the company running the chatbot.

  5. Q5

    Is there a spoiler-free recap tool for paperbacks?

    Kindle's Story So Far only covers Kindle ebooks. For print, an app that finds your page from a photo or page number, such as Magic Bookmark, does this for the titles it supports.

  6. Q6

    Did you test chatbots for this guide?

    No. This guide explains how the tools work from their makers' documentation. We did not run or time any tests.

Sources

  1. 1OpenAI Help Center: Does ChatGPT tell the truth?read 2026-10-30
  2. 2Claude Help Center: Claude is providing incorrect or misleading responses. What's going on?read 2026-10-30
  3. 3About Amazon: How Kindle helps you stay immersed in every book you read (published 2 April 2025, updated 2 July 2026)read 2026-10-30
  4. 4Magic Bookmark: Book Recaps, US App Store listing (incl. App Privacy and version history)read 2026-10-30
  5. 5Magic Bookmark website (home page, pricing and FAQ)read 2026-10-30
  6. 6Thus Far: No-Spoiler Book Club, US App Store listing (incl. App Privacy and version history)read 2026-10-30
  1. 01General chatbotsHow general chatbots work as reading companions: what they are good at, why they do not know your page, how web search and memory let spoilers in, privacy labels and prices. Checked 2026-10-30.03 Apps
  2. 02How page-locked recaps workHow an AI reading companion finds your place, matches it to the book, cuts the story off at your page and writes a recap, cast list or answer. With diagrams, limits and what each maker says.01 Guides
  3. 03Spoiler-safe question bankForty questions you can ask an AI reading companion or chatbot about a book you are mid-way through, sorted by purpose, plus the questions to avoid and safer rewrites.02 Resources
  4. 04How to check an AI recapAI recaps can be wrong even when they stop at your page. Six checks to catch invented details, the wrong place and hidden spoilers, with red flags to watch for. No invented test results.01 Guides