How to analyze a WhatsApp conversation
Reading a chat well means looking at the shape of months, not the wording of the last message. Here is the method — and what it costs you in time if you do it by hand.
Step 1 — Export the chat
In WhatsApp, open the conversation, tap the contact or group name, scroll down to Export chat and pick Without media. You get a .zip containing the conversation as plain text with timestamps. iMessage, Instagram, Tinder, Hinge and Snapchat have no equivalent export, which is why screenshots are the practical route there.
Step 2 — Measure effort balance
Who opens conversations, who extends them, who lets them end. Do it for a normal week early on and a normal week recently. A stable asymmetry — one person consistently initiating — is the single most informative measurement in a chat, and it is invisible when you only read the latest exchange.
Step 3 — Trace the conflict loop
Pick two or three disagreements and write their shape: what triggered it, how it escalated, who moved first, and whether it ended in repair or in silence someone later stepped over. Most relationships run one loop with different content each time. Naming the loop is what makes it changeable.
Step 4 — Check what happened to vulnerability
Find the moments where one of you said something genuinely honest. Note the reply: moving closer, deflecting with humour, changing the subject, going quiet. This predicts what will happen to the next honest thing you say better than any promise either of you made.
Step 5 — Compare periods
Read one early week and one recent week side by side: message length, warmth, specificity of plans, how quickly conflict resolves. Drift is easy to see in comparison and impossible to see day by day.
Doing this manually on a two-year chat takes hours, and you will unconsciously skip the parts that do not fit the story you already hold. That bias is the main reason automated analysis reads differently from your own memory.
Where Subtext fits
Subtext runs these steps automatically on a WhatsApp export (up to 10 MB, with very long chats truncated to the most recent messages and the analyzed period shown to you) or on a single screenshot from any messaging app. You get the effort balance, the recurring loops, how vulnerability was handled, and specific advice on what to do next — plus Before You Send for checking a draft reply. Conversations are processed in EU infrastructure and are not used to train models.
Analyze your chat with Subtext