Audit Invoices for Hidden Fees

Turn a dense, multi-page contractor invoice or medical Explanation of Benefits (EOB) into a clean, itemized spreadsheet to spot double-billing and hidden fees—using Mistral Le Chat, which is completely free forever with no credit card required, though it soft-caps at about 25 document chats per day.

Mistral Le Chat Verified March 11, 2026

Step 1 of 4

1. Create a free account

Go to chat.mistral.ai and sign up with an email or SSO. No credit card is required.

2. Upload the document

Click the attachment (paperclip) icon and upload your multi-page PDF (invoice, EOB, or estimate).

3. Run the extraction prompt

Paste the following text into the chat box:

Extract every line item from this document into a CSV format inside a code block with these exact columns: Date, Description, Quantity, Unit Price, Total, Suspected Category, and Notes.

Do not summarize or omit any rows. If you see duplicate line items, or fees that look like standard hidden markups (e.g., "admin fee", "processing fee", or "miscellaneous" over 3% of the total), flag them in the "Notes" column.

4. Export and audit

Copy the code block Mistral generates, paste it into a blank text file, and save it as audit.csv. Open it in Excel or Google Sheets, sort by the "Total" column, and review the "Notes" column for flagged errors.

What breaks

The Daily Cap: The free tier soft-caps at roughly 25 messages per day. If you are auditing a massive 50-page construction invoice that requires multiple back-and-forth prompts, you might hit the limit and have to wait until the next day.
Reset Timing: The limit resets daily, but Mistral does not publish the exact UTC hour the counter rolls over.
Scan Quality: Like all OCR, it struggles with severely blurred thermal scans or heavily creased receipts.
Ladder Tier: T0 (Free forever, no card required).

Three non-obvious things this teaches

1. The CSV Code Block Trick

Asking for CSV format specifically "inside a code block" forces the LLM to maintain strict column alignment and comma separation, preventing the "markdown table" formatting issues that break when you try to paste them into Excel.

2. The "Suspected Category" Semantic Check

Adding a "Suspected Category" column forces the OCR model to semantically evaluate each line item. This catches miscoded medical procedures (e.g., an "admin fee" that is actually a duplicate lab charge) because the model has to assign it a logical bucket.

3. The Library Benefit Dead End

Searching for "free receipt scanner app library card" reveals that while libraries offer great free tech, they don't currently offer AI document extraction benefits—meaning you have to rely on lab free tiers like Mistral's rather than assuming your local library has a software license for you.