How to Auto-Sort Thousands of AI Conversations into Folders
Learn how to use AI Chat Importer's Auto-Sort feature to automatically organise thousands of ChatGPT, Claude, and DeepSeek conversations into folders — using a local AI model that never uploads your data.
If you've been using ChatGPT, Claude, or DeepSeek for more than a few months, you've probably got hundreds — maybe thousands — of conversations sitting in your archive. Research threads, debugging sessions, writing drafts, random questions.
Finding anything specific means scrolling endlessly or hoping your memory of the title is close enough to search for.
Auto-Sort fixes this. It uses a local AI model to read your conversations, suggest a folder structure based on topics and themes, and sort everything automatically — without sending a single message to any external server.
This guide walks through exactly how it works and how to use it.
What is Auto-Sort?
Auto-Sort is a feature in the AI Chat Importer desktop app that analyses your imported conversations and automatically classifies them into folders.
It works by running a local AI model (via Ollama) directly on your machine. The model reads the content of each conversation, identifies topics and themes, and assigns each one to a folder from a suggested taxonomy.
Because everything runs locally, your conversation content never leaves your device — there's no cloud API call, no external AI service, no data upload of any kind.
What you need before you start
- The AI Chat Importer desktop app installed and activated
- Ollama installed on your machine or local network
- A compatible AI model downloaded —
llama3.1:8bis recommended for most users
If you haven't set up Ollama yet, follow the Ollama setup guide first — it covers installation, model selection based on your hardware, and connecting it to AI Chat Importer.
How Auto-Sort works
Auto-Sort runs in two phases before touching a single file.
Phase 1 — Discovery
First, Auto-Sort scans all your conversations and builds a picture of what topics appear across your archive. It sends batches of conversations through the local model and accumulates a list of suggested folder names — things like "Coding & Development", "Writing & Content", "Research & Analysis", or whatever themes appear in your specific archive.
It then consolidates overlapping suggestions into a clean taxonomy of up to 15 folders.
Phase 2 — Review
Before anything is sorted, you see the full suggested folder list as a review screen. Every folder is shown with example conversation titles so you can judge whether the suggestion makes sense.
You can:
- Rename any folder
- Delete folders you don't want
- Merge similar folders
- Add your own folders
Nothing is written to disk until you confirm.
Phase 3 — Classification
Once you confirm the taxonomy, Auto-Sort classifies each conversation against your approved folder list and shows you the assignments one batch at a time. You can move any conversation to a different folder before committing — and those corrections are fed back into subsequent batches so the model learns from your preferences.
Running Auto-Sort on an existing archive
If you've already imported your conversations and want to sort them:
- Open AI Chat Importer and click the settings cog in the sidebar
- Select "Auto-Sort Conversations"
- Auto-Sort will run the discovery phase and present a suggested folder taxonomy
- Review and edit the suggested folders, then confirm
- Review the conversation assignments and make any corrections
- Click "Confirm & Sort" — your conversations are moved into folders
The whole process for a few hundred conversations typically takes a few minutes depending on your hardware.
Auto-Sort on large imports (Batch Queue)
If you're importing a large file — 60 or more conversations — Auto-Sort intercepts the import and offers to sort as it goes.
Instead of importing everything to the inbox first and sorting later, the Batch Queue processes conversations in groups of 50, running classification in the background across sessions. You don't have to do it all in one sitting.
The queue persists between app restarts, so if you close the app mid-sort it picks up where it left off.
Tips for better results
Use a capable model. llama3.1:8b gives significantly better classification accuracy than smaller 3B models. If accuracy matters more than speed, stick with the 8B model.
Review the taxonomy carefully. The folder names Auto-Sort suggests directly influence how conversations get classified. Vague folder names lead to vague sorting. If you rename "General" to "Personal Projects" you'll get more meaningful assignments.
Use corrections. If a conversation lands in the wrong folder during the review step, move it. Auto-Sort records these corrections and applies them to subsequent batches — so it gets more accurate as you go.
Re-sort after big imports. Auto-Sort works best when run after a large import. If you import a small batch of new conversations, you can simply drag them into folders manually — the Batch Queue is designed for large volumes.
Privacy — how Auto-Sort handles your data
The entire process runs on your machine. When Auto-Sort sends a conversation to the AI model, that request goes to Ollama running locally — not to any cloud service.
No conversation content is transmitted externally at any point. The only network calls AI Chat Importer makes are to verify your license key on activation — never to process your data.
You can verify this yourself: open the Network tab in your system monitor while Auto-Sort is running. You'll see local requests to your Ollama instance and nothing else.
Getting started
Auto-Sort is available in the AI Chat Importer desktop app for Windows — £29 one-time, no subscription.
If you haven't set up Ollama yet, the Ollama setup guide covers everything from installation to connecting it to the app.