How to summarize a long WhatsApp chat or group

For a few hundred unread messages, WhatsApp's own message summaries can catch you up, where they are available. For months or years of messages, you need the chat as an export first. Then you can paste it into an AI chatbot, read it yourself with notes, or use a summary built from the whole export, with every point linked to its message.

Four ways compared

WhatsApp's message summaries cover unread messages only

In June 2025 WhatsApp added message summaries, which use Meta AI to summarize the unread messages in a chat. They launched in English in the US. WhatsApp's help pages now list English, Indonesian, Portuguese and Spanish, in a limited set of countries, so the option may not be on your phone yet.

When a chat has enough unread messages, try it:

1. Open the chat.
2. Above the first unread message, tap the label that shows how many unread messages there are.
3. The first time, accept the terms and tap Continue.
4. Read the summary in the chat.

WhatsApp says only you see the summary, and that the feature is off by default.

Its limits:

• It only covers messages you haven't read. Last spring's planning, already read, is out of reach.
• The summary disappears when you scroll away or leave the chat. There is nothing to keep or forward.
• It is not available in a chat with Restricted chat turned on, because that setting blocks Meta AI features.
• WhatsApp's own note says messages generated by AI might not be accurate.

Pasting the chat into an AI chatbot

For anything longer, start with an export. The export guide shows the steps. A chatbot only needs the text, so export without media on your phone, or use Export chat on WhatsApp Web, which never includes media. WhatsApp Web also lets you pick a Custom date range, which helps if you only care about one stretch of the chat. Paste or upload the .txt from the export, then ask for a summary.

This can work for a short chat. With a long one, watch for four problems.

Length. Every chatbot has a limit on how much text it works with at once. A group with 20,000 messages of about 10 words each is around 200,000 words. Over the limit, a tool may refuse the file or use only part of it. A quick test: ask what happened in the first month and in the last month. If either answer is vague, it may not have read the whole thing.

Voice messages. In the .txt, a voice message is only a file name or “audio omitted”. Whatever was said in it is missing from the summary.

Privacy. You are handing everyone's messages to another company, so check its settings first. OpenAI, for example, says it may use content from its plans for individuals to train its models, unless you turn off Improve the model for everyone under Settings > Data controls in ChatGPT. The other people in the chat did not choose that service, so think about what the chat contains.

No links back. A chatbot's summary might say “Ana agreed to pay half” without saying where. To check, you search the chat yourself. Dates are another trap. An export writes dates the way the phone does, with the day first or the month first, and a tool that reads 03/04 the wrong way moves April 3 to March 4.

Reading the export yourself

Reading it yourself is free and private, and it is as accurate as your attention.

1. Make it easy to read. Open the export in the free chat viewer. It works in your browser, uploads nothing, and saves a printable PDF, a TXT, a CSV with one row per message, or an HTML page. A .txt also opens in any text editor: see opening a WhatsApp .txt file.
2. Go month by month. Open the CSV in a spreadsheet and filter by date or by person.
3. Search for the words that carry decisions. Try “agreed”, “OK”, “deadline”, “invoice”, “pay”, “refund”, “tomorrow”, a currency sign, and the names of people and places.
4. Keep a running list. For each point, write the date, who said it and a few words from the message. That list is your summary, with its sources.

Budget the time. At three seconds a message, 5,000 messages take about four hours. Voice messages need listening on top. WhatsApp's own voice message transcripts help, but they are made one message at a time, on your phone.

A summary built from the whole export

This is what LogMyChat does. After one payment for that chat, every voice message is transcribed in the language spoken, under the date and time it was sent. The summary is then written from the whole chat, transcripts included. It has six parts: an overview, what is still open, key moments, promises (who promised what), money mentioned and dates mentioned. Every item links to its message, and each message has a date, a time and a number like M006120. The summary is written in the chat's main language.

For this, export on your phone with media, so the voice messages come too.

A made-up example: Ravi is a freelance wedding videographer. For 11 months he has shared a group with the couple, Ade and Lina, and their planner, Sophie. The group holds 9,300 messages and 140 voice messages, about 130 minutes in all. Under “still open”, the summary lists two items:

• Extra hour of filming: Ravi quoted $300, no answer yet (M006120)
• Drone at the venue: Sophie was checking whether it's allowed (M007845)

Then Ravi uses Ask: “Did Ade and Lina agree to the extra hour?” The answer says no message confirms it. On May 2 Ravi quoted $300 (M006120), and Lina replied “let us think about it” (M006134). Both messages are linked. When the chat doesn't contain an answer, Ask says so. You can ask up to 100 questions per chat, in any language.

A chat like Ravi's is a Medium chat: $12, paid once. See every size. If the group keeps going, export it again later. You pay only the difference if it grew into a bigger size, and each chat gets up to 3 summaries in total.

The free overview needs no upload. On unlock, only the message text and the voice files are sent, over HTTPS, and the audio is discarded after transcription. The text goes to an AI text model only to write the summary and the answers. It is never sold or used to train AI, and photos and videos stay on your phone. There is more on summarizing a busy group and on keeping track of a client chat.

How to check a summary before you rely on it

Any summary, from any tool, can be wrong. It can miss a message, mix up two people or take a joke at face value. Check it before you forward it, pay on it or quote it.

1. Open the source of every point you will act on. Read a few messages before and after it. An “OK” may answer a different question.
2. Check the names. A group with two Sams, or one person writing from two numbers, is easy to mix up.
3. Check dates and amounts against the message. “The 15th” needs the month the message was written in. “Three hundred” in a voice message could be dollars or euros.
4. Look for what is missing. If you know the price changed, search “price”, and read that part yourself if the summary skips it. A summary of a two-year chat should mention the first months as well as the last weeks.
5. Check important voice transcripts against the recording. Transcripts are usually very good, but noise, accents and people talking over each other cause mistakes.
6. Ask a question you already know the answer to. If it gets that wrong, trust the rest less.

If the chat is part of a dispute, keep the original export too. Many people share the dated record with a lawyer or mediator. Ask whoever handles your case which format they need. This is not legal advice.

Common questions

Can WhatsApp summarize a whole chat?

No. Its message summaries cover only the unread messages in one chat, and only in some countries and languages. For a whole chat, export it first.

Is it safe to paste a WhatsApp chat into ChatGPT?

It depends on the chat and on your settings. OpenAI says it may train its models on content from its plans for individuals unless you turn that off under Data controls. The chat also holds other people's messages.

Does the summary include voice messages?

Not if you paste a .txt into a chatbot: voice messages appear only as file names or “audio omitted”. In LogMyChat, every voice message is transcribed first, so the summary and the answers cover what was said in them too.