StatementEdge

Free tool

Bank from statement detector — name the bank from the PDF

Drop a PDF statement and we'll match it against ~100 bank layout fingerprints — covering the US, UK, EU, India, Canada, Australia, Asia, the Middle East and the major fintechs. We show you the top candidate, the signals that triggered the match, and the confidence.

The PDF is parsed with pdf.js inside your browser. The file never leaves your device.

Runs in your browser
No upload, no signup. Your file never leaves your device.
Free and unmetered
Use it as often as you need. No daily quota, no credit card.
EU-built, GDPR-first
Hosted in Frankfurt. Built by a small EU team that takes privacy seriously.

What this is, and what it isn't

This is a layout-fingerprint detector. It reads the text layer of your PDF, scans it against a curated list of bank signals (brand names, customer-service URLs, SWIFT/BIC codes and proprietary terminology like "Net Banking" or "Sort Code"), and ranks the matches. The bank with the highest combined signal score wins.

It is notOCR. If your PDF is a scanned image with no text layer, the in-browser detector has nothing to read; we'll say so and point you at the main converter, which runs a Gemini vision pipeline that can identify the bank without relying on a text layer.

It is also not a forensic tool. The fingerprints are best-effort heuristics. A bank mentioned in a transaction description (e.g. an inbound transfer from Chase on a non-Chase statement) can pull a few points toward the wrong candidate. That's why we also show the runner-up scores — if it's close, look at the matched signals and decide for yourself.

How the scoring works

Each fingerprint contains several signals, each with a weight. The strongest signals are things that almost never appear by accident in someone else's statement: the bank's full legal name in a header, the bank's own customer-service URL, or its SWIFT/BIC code. Weaker signals are things like a casual brand mention ("Chase" without any context) — these still help when they stack up but don't carry a confident match on their own.

Confidence is reported as one of:

  • High — top score is ≥ 16 and meaningfully ahead of the runner-up. Almost always correct unless the statement explicitly references another bank.
  • Medium — top score is ≥ 10, but the runner-up is close. Treat as a strong guess.
  • Low — only a few weak signals matched. Best to verify visually.
  • None — no signals matched any fingerprint.

The ~100 fingerprints, by region

The detector covers, at minimum:

  • United States (15): Chase / JPMorgan, Bank of America, Wells Fargo, Citibank, Capital One, U.S. Bank, PNC, TD Bank, Truist, Fifth Third, Regions, Discover, Ally, Charles Schwab, American Express.
  • United Kingdom (12): Barclays, HSBC UK, Lloyds, NatWest, Halifax, Santander UK, Nationwide, TSB, RBS, First Direct, The Co-operative Bank, Metro Bank.
  • Germany (6): Deutsche Bank, Commerzbank, Sparkasse, DKB, ING-DiBa, N26.
  • France (5): BNP Paribas, Société Générale, Crédit Agricole, Crédit Mutuel, La Banque Postale.
  • Italy (4): UniCredit, Intesa Sanpaolo, Banco BPM, BPER Banca.
  • Spain (4): BBVA, CaixaBank, Banco Santander, Banco Sabadell.
  • Netherlands (3): ING, ABN AMRO, Rabobank.
  • Ireland (3): AIB, Bank of Ireland, Permanent TSB.
  • Belgium / Austria / Portugal / Switzerland (9): KBC, Belfius, Erste, Raiffeisen, Millennium BCP, Novo Banco, UBS, Credit Suisse, PostFinance.
  • India (15): HDFC, ICICI, State Bank of India, Axis, Kotak Mahindra, IDFC First, Yes Bank, IndusInd, Punjab National Bank, Bank of Baroda, Canara, Union Bank of India, Federal Bank, IDBI, RBL.
  • Singapore / Malaysia (5): DBS, OCBC, UOB, Standard Chartered, Maybank.
  • Middle East (4): Emirates NBD, ADCB, Mashreq, First Abu Dhabi Bank.
  • Australia (4): Commonwealth Bank, ANZ, Westpac, NAB.
  • Canada (5): RBC, TD Canada Trust, Scotiabank, BMO, CIBC.
  • Fintech (8): Revolut, Wise, Monzo, Starling, Chime, SoFi, Cash App / Block, PayPal.
  • Global parents (3): HSBC group, ING group, Santander group — for statements that use the parent-brand template instead of a country-specific one.

What to do once you know the bank

Drop the statement into our main converter— it produces a clean CSV, Excel, QuickBooks, Xero or Sage-ready output, reconciled to the cent. Knowing the bank in advance is a good sanity check: if the converter's detected bank disagrees with the detector, something is off (a multi-bank PDF, an unusual template, or a scanned page with bad text).

For more pre-flight checks, the statement format validator reports whether the PDF is digital or scanned, encrypted or open, and how many pages it spans. For password-protected PDFs, the PDF password remover handles unlocking in your browser.

Why detection sometimes fails

  • Scanned PDFs without a text layer.The detector reads text, not pixels. A scan returns nothing for it to fingerprint. We'll tell you the PDF looks like a scan and direct you to the vision-based converter.
  • White-label or private banking templates. Some private and business banking statements deliberately omit the parent brand from the header. The bank name may only appear on the cover page or in the footer of the last page.
  • Smaller regional banks.The fingerprint list covers the largest ~100 banks globally. If you have a statement from a smaller credit union, mutual, or community bank, we won't match it.
  • Banks mentioned but not the issuer.If your statement has a lot of transfers to/from a different bank, that bank's name will appear in transaction descriptions. The detector tries to weight unambiguous signals (URLs, SWIFT codes) more heavily, but a brand-name-only signal can still mislead.

Privacy

The detector runs entirely in your browser. The PDF is parsed locally using pdf.js (an open-source PDF library). No upload happens at any point. You can verify this by opening DevTools → Network and watching for outbound requests while the tool runs — the only outbound traffic is the lazy-loaded pdf.js worker from a CDN.

FAQ

Does this work on scanned bank statements?
Only partially. If the scanned PDF still has a text layer (some bank-issued scans do), detection works normally. If it's an image-only scan, the in-browser detector has nothing to read; we'll tell you it's a scan and direct you to our main converter, which uses a Gemini vision pipeline that does work on scans.
How accurate is the detection?
For the major banks listed, near-perfect on real bank-issued statements — the brand name, customer-service URL and SWIFT/BIC code together produce an unambiguous match. The tool reports a confidence level (high / medium / low) so you can see how strong the evidence is. It can be misled by statements that heavily reference another bank in transactions.
Why does it sometimes guess the wrong bank?
Layout fingerprints are heuristics. The most common cause of a wrong guess is a statement that mentions another bank repeatedly in transaction descriptions — e.g. inbound transfers from Chase appearing on a non-Chase statement. The runner-up scores let you see whether the match was close.
My bank isn't listed — what does that mean?
It means we don't have a fingerprint for it yet. The detector covers the ~100 largest banks globally, but there are tens of thousands of smaller institutions. The main converter at / doesn't depend on the fingerprint list — it uses vision-based parsing and works on any statement layout.
Is anything uploaded?
No. Parsing happens entirely in your browser using pdf.js. The PDF is never sent to our servers or anyone else's. You can verify by inspecting the Network tab — the only outbound requests are the pdf.js worker and font files loaded from a public CDN.
How can I get my bank added?
Send us the bank name and a redacted sample (any sensitive details blacked out) via /contact. We add new fingerprints regularly when there's a clear, repeatable signal in the layout — a brand name, a customer-service URL, or a SWIFT/BIC code.

Related tools and guides