# OCR Bankruptcy Documents: Schedules, Claims, Pay Advices

- Author: Schema Law (https://www.schemalaw.com/)
- Published: 2026-09-07T10:54:36Z
- Updated: 2026-10-11T19:22:20Z
- Category: Practice areas
- Canonical: https://www.schemalaw.com/blog/ocr-bankruptcy-documents

> How to OCR bankruptcy documents from intake to the claims register: creditor lists from bills, pay advices and statements for the trustee, and proofs of claim.

To OCR bankruptcy documents, sort the debtor's paper by the job each page will do: bills become a creditor list, pay advices and account statements go to the trustee, and scanned claim attachments get searched. Convert each group separately, then read every name, address, account number and dollar figure against the page before anything is filed.

The examples follow an invented consumer case. Dolores Whitcomb, a dental hygienist, brings her lawyer two grocery bags of mail and a folder of pay stubs. Her creditors, her employer and her bank are made up as well.

Bills and claim attachments should be searched for names before you type a schedule. Find each creditor and the balance you will list for the debtor in a chapter 7 or chapter 13 case. On the claim papers, find the proof of claim total and any mention of the automatic stay or a discharge. Confirm account numbers and amounts on the page before the set goes to the trustee.

## What paper does a bankruptcy debtor hand over at intake?

A bankruptcy debtor hands over whatever proves the four things the forms ask about: who is owed and how much, what comes in, what is owned and what goes out each month. On paper that means bills, collection letters, pay stubs, account statements, tax returns, titles, deeds and court papers from any lawsuit.

The federal judiciary's overview of [Chapter 7 bankruptcy basics](https://www.uscourts.gov/court-programs/bankruptcy/bankruptcy-basics/chapter-7-bankruptcy-basics) lists the information a debtor compiles to complete the official forms: "A list of all creditors and the amount and nature of their claims", the source, amount and frequency of income, a list of all property, and a detailed list of monthly living expenses. Each item on that list has paper behind it. Sorting the pile by those four headings is the first pass.

A sort of Ms. Whitcomb's two bags, by what each kind of page is for:

| Paper in the pile | What it supports | What to do with the scan |
| --- | --- | --- |
| Credit card statements, medical bills, collection letters | The creditor list and the debt schedules | Convert to text; pull names, addresses, account numbers, balances |
| Pay stubs | Income, and the payment advices the rules call for | Keep as page images; convert a copy to check totals |
| Bank and investment statements | Assets, and the statements the trustee sees | Keep as page images; convert to a workbook if transactions need review |
| Tax returns | Income history, and the return the trustee receives | Keep as page images; convert a copy for search |
| Vehicle title, deed, mortgage statement, lease | Property and secured debts | Convert to searchable PDF |
| Summons, complaint, judgment, garnishment order | Lawsuits and collection actions | Convert to searchable PDF; note case number and court |

Two neighbors hold the detail for two of those rows. Reading a return's form pages and schedules is covered in the post on [converting prior-year tax returns and their schedules](https://www.schemalaw.com/blog/ocr-tax-returns#converting-prior-year-tax-returns-and-their-schedules). When the debtor's house is already in foreclosure, the note, assignments and servicer payment history follow the method in the post on [OCR for foreclosure documents](https://www.schemalaw.com/blog/ocr-foreclosure-documents).

### Scan first, sort second, or the other way round?

Sort the paper before it goes through the scanner if you can. A pile fed in as it came out of the bag produces one long file in which a March credit card statement sits between a pay stub and a hospital bill. Six short files named by type are easier to convert, easier to check and easier to send on. If the client emailed phone photos instead of paper, save each photo under a name that says what it shows before you do anything else.

## Turning a debtor's bills and collection letters into a creditor list

Turn bills and collection letters into a creditor list by converting each one to text, then copying four things into one row per debt: who is owed, the mailing address for notices, the account number as printed, and the balance with its date. Add the page each row came from so the row can be proved later.

The list matters twice. Under [11 U.S.C. 521](https://www.law.cornell.edu/uscode/text/11/521), subsection (a)(1)(A) requires the debtor to file "a list of creditors". [Federal Rule of Bankruptcy Procedure 1007](https://www.law.cornell.edu/rules/frbp/rule_1007), at (a)(1), requires a voluntary debtor to file with the petition "a list containing the name and address of each entity included or to be included on Schedules D, E/F, G, and H of the Official Forms." The same names and addresses then appear on the schedules themselves. How your court wants that list formatted and uploaded is a local matter, so read your own court's instructions.

A worksheet layout that holds what the forms will need:

| Column | Example | Why it is there |
| --- | --- | --- |
| Creditor name as printed | Harborline Card Services | The legal name on the bill, not the brand on the envelope |
| Notice address | PO Box 4410, Tulsa, OK 74101 | The address for correspondence, which may differ from the payment address |
| Account number as printed | ending 7731 | Stored as text so leading zeros and spaces survive |
| Balance | 4,218.66 | A number, with the statement date beside it |
| Statement or letter date | 14 Aug 2026 | Shows how stale the balance is |
| Collector or law firm, if any | Brant and Lowe Recovery | A second party that also gets notice |
| Original creditor named in the letter | Harborline Card Services | Links a collector's row to the original debt |
| Source file and page | Bills.pdf, p. 9 | Where the row can be checked |

### One debt, three letters

A single debt can show up in the pile three times: the original card statement, a letter from a collection agency, and a letter from a law firm. Do not merge them while you are converting. Give each sender its own row and fill in the "original creditor" column, then decide with the lawyer how the debt is listed. In the Whitcomb pile, Harborline's balance of $4,218.66 appears again in a Brant and Lowe letter as $4,611.02 with fees added. Both figures, both dates and both addresses belong in the worksheet.

### Addresses are where recognition hurts most

A bill can print several addresses: where to send payment, where to send disputes, a return address in the envelope window, and a separate address for bankruptcy notices in small type on the back. OCR will read all of them and cannot tell you which one to use. Search the converted text of each bill for "bankruptcy", "correspondence" and "notices" to find the address the creditor itself names, and check the back of every page was scanned.

Post office box numbers and ZIP codes are strings of digits with no context to catch a misread. Read each one against the image. The post on [why OCR gets numbers wrong](https://www.schemalaw.com/blog/why-does-ocr-get-numbers-wrong) explains which digits get confused with which.

## Pay advices and bank statements a bankruptcy trustee asks for

The pay advices and bank statements a trustee asks for should travel as clear page images, complete and in date order, with identifying numbers cut down as the rules allow. OCR helps you prepare them: a converted copy lets you confirm every pay period is present and every figure on the forms agrees with the paper.

What must be supplied comes from the Code, the rules and the trustee's own instructions. Three provisions to read in full:

- **Payment advices.** [Rule 1007](https://www.law.cornell.edu/rules/frbp/rule_1007), at (b)(1)(E), calls for "copies of all payment advices or other evidence of payment that the debtor received from any employer within 60 days before the petition was filed", with all but the last four digits of the debtor's Social Security number or individual taxpayer identification number deleted. The rule's opening words let a court order otherwise, so check whether yours has.
- **Documents at the meeting of creditors.** [Rule 4002](https://www.law.cornell.edu/rules/frbp/rule_4002), at (b)(2), lists "evidence of current income, such as the most recent payment advice" and, unless the trustee or United States trustee instructs otherwise, "a statement for each depository or investment account" for the period that includes the filing date.
- **The tax return.** The same rule, at (b)(3), sets the return, a transcript of it, or a written statement that the documents do not exist as what the trustee receives at least 7 days before the first date set for the meeting.

Your trustee may ask for more, and for a particular delivery method. Follow the trustee's written instructions over any general checklist.

### Preparing the pay advice set

1. **Lay the stubs out by pay date.** Ms. Whitcomb is paid every two weeks, so a 60-day window should hold four or five stubs. Count them before you scan.
2. **Scan at a setting that keeps small print legible.** Year-to-date columns are printed small. Zoom in on the scan and confirm you can read every digit.
3. **Convert a working copy to Excel.** Each stub's table lands on its own sheet. Type the gross, each deduction and the net into one summary tab, one row per pay date.
4. **Run the year-to-date test.** Each stub's year-to-date gross should equal the previous stub's figure plus the current gross. A break in that chain means a missing stub or a misread digit.
5. **Compare the summary to the income form.** The monthly figures on the income schedule and the means test form were built from these stubs. If they do not tie, find out why before the trustee does.
6. **Send the images, not the workbook.** The rules call for the payment advices themselves. The workbook is your own proof that the set is complete.

### Bank statements

Statements are longer and the question asked of them is different: is every page here, and does anything on them need explaining? Check that the page numbers printed on each statement run without a gap ("Page 3 of 6") and that each statement's opening balance equals the prior statement's closing balance. If the lawyer needs to review transfers or large withdrawals, the method in the post on [converting scanned bank statements to Excel](https://www.schemalaw.com/blog/bank-statements-to-excel-for-family-law#converting-scanned-statements-to-excel) turns the transaction lines into rows that can be sorted and filtered.

## Are filed bankruptcy petitions and schedules already searchable on the docket?

Some are and some are not, so test each file. A petition or schedule saved to PDF from form software carries real text and searches as downloaded. A page that was printed, signed and scanned before filing is a picture, and an attachment can be one too. Search for a word you can see before you decide.

Whether a filed PDF had to be searchable depends on the court. The Bankruptcy Court for the District of Nevada, as one example, states in its page on [making a PDF text-searchable using OCR](https://www.nvb.uscourts.gov/electronic-case-filing/resources-documentation/procedures/text-searchable-pdfs/) that "pleadings/main documents filed electronically in CM/ECF must be in text searchable PDF format" and that attachments and exhibits "should be in text searchable PDF format whenever possible." Read your own court's procedures for its rule. The judiciary's PACER help page on [how to create a searchable PDF for the court](https://pacer.uscourts.gov/help/faqs/how-do-i-create-searchable-pdf-court) describes the route that produces real text: open the document in a word processor and save it as a PDF.

A quick sort of what comes off a bankruptcy docket:

| Docket item | Test result to expect | If it fails the test |
| --- | --- | --- |
| Petition, schedules and statements generated by software | Words select and search | Nothing to do |
| The same forms filed by a debtor without a lawyer, on paper | May be a scan made by the clerk's office | Run OCR on a copy |
| Signature pages and declarations | May be scanned inserts in an otherwise searchable file | Run OCR on those pages |
| Exhibits to motions: contracts, statements, letters | Mixed | Test every exhibit, the last as well as the first |
| Proof of claim forms and their attachments | Mixed, attachments especially | Run OCR on a copy before review |
| Orders and notices generated by the court | Test them; do not assume | Run OCR on a copy |

Test more than page one. A file assembled from several sources can have a searchable motion in front of forty scanned exhibit pages, and a search that finds a word on the first page tells you nothing about the rest. The post on [why you can't copy text from a PDF](https://www.schemalaw.com/blog/why-cant-i-copy-text-from-a-pdf) gives the three checks that settle whether a page holds text.

Keep the docket copy untouched. Save the downloaded file under its docket number, run OCR on a duplicate, and name the duplicate so nobody mistakes it for the filed version.

## Proofs of claim and the scanned attachments behind them

A proof of claim is a short form with paper stapled behind it, and the paper is where the review happens. Convert the attachments so you can search them for the account number, the dates and the amounts the form states, then compare what the creditor filed with what the debtor's own records show.

The form is Official Form 410, published on the judiciary's [Proof of Claim form page](https://www.uscourts.gov/forms-rules/forms/proof-claim-0). [Rule 3001](https://www.law.cornell.edu/rules/frbp/rule_3001) describes it at (a): "A proof of claim is a written statement of a creditor's claim. It must substantially conform to Form 410." At (c)(1), the rule says that "If a claim or an interest in the debtor's property securing the claim is based on a writing, the creditor must file a copy with the proof of claim", with an exception for certain consumer credit claims that the rule handles separately. Read (c) in full for what must accompany a claim in an individual debtor's case.

Those attachments are copies of copies: a card agreement from a servicer's archive, an account statement, an assignment or bill of sale from one debt buyer to another, a recorded mortgage. Expect scans of mixed quality, and test each one.

The claims register itself is the clerk's list. [Rule 5003](https://www.law.cornell.edu/rules/frbp/rule_5003), at (b), provides: "When it appears that there will be a distribution to unsecured creditors, the clerk must keep in a claims register a list of the claims filed in the case."

### A claims review worksheet

Build one row per filed claim and fill it from the converted text, then prove each cell against the image.

| Column | Filled from |
| --- | --- |
| Claim number and date filed | The register |
| Creditor name on the form | Form 410, part 1 |
| Name of any prior holder of the debt | The form and the attachments |
| Last four digits of the account | The form; compare with the debtor's bills |
| Amount claimed | The form |
| Amount on the debtor's schedule | Your own filed schedule |
| Difference | Formula |
| Secured or unsecured, as claimed | The form |
| Attachments present | Your reading of the pages: agreement, statement, assignment, itemization |
| Attachment pages checked by eye | Initials and date |

In the Whitcomb case, claim 4 is filed by Northgate Receivables for $4,611.02 on an account ending 7731. The worksheet row matches it to the Harborline debt scheduled at $4,218.66, shows a difference of $392.36, and notes that the attachment is a one-page account summary with no itemization of the added amount. What follows from that is the lawyer's call. The worksheet's job is to put the two numbers and the page reference side by side.

Professionals' fee applications sit on the same docket and raise a different task. Their time entries are tables, and the post on [extracting billing records to Excel](https://www.schemalaw.com/blog/extract-billing-records-to-excel) covers turning those into rows.

## Which figures in a converted bankruptcy file must be checked against the page?

Check every figure that will be typed into a form, sent to the trustee or argued from: creditor names and addresses, account digits, balances, pay amounts, dates, and claim amounts. Recognition is strongest on clean print and weaker on faint faxes, stamps over text and handwriting, and nothing in the output marks a passage as uncertain.

That last point is how the Schema Law homepage describes its own results in its [common questions](https://www.schemalaw.com/#questions), and it is a fair description of converted text in general: a wrong digit looks exactly as confident as a right one. So the check is by eye, against the image, for a defined list of items.

- **Creditor names.** One wrong letter in a name can send a notice to nobody. Read each name against the bill.
- **Notice addresses.** Street numbers, box numbers, suite numbers and ZIP codes, digit by digit.
- **Account numbers.** Compare the last four digits on the bill, on your worksheet and on any proof of claim.
- **Balances and their dates.** A balance without its statement date cannot be defended.
- **Pay advice figures.** Gross, net and year-to-date, tested with the year-to-date chain.
- **Dates of lawsuits, judgments and garnishments.** These feed questions on the statement of financial affairs.
- **Claim amounts and filing dates.** Read from the form image, not from a converted copy.
- **Handwritten pages.** A handwritten ledger from a landlord or a relative's loan note is read and typed by a person. Treat any converted text from it as a rough draft.

### Identifying numbers in the working files

The converted files now hold full account numbers and, on pay stubs and tax returns, Social Security numbers as searchable text. That is useful for finding them and a risk if the files travel. For anything filed with the court, [Rule 9037](https://www.law.cornell.edu/rules/frbp/rule_9037), at (a), limits a filing that contains such identifiers to, among other things, "the last four digits of the financial-account number" unless the court orders otherwise. Read the full rule and its exemptions before you file an exhibit. The post on [OCR for documents with Social Security numbers](https://www.schemalaw.com/blog/ocr-documents-with-social-security-numbers) covers handling the working copies.

### Where the conversion happens

Schema Law takes PDF, PNG or JPEG uploads of up to 20 MB, reads each page into text and tables, and shows the result page by page so you can compare it with the scan before you export. Word, Excel, CSV, searchable PDF and plain text are among the export formats on every plan. The plans differ in how many pages you can convert, as the [pricing page](https://www.schemalaw.com/pricing) sets out. You can delete a document, with everything parsed from it, whenever you choose.

A closing check before the petition is filed: pick five rows of the creditor worksheet at random, open the source page for each, and read all eight columns against it. If any of the five has an error, check every row.

## Try it on a real document

Upload a scanned PDF and export it as Word, a searchable PDF or plain text on the free plan. Your documents are never used to train an AI model. [Start free](https://www.schemalaw.com/api/auth/google)