Upload the 7501 PDFs your broker or courier sent you and get one spreadsheet back: entry number, entry type, dates, importer of record, and every line with its HTS numbers, entered value, rate and duty, plus the fee summary and totals.
Built for duty refund and protest files, drawback claims, landed-cost bookkeeping and post-entry audits. Reads both the 02/26 layout and the older one, which number their boxes differently.
Upload a document to extract
Drop files here or click to upload
Up to 50 files
The free demo extracts the first 5 files, the rest unlock when you upgrade
Uploading...
Try one real entry summary with its continuation sheets before you plan the whole file.
If you have an ACE Portal account, export the Entry Summary Detail Report (ES-003) to Excel and stop there: it is CBP's own data and it costs nothing. If what you actually hold is a folder of Form 7501 PDFs from a broker, a courier or a company you acquired, run them through an extraction tool that reads the header blocks, every line item on the continuation sheets and the fee summary, and writes one row per line to Excel or CSV. Then check the entry number check digits and confirm that block 44 equals blocks 41, 42 and 43 added together before anyone relies on the totals.
The trap is the edition. CBP's 02/26 edition inserted four new blocks at number 21, so every box after them is numbered four higher than on the 7/21 edition. A template, a macro or an AI prompt written around "block 37 is duty" reads the duty on an older form and the rate column on a new one, and nothing errors. A refund file spanning 2025 and 2026 entries can hold both layouts, so the extraction has to read the labels, not the box numbers.
What this page is and is not. DocuOCR is document extraction software: it turns the PDFs into data. It is not a customs broker, it does not file anything with CBP and nothing here is legal or trade-compliance advice. Refund eligibility, protests and drawback claims belong with your licensed broker or trade counsel. We make sure the numbers they work from came off the paper correctly.
We laid CBP's current Form 7501 (edition 02/26, dated February 11, 2026) next to the 7/21 edition. Blocks 1 to 20 are unchanged. The new edition adds four origin fields that certain steel and aluminum imports must report, and everything after them shifts. These are the boxes an extraction template or a reviewer is most likely to point at.
| Field | 7/21 edition | 02/26 edition | Why it matters |
|---|---|---|---|
| Country of Melt and Pour | Not on the form | 21 | New on 02/26 |
| Primary / Secondary Country of Smelt | Not on the form | 22 / 23 | New on 02/26 |
| Country of Cast | Not on the form | 24 | New on 02/26 |
| Location of Goods / G.O. Number | 21 | 25 | Old 21 is now a steel origin field |
| Importer Number | 23 | 27 | Your IRS or CBP-assigned number |
| Importer of Record name and address | 26 | 30 | |
| Line Number | Column 27 | Column 31 | Start of every line item |
| HTSUS No. and AD/CVD No. | Column 29 | Column 33 | Where Chapter 99 lines appear |
| Entered Value, Charges, Relationship | Column 32 | Column 36 | |
| HTSUS Rate, AD/CVD Rate, IRC Rate | Column 33 | Column 37 | A rate, not an amount |
| Duty and I.R. Tax (per line) | Column 34 | Column 38 | The line-level dollars |
| Total Entered Value | 35 | 39 | |
| Duty (entry total) | 37 | 41 | Old 37 is now the rate column |
| Tax | 38 | 42 | |
| Other (fees, AD/CVD deposited) | 39 | 43 | Fee summary feeds this block |
| Total | 40 | 44 | Old 40 is now the declaration |
| Broker / Importer File Number | 43 | 47 |
On the 7/21 edition, block 37 is the total estimated duty and block 40 is the grand total. On the 02/26 edition, column 37 holds the HTS rate, the AD/CVD rate and the IRC rate, and block 40 is the importer's declaration with no number in it at all. A position-based OCR template built on older forms keeps returning values, they are just the wrong values, and a duty refund estimate built on them is off by the difference between a rate and an amount.
DocuOCR reads the printed labels ("Duty", "Total Entered Value", "Importer Number") and the structure of the line table rather than fixed coordinates, so the same extraction works on both layouts. CBP also lets filers print extra information anywhere on the form as long as it does not cover a required element, and broker software lays the form out a little differently from one vendor to the next. That variation is what breaks templates, and it is why this job suits AI extraction better than a zonal OCR setup.
The export repeats the header values on every line, so you can pivot by entry, by HTS number or by country of origin without lookups. On Plus you save this as a custom template and every later batch comes back with the same columns in the same order.
Filer code and entry number, entry type code, summary date, entry date, import date, port code, surety number and bond type.
Importer number, importer of record name and address, ultimate consignee, and the broker or filer with its file number.
Importing carrier, mode of transport, B/L or air waybill number, manufacturer ID, country of origin, exporting country and ports.
Line number, description, HTS numbers including Chapter 99 lines, AD/CVD case number, weights, quantities, entered value, charges and relationship.
HTS rate, AD/CVD rate and IRC rate from column 37, and the duty and tax dollars for each line from column 38.
Each fee with its class code, for example 499 for the merchandise processing fee and 501 for the harbor maintenance fee, plus AD and CVD deposits.
Total entered value and the duty, tax, other and total blocks, so you can reconcile them against the lines.
Country of melt and pour, primary and secondary country of smelt and country of cast, where the entry reports them.
Money gets claimed off these numbers, so treat the export as a draft until it passes arithmetic the form itself defines. Every check below comes from CBP's own instructions for the form.
Block 1 is eleven characters: a three-character filer code, a seven-digit number and a check digit. CBP publishes the formula in Appendix 1 of the instructions: letters in the filer code convert to digits (A=1, B=2 and so on), every second digit is doubled, the digits are summed, and the check digit is 10 minus the last digit of that sum. CBP's worked example is filer code B76 with number 0324527, which gives check digit 8. A misread digit almost always breaks it.
The instructions define the total as the sum of duty, tax and other, excluding deferred I.R. tax and any AD/CVD that was bonded rather than deposited. If the extracted total does not match the three blocks above it, one of the four was misread, or the entry carries deferred tax or bonded AD/CVD and needs a person to look.
The Duty and I.R. Tax amounts in column 38 across all lines should reconcile to the totals in blocks 41 to 43, and the entered values in column 36 to block 39. Bonded AD/CVD is shown in parentheses on the line and left out of the totals, which is the usual reason a clean read still does not tie.
Every field comes back with a confidence score. Set a threshold, send anything below it to the review screen, and approve the rest. On a stack of 7501s the fields that land in review are usually faint dollar amounts on photocopies, which are exactly the ones worth a second look.
How to choose a threshold, and why a small per-field error rate compounds across a form with dozens of fields, is covered on OCR confidence scores compared.
Drop in the 7501 PDFs, scans or faxed pages, continuation sheets included. Plus uploads up to 50 files at once and has no page limit per document.
The header blocks, every line and the fee summary are read into named fields, with a confidence score on each one.
Low-confidence values are highlighted next to the source page so a person confirms them in seconds instead of rekeying the form.
Download Excel or CSV with one row per line, or JSON. On Pro, pull the same data over the REST API into your own system.
After the Supreme Court's February 20, 2026 decision in Learning Resources v. United States, CBP set up CAPE in the ACE Portal to refund IEEPA duties. The declaration itself is only a CSV of entry numbers, but deciding which entries to list, and checking what came back, means knowing the Chapter 99 duty on each one. Importers whose shipments were cleared by a courier often hold PDFs rather than ACE access.
A drawback claim ties exports back to the import entries they came from, line by line, with the duty paid on each. Preparers rebuilding years of imports from entry summary PDFs spend most of their time on data entry, which is the part extraction removes.
Controllers booking duty, MPF and HMF to inventory cost in QuickBooks, NetSuite or Xero need the fee lines separated from the duty, per entry. The fee summary comes out with its class codes, so 499 and 501 map to their own accounts.
Checking that a broker classified and valued correctly means comparing hundreds of entries against the commercial invoices. With the lines in a spreadsheet, a pivot by HTS number and rate shows the outliers in minutes.
When a company changes brokers or is acquired, the entry history often arrives as a box of PDFs from the old provider. Extracting it once gives the new team a searchable record of past classifications and duty.
Anyone underwriting or reviewing a company's duty refund position works from the same entry-level duty figures, usually supplied as the borrower's PDFs. A consistent extraction with confidence scores makes that review repeatable.
From CBP's IEEPA duty refunds page: "Only the IOR for the listed entries or the authorized customs broker that filed the entries on behalf of the IOR may file the CAPE Declaration." The declaration is filed "using a Comma-Separated Values (.CSV file), through their web-based ACE Secure Data Portal (ACE Portal) account", and "Each individual CAPE Declaration has a limit of 9,999 entries." CBP adds that it "does not require any other information in the .CSV file", that refunds are "paid electronically via Automated Clearing House (ACH)", and that "valid IEEPA refunds will generally be issued within 60 - 90 days following acceptance of the CAPE Declaration." Eligibility has changed phase by phase since April 2026, so confirm the current rules with your broker before you build a list.
There are four ways to get entry summary data into a spreadsheet, and the cheapest one is not ours. Pick by what you actually have access to.
| Route | Who runs it | Covers | Cost | Our advice |
|---|---|---|---|---|
| ACE ES-003 export | You, with an ACE Portal account | Every entry under your importer number | Free | Use it first. It is CBP's own data. |
| Broker data export | Your current broker | Entries that broker filed | Usually free, sometimes a fee | Ask for it before paying anyone. |
| Keying PDFs by hand | Your staff or a temp | Whatever PDFs you hold | Staff hours | Fine for 30 entries, not for 3,000. |
| DocuOCR on the PDFs | You, from a browser | Whatever PDFs or scans you hold, any broker | From $49 a month | When PDFs are all you have. |
DocuOCR is priced per page. An entry summary with a handful of lines is one page; an entry with many lines adds continuation sheets. Take a file of 1,500 entry summaries averaging three pages each: that is 4,500 pages, which fits inside one month of the Plus plan at $149 (10,000 pages, custom templates, bulk upload, no page limit per document). A smaller file of 800 one-to-three page entries fits Starter at $49 for 2,500 pages, as long as no single entry runs past 25 pages. Annual billing halves both. Full plan details are on the pricing page.
Work it out with your own numbers. If someone on your team needs ten minutes to key one entry summary with its lines and fees, 1,500 entries is 250 hours before anyone checks the work. Typing errors in a duty column do not announce themselves, which is why the checks above matter whichever route you take. The same arithmetic applies to the freight side of the file: shipment paperwork is handled on bill of lading OCR, and air shipments on air waybill OCR.
Yes, two ways. If you have an ACE Portal account, run the Entry Summary Detail Report (ES-003) and export it to Excel, which is free and comes straight from CBP. If all you have are 7501 PDFs or scans from a broker or courier, an extraction tool reads the header blocks, every line and the fee summary into a spreadsheet, one row per line.
Ask the customs broker that filed the entry, including FedEx, UPS or DHL if they cleared your shipments. Request the entry summary PDFs with continuation sheets for a date range, and ask for an ACE entry summary export at the same time, because a structured export is better than a PDF when one exists. Brokers usually want the request in writing from the importer of record.
It depends on the edition. On the current 02/26 edition the line-level duty sits in column 38 and the entry totals are block 41 (duty), 42 (tax), 43 (other) and 44 (total). On the 7/21 edition the same boxes were column 34 and blocks 37 to 40. Block 37 on the new form is the rate column, not an amount.
The 02/26 edition added four blocks for steel and aluminum origin: 21 Country of Melt and Pour, 22 Primary Country of Smelt, 23 Secondary Country of Smelt and 24 Country of Cast. Every block after them moved up by four numbers, so location of goods went from 21 to 25, the line number column from 27 to 31, and the total from block 40 to block 44.
Eleven characters. CBP's instructions define it as a three-character filer code, a seven-digit entry number with leading zeros shown, and a one-digit check digit computed from the previous ten characters. Because the check digit is a formula, a misread digit anywhere in the number usually fails the check, which makes the entry number the easiest field on the form to validate.
Look for the Chapter 99 lines. IEEPA duties were reported under HTS headings 9903.01 and 9903.02, paired with the regular classification of the goods on the same line, and the duty for each shows in the Duty and I.R. Tax column. Extracting every line with its HTS numbers lets you filter the IEEPA lines and total them per entry.
CBP says only the importer of record for the listed entries, or the licensed customs broker that filed those entries for the importer, can file. The declaration is a CSV file of entry numbers uploaded through the ACE Portal, up to 9,999 entries per declaration, and CBP asks for no other data in the file. Refunds are paid by ACH.
Yes, if the scan is legible. AI extraction reads printed ABI-generated forms, photocopies and faxed pages, and returns a confidence score per field so the uncertain ones go to review instead of straight into your spreadsheet. Very faint or skewed scans still need a person to confirm the dollar amounts, which is exactly what the review queue is for.
On DocuOCR it is priced per page. Starter is $49 a month for 2,500 pages and Plus is $149 for 10,000 pages with custom templates and bulk upload. A file of 1,500 entry summaries averaging three pages each is 4,500 pages, which fits one month of Plus. Annual billing halves both prices.
When you have ACE Portal access, yes, and we would use it first. ES-003 is CBP's own data, it covers every entry filed under your importer number and it exports to Excel at no cost. OCR earns its place when the PDFs are all you have: shipments cleared by a courier, a previous broker, an acquired company, or a lender reviewing someone else's file.
Upload one 7501 above, check the lines against the paper, then run the rest of the folder. You get entry numbers you can validate, lines you can pivot and totals you can reconcile, from both the old and new layouts. If you already have ACE Portal access, pull ES-003 first and use this for whatever the report does not cover.
