People use WhatsApp to have their most important conversations. With partners, family, close friends. Over months and years, those chats become a record of the relationship.
But nobody scrolls back through 10,000 messages to understand patterns. AI can.
This is not about spying or surveillance. It is about understanding communication patterns in your own conversations:
Turn this thread into meeting notes and action items.
Analyze your chatThese insights are useful for self-reflection, couples therapy preparation, or simply appreciating a long friendship.
The distinction between insight and surveillance matters. When you export a WhatsApp chat, you are exporting a record of your own participation in that conversation. The analysis helps you understand your own communication habits as much as the relationship itself. Recognising that you consistently initiate difficult topics, or that your response times drop during stressful periods, is the kind of self-knowledge that is hard to arrive at through memory alone but straightforward to identify in the data.
WhatsApp conversations differ from most other communication channels in a few important ways. They tend to be continuous rather than episodic. They span years, not just recent weeks. They mix text, voice messages, images, and links in a single thread. And because WhatsApp is used for personal rather than professional communication, the language is closer to how people actually think and feel. All of that makes WhatsApp chat history analysis more revealing than, say, reviewing email threads.
ThreadRecap analyzes the conversation as a whole and identifies themes, patterns, and notable moments.
WhatsApp limits individual exports to 40,000 messages without media or 10,000 messages with media per export file. For long relationships, that cap is worth keeping in mind. A two-year daily conversation can easily exceed 40,000 messages, which means you may need multiple export files to cover the full history. ThreadRecap handles ZIP files up to 2 GB and can process exports containing 75,000 or more messages in total, so uploading several sequential exports from the same conversation is a practical way to get analysis covering a longer period.
To export on iOS, open the chat, tap the contact name, scroll to Export Chat, and choose whether to include media. On Android, tap the three-dot menu inside the chat, then More, then Export Chat. WhatsApp's export goes back as far as the chat history stored on your device, which for older devices or frequently cleared chats may be shorter than you expect.
A typical relationship-focused recap includes:
Tone shifts are one of the more practically useful outputs. A conversation that starts warmly and becomes clipped and transactional over twelve months is telling you something that would be difficult to notice message by message but is visible when the whole history is considered together. ThreadRecap looks at emotional register across time segments rather than just averaging the overall sentiment, which means gradual changes surface rather than being flattened by the aggregate.
The unresolved threads category identifies conversations that opened a topic, decision, or question and then moved on without a clear conclusion. These are often the source of low-grade friction in relationships. Someone raises the idea of visiting family, the conversation drifts, and the topic never gets resolved. Seeing a list of these threads can be a useful prompt for follow-up, particularly when preparing for a direct conversation or a therapy session.
Balance is measured across both message volume and conversation initiation. Two people can exchange equal numbers of messages, but if one person starts nearly every conversation, the dynamic is not as balanced as the raw count suggests. ThreadRecap reports on both dimensions, giving a more complete picture of how the conversational load is distributed.
Text messages are words. Voice messages are words plus tone, emphasis, laughter, and pauses.
When ThreadRecap transcribes voice messages, the content is included in the analysis. A voice message where someone says "I'm really proud of you" gets captured as a notable moment. A frustrated rant about work gets captured as a theme.
ThreadRecap uses OpenAI's transcription models to transcribe voice messages in .opus and .m4a formats, which are the two formats WhatsApp uses depending on device and recording conditions. Clear audio transcribes best, which is enough for the transcripts to be meaningfully included in a relationship analysis. Where background noise is heavy, accuracy drops, but the transcripts are still included rather than silently excluded. This matters because voice messages are often sent precisely when something feels important enough to say out loud rather than type, making them disproportionately significant relative to their frequency.
For couples using ThreadRecap before a therapy session, the analysis gives both partners a shared, neutral reference point. Rather than each person reconstructing events from memory, which is selective and subjective, the analysis provides a structured summary of what was actually discussed and how the tone evolved. Therapists working with communication patterns can use this as a starting point rather than spending session time on recall. It is worth noting that the analysis reflects what was written, not what was felt or left unsaid, so it complements rather than replaces the conversation itself.
ThreadRecap is not a surveillance tool. You can only analyze chats you are part of, from your own phone's export. You cannot analyze someone else's private messages.
The goal is self-reflection and documentation, not monitoring.
This constraint is built into how WhatsApp exports work at a technical level. An export can only be generated from a device that is a participant in the conversation. There is no mechanism within WhatsApp to export a conversation you are not part of. ThreadRecap's terms of use reinforce this: users confirm they are a participant in any chat they upload for analysis.
Relationship chats are deeply personal. ThreadRecap's privacy approach is designed for this:
The local unzipping step is worth understanding in detail. When you upload a ZIP file to ThreadRecap, the browser handles the decompression entirely on your device. The photo, video, and other media files in the export are never transmitted to a server. What is processed is the text content and the audio you choose to include, and from a video that means the audio track your browser extracts from it, never the picture. This means that even in a long relationship chat that includes years of shared photos, none of those images are ever uploaded. The design was deliberate: media in relationship chats is the most sensitive category of content, and removing it from the upload path entirely is a more robust privacy protection than encryption alone.
Saved analyses are encrypted at rest, and the deletion option is complete rather than soft. When you delete stored data, it is removed from ThreadRecap's storage, not archived or retained.
Pick a conversation that matters to you. Export it, upload it, and see what ThreadRecap finds. When the conversation grows, upload a newer export to update the same analysis with what changed.
Upload your export and get decisions, action items, and a clean summary you can forward in minutes.
Discover communication patterns, recurring themes, and meaningful moments in your WhatsApp conversations with AI-powered relationship insights.