How to Build an AI Receipt Scanner and Data Extractor

Receipts pile up in shoeboxes and email inboxes. I built an ai receipt scanner for data extraction that reads receipt images, pulls out vendor, date, amount, tax, and category, then stores everything in a structured format for accounting. No more manual data entry from crumpled paper.

Point a camera. Get structured data.

What You Need Before Starting

• Python 3.8+
• An AI API key with vision capabilities (Claude or GPT-4V)
• A receipt storage directory
• SQLite for extracted data

Step 1: Scan Receipts with AI Vision

import anthropic
import base64
import json

client = anthropic.Anthropic()

def scan_receipt(image_path):
    with open(image_path, "rb") as f:
        image_data = base64.standard_b64encode(f.read()).decode("utf-8")

    media_type = "image/jpeg" if image_path.endswith(".jpg") else "image/png"

    message = client.messages.create(
        model="claude-sonnet-4-20250514",
        max_tokens=500,
        messages=[{
            "role": "user",
            "content": [
                {"type": "image", "source": {"type": "base64", "media_type": media_type, "data": image_data}},
                {"type": "text", "text": """Extract from this receipt and return JSON only:
{"vendor": "", "date": "YYYY-MM-DD", "subtotal": 0.00, "tax": 0.00, "total": 0.00, "items": [{"description": "", "amount": 0.00}], "payment_method": "", "category": ""}
If a field is not visible, use null."""}
            ]
        }]
    )
    return json.loads(message.content[0].text)

Step 2: Store Extracted Data

import sqlite3
from datetime import datetime

def init_receipt_db(db_path="receipts.db"):
    conn = sqlite3.connect(db_path)
    conn.execute("""
        CREATE TABLE IF NOT EXISTS receipts (
            id INTEGER PRIMARY KEY AUTOINCREMENT,
            vendor TEXT,
            receipt_date TEXT,
            subtotal REAL,
            tax REAL,
            total REAL,
            category TEXT,
            payment_method TEXT,
            image_path TEXT,
            items_json TEXT,
            extracted_at TEXT
        )
    """)
    conn.commit()
    return conn

def store_receipt(conn, data, image_path):
    conn.execute(
        "INSERT INTO receipts (vendor, receipt_date, subtotal, tax, total, category, payment_method, image_path, items_json, extracted_at) VALUES (?,?,?,?,?,?,?,?,?,?)",
        (data.get("vendor"), data.get("date"), data.get("subtotal", 0),
         data.get("tax", 0), data.get("total", 0), data.get("category", ""),
         data.get("payment_method", ""), image_path,
         json.dumps(data.get("items", [])), datetime.now().isoformat())
    )
    conn.commit()

Step 3: Batch Process a Folder

from pathlib import Path

def process_receipt_folder(folder_path, conn):
    processed = 0
    errors = []

    for image_file in Path(folder_path).glob("*"):
        if image_file.suffix.lower() not in [".jpg", ".jpeg", ".png"]:
            continue
        try:
            data = scan_receipt(str(image_file))
            store_receipt(conn, data, str(image_file))
            processed += 1
            print(f"Processed: {image_file.name} - {data.get('vendor', 'Unknown')} ${data.get('total', 0)}")
        except Exception as e:
            errors.append({"file": image_file.name, "error": str(e)})

    print(f"Processed {processed} receipts. {len(errors)} errors.")
    return {"processed": processed, "errors": errors}

Step 4: Validate Extracted Data

def validate_receipt(data):
    issues = []
    if not data.get("vendor"):
        issues.append("Missing vendor name")
    if not data.get("date"):
        issues.append("Missing date")
    if not data.get("total") or data["total"] <= 0:
        issues.append("Invalid total amount")
    if data.get("subtotal") and data.get("tax"):
        expected = round(data["subtotal"] + data["tax"], 2)
        if abs(expected - data.get("total", 0)) > 0.02:
            issues.append(f"Total mismatch: subtotal+tax={expected}, total={data['total']}")
    return issues

Step 5: Export for Accounting

import csv

def export_receipts(conn, month, output_path):
    rows = conn.execute("""
        SELECT vendor, receipt_date, total, tax, category
        FROM receipts WHERE strftime('%Y-%m', receipt_date) = ?
        ORDER BY receipt_date
    """, (month,)).fetchall()

    with open(output_path, "w", newline="") as f:
        writer = csv.writer(f)
        writer.writerow(["Vendor", "Date", "Total", "Tax", "Category"])
        for row in rows:
            writer.writerow(row)

    print(f"Exported {len(rows)} receipts for {month}")

What to Build Next

Add duplicate detection that compares new receipts against existing ones by vendor, date, and amount. Duplicate receipts are a common bookkeeping error. Catching them automatically saves correction time later.

Related Reading

• Building Your First Automation: A Complete Guide - automation fundamentals
• Input, Process, Output: The Universal AI Framework - AI processing patterns
• Cost of Manual vs Cost of Automated - the cost of manual receipt entry