Chakam
Pasting a chat into ChatGPT and asking “what do you think of this” is a genuinely useful first move — it’s fast, free to try, and good at the kind of question that just needs a second opinion. It’s also not built for this specific job, and the gap shows up fast once a chat gets long or the question gets specific. Here’s the actual step-by-step, followed by exactly where it holds up and where it doesn’t.
How to Export a WhatsApp Chat
Before ChatGPT can read anything, the conversation has to leave WhatsApp as a file.
On iPhone
Open the chat, tap the contact or group name at the top, scroll down and tap Export Chat, then choose Without Media (media makes the file huge and isn’t needed for text analysis). Share it to Files, email, or directly into the ChatGPT app if it’s installed.
On Android
Open the chat, tap the three-dot menu, go to More → Export Chat, and choose Without Media. WhatsApp generates a .txt file you can share the same way.
How to Analyze the Exported Chat With ChatGPT
Once the .txt file is out of WhatsApp, upload it directly into a ChatGPT conversation (most current versions support file uploads) or, for a shorter chat, open the file and paste the raw text in. Either way, the prompt matters more than the upload method — a vague “what do you think of this chat” gets a vague answer. A specific prompt gets something usable:
- “Based on this conversation, who initiates more often — me or the other person?”
- “Are there any messages here that went unanswered? List them.”
- “Summarize the tone of this conversation over time — does it get warmer or cooler toward the end?”
Where ChatGPT Actually Works Well for This
For a short, specific question about a short chat — a few dozen messages, one clear thing you’re trying to understand — this approach genuinely works. It’s good at summarizing tone, flagging an obviously dismissive reply, or explaining a confusing message in plain language. For a single conversation you want a second opinion on, it’s a reasonable first stop.
Want a structured, private breakdown instead of a one-off chat with an AI?
Where It Falls Short
Long Chats Lose the Thread
A chat history with a real partner or an ongoing situationship can run to months of messages — thousands of lines. Pasted or uploaded as raw text, that length tends to exceed what the model can actually hold onto at once. The answer you get back often reflects the most recent or most vivid part of the conversation, not the whole thing, even when you ask about “the whole chat.”
No Memory Between Sessions
Ask ChatGPT for a trend over the last six months, close the chat, come back tomorrow to ask a follow-up — it doesn’t remember the file. Every session starts over, which means every real analysis has to be a single long conversation, re-uploading and re-establishing context each time you want a new angle on the same chat.
It’s a General Model, Not Built for This
ChatGPT has no concept of “red flag,” “ghost risk,” or “attachment style” as structured things to measure — it can discuss them if you ask, in the way a knowledgeable friend would, but it isn’t counting anything. Ask it to quantify how often someone initiates versus responds across 400 messages, and the number it gives back is an estimate, not a count — language models are not reliable at precise arithmetic over long documents, even when the answer looks confident and specific.
The Privacy Question
This is the part that’s easy to skip past. A chat export contains someone else’s private messages, not just your own — and uploading it to ChatGPT sends that conversation to a third-party service, governed by OpenAI’s data policies, not something either party in the original conversation agreed to. For a work conversation or a casual question, that’s a low-stakes trade. For a chat with a partner about something sensitive, it’s worth pausing on before uploading.
How Chakam Handles This Differently
Chakam was built specifically for this job, not adapted from a general chatbot. The chat is read and analyzed on your device — nothing is uploaded to a server by default — and instead of a single freeform answer, it produces structured reports: a Red Flag Report, response-time and initiation breakdowns, attachment-style signals, and a Ghost Risk Meter, each one an actual count across the full chat history, not an estimate from whatever fit in one context window.
chakam-report.jpg
Which One Should You Actually Use?
For a quick, low-stakes question about a short exchange — “does this text sound passive-aggressive to you” — ChatGPT is fast and free, and there’s no reason not to use it. For anything involving a real relationship history, a question that needs an actual count rather than an estimate, or messages you’d rather not hand to a third-party chatbot, a tool built specifically for chat analysis is going to get you a more reliable answer, read more privately.
Frequently Asked Questions
Is it safe to upload a WhatsApp export to ChatGPT?
It depends on what’s in it and who else is in the conversation. Your own messages are yours to share, but a chat export includes the other person’s messages too, sent to a third-party AI service without their knowledge. For a low-stakes question that’s usually a reasonable trade; for sensitive relationship content, it’s worth considering an on-device alternative instead.
Can ChatGPT analyze a group chat export?
Yes, mechanically — it reads a group export the same way as a one-on-one chat. The results tend to get less reliable as the number of participants and the chat length both grow, for the same reason a single long one-on-one chat gets less reliable: more content than the model can actually hold onto in one pass, with more voices to keep straight.
Why does ChatGPT give different answers if I ask the same question twice?
Language models aren’t deterministic by default — the same prompt can produce a differently worded, and sometimes differently reasoned, answer each time. For a subjective question (“does this sound sincere”) that’s a minor quirk. For something you want a precise count on, like how many messages went unanswered, it’s a real reason to be skeptical of any single answer it gives you. How to use AI in your relationship covers the broader tradeoffs of leaning on general AI tools for relationship questions.