Schema LawSign in with Google

Family Law Document OCR: Orders, Tax Returns, Pay Stubs

By Schema LawPublished Updated

In a family law file, OCR the documents you will search, quote or add up: prior orders and judgments, marital agreements, tax returns, pay stubs, the other side's declarations with their exhibits, and rental or business schedules. Use searchable PDF for anything you cite, Word for language you will reuse, and Excel for figures.

Family law files are built from other people's paperwork. Very little of it was typed in your office, and most of it reaches you as a scan, a fax or a phone photo. The sections below take the file one kind of document at a time.

Orders and agreements repay a direct search. Look for divorce, legal separation, and no-fault divorce so you know which judgment is on the page. In a prenuptial agreement, check the schedules that list marital property. In a support order, check every alimony or spousal support figure against the image before it moves into a disclosure or a modification draft.

Which documents fill a family law file?

A family law file fills with six kinds of document: the governing orders and agreements, income records, account statements, filings received from the other party, records of property and businesses, and records about the children. They differ in who made them, how they arrive, and which export serves them best.

KindExamplesArrives asMost useful export
Governing documentsJudgment, parenting plan, support order, premarital agreementCourt copy, often a scan of a scanSearchable PDF, Word for drafting
Income recordsTax returns, W-2s, 1099s, pay stubsClient photos, preparer PDFs, payroll printoutsExcel for figures, searchable PDF to cite
Account statementsBank, card, brokerage, retirementPaper from a drawer, subpoena returnsExcel
Received filingsDeclarations, motions, attached exhibitsServed copies, court downloadsSearchable PDF
Property and businessDeeds, appraisals, rent rolls, profit and loss statements, ledgersMixed, often bound reportsExcel for schedules, searchable PDF for the rest
ChildrenSchool, medical, agency and evaluator recordsAgency packets, handwritten logsSearchable PDF

Two of those rows have guides of their own. Statements and the reconciliation that goes with them are covered in the post on bank statements to Excel for family law. School files, evaluations, message printouts and parenting logs are covered in the post on OCR for custody records. This post covers the other four.

Where to start when the box is full

Convert in the order the case will ask questions. The governing order comes first, because every later document is read against it. Income records come second, because support turns on them and they are short. Received filings are converted on the day they are served. Property and business records wait until you know which assets are in dispute.

What do scanned tax returns and pay stubs give you once converted?

Converted tax returns and pay stubs give you numbers in cells and text you can search across years. A return becomes a set of labeled lines you can line up year against year. A stub becomes a small table of earnings, deductions and year-to-date totals that you can compare with what a party has declared.

Tax returns

A return is a stack of forms, each a grid of numbered lines with an amount at the right edge. After conversion, the line labels read well because they are printed. The amounts need more care, for three reasons:

  • The amount sits far from its label. On a wide form, recognition can attach a figure to the line above or below. Read the line number next to each amount you use.
  • Blank lines matter. A return with twenty blank lines and three filled ones can come out with the three amounts closed up together. Check which line each amount sits on as well as the amount.
  • Client copies are often rough. A return photographed page by page on a kitchen table has shadows and curl. The post on why OCR gets numbers wrong lists the digit confusions to expect.

Do not stop at the first two pages. The schedules behind them often answer the questions a family case raises. A rental property shows up on Schedule E, which the IRS says is used to report income or loss from rental real estate, royalties, partnerships, S corporations, estates and trusts. A depreciation schedule lists assets nobody mentioned. A state return may show an address.

If the copies you have are incomplete or unreadable, there is another route to the figures. The IRS lets a taxpayer view, print or download transcripts of past tax returns, along with wage and income statements. A transcript is not the return itself. The same IRS page sends anyone who needs a photocopy of a return to a separate form, and it notes that transcripts partially mask personally identifiable information while the financial data stays fully visible. Whether you can require the other party to get one depends on your court's rules.

A year-over-year sheet

Once three or more years are converted, put the same lines side by side. Type or paste each figure while looking at the scan.

Line202320242025Source pages
Wages118,400121,95097,300Return2023 p1, Return2024 p1, Return2025 p1
Business income22,1509,800(4,200)Schedule C, each year
Rental income, net6,9007,1007,350Schedule E, each year
Retirement contributions shown on W-29,0009,50023,000W-2, each year

The figures are invented. The point of the layout is that changes stand out: wages fall in the year of separation while retirement contributions more than double. The sheet does not say what that means. It tells you where to ask.

Pay stubs

A pay stub is three small tables on one page: earnings by type, deductions, and year-to-date totals. Converted to Excel, each lands in cells. What to take from a run of stubs:

  • gross pay per period and the pay frequency, which fixes the monthly figure
  • each earnings type separately: regular, overtime, bonus, commission
  • voluntary deductions, which reduce take-home pay but not earnings
  • the year-to-date gross on the latest stub, which should equal the sum of the periods before it

That last check catches a missing stub. If twelve stubs sum to less than the year-to-date figure on the twelfth, a pay period, or a bonus paid on a separate check, is absent from what you were given.

Excel, CSV and JSON exports are included on every Schema Law plan. The free plan takes one document a day of up to three pages, which is enough to see how your own client's stubs read before you decide. The pricing page has the page limits.

Prior orders, judgments and marital agreements as scans

Prior orders, judgments and marital agreements should be converted twice over: to searchable PDF, so you can find and cite a paragraph in the signed original, and to Word, so you can lift the operative language into a new motion or a proposed modification without retyping it.

These are the documents the rest of the file is measured against, and they can be the worst copies in it. An order entered eight years ago may have been copied by the clerk, faxed to a client, and photographed for you.

What to proofread in an order

Read these against the image before any converted text goes into a pleading:

  1. The case number and the date of entry. A file stamp often overlaps the caption, and the stamp is where the date lives. The post on whether OCR can read stamps over text explains what to expect.
  2. Every dollar amount and percentage. Support of 1,850 a month is not 1,350 a month.
  3. Days and times in a schedule. "Thursday" and "Tuesday" are different weeks for a family.
  4. Paragraph numbers. A dropped number shifts every later reference.
  5. Handwritten changes. Judges strike words and write in others, and parties initial changes in the margin. Recognition will usually return the typed text and miss or mangle the ink. The handwriting controls, so the converted text is wrong until you fix it by hand.

Then mark the Word copy plainly as a working copy. The signed image is the order. General guidance on turning orders into editable text is in the post on OCR for court filings.

Marital agreements

A premarital or postmarital agreement is a contract with schedules. Each party's property is usually listed in an exhibit at the back, as a table of assets and values. Convert the body to Word and the schedules to Excel. The schedules are what you compare against today's disclosures, and a row that appeared on the schedule and is missing now is worth a question.

Check that the scan is complete before relying on it. Count the pages against the page numbers printed in the footer, confirm that every exhibit named in the body is attached, and confirm the signature and notary pages are present. An agreement with a missing schedule is a different problem from a bad conversion, and you want to know which one you have.

Declarations and their attached exhibits from the other side

Convert the other side's declarations and exhibits to one searchable PDF per filing on the day they arrive, keeping the page order exactly as served. A declaration is clean typed text. Its exhibits are scans of everything else, and they are where the checking goes.

A received declaration might be six typed pages followed by forty pages of exhibits: text message screenshots, a lease, three bank statements, photographs of a kitchen, a letter from a teacher. If the filing came as an electronic download, the typed pages may already have a text layer while the exhibits have none. Search for a word you can see on an exhibit page to find out.

A response table

Once the declaration is converted, copy each numbered paragraph into a table. This is the working document for your client's reply.

Para.What it saysExhibit citedOur responseOur proof
7Respondent has not paid the mortgage since MarchEx. C, p. 2Paid April through July from the joint accountStatements, Joint_3382_2026-04 to -07
12Petitioner earns 4,200 a monthnoneStubs show 5,150 average over six monthsPay stub sheet, rows 2 to 14

The entries are invented. The discipline is real: the second column is copied from the converted text, not paraphrased from memory, so the reply answers what was actually sworn.

Checking their exhibits against their text

With the whole filing searchable, test the declaration against its own attachments.

  • Search each exhibit letter. An exhibit cited and not attached, or attached and never cited, is worth noting.
  • Search each dollar figure in the declaration and see whether it appears in the exhibits.
  • Search each date. A declaration that says "on June 3" over an exhibit dated June 13 has a problem.

Quote sworn statements with care. Pull the words from the converted text, then read them against the page image, because a dropped "not" is the kind of error that ends up in a reply brief.

Rental income and business records in a divorce file

Rental and business records in a divorce file are schedules: rent rolls, profit and loss statements, general ledgers, depreciation tables and appraisals. Convert the schedules to Excel so they can be totaled and compared with the tax returns, and convert the narrative reports around them to searchable PDF.

Rental property

For each property, the papers that matter are the lease or leases, a list of tenants and rents, the mortgage statement and the expense records. The list of tenants and rents is a rent roll, and the method for turning a scanned one into a workbook, including checks on the totals, is in the post on rent roll to Excel.

The comparison that follows is simple to state. Rents received according to the rent roll, rents deposited according to the bank statements, and rents reported on Schedule E should describe the same property in the same year. Put the three annual figures in one row. If they differ, the difference is a question for the party who owns the building.

A closely held business

Business records arrive in whatever form the bookkeeping software printed and somebody scanned.

RecordShape on the pageWhat to watch after conversion
Profit and loss statementIndented account names, one or more amount columnsIndentation is lost, so subtotals and detail lines look alike. Mark which rows are totals.
Balance sheetTwo blocks, assets then liabilities and equityConfirm the two sides still balance.
General ledgerLong, many pages, running balance per accountRepeated headers on every page. Negative amounts in parentheses.
Payroll registerWide table, one row per employeeNarrow columns merge. Check the column count on every page.
Business valuation or appraisalNarrative report with tables and appendicesConvert for search. The tables inside it need their own check.

A statement that balances on the scan should balance in the workbook. Total revenue less total expenses should equal the printed net income, and total assets should equal liabilities plus equity. Run those two sums before anything else. How columns go wrong, and how to repair them, is explained in the guide to OCR on exhibits with tables.

Whether a payment by the business was a personal expense, and what the business is worth, are questions for you and any expert you retain. The converted records only make the entries findable.

How do you search a family law file for one date or dollar figure?

Search a family law file for a date or an amount by trying every way the page could have printed it, across every converted document at once. A date has many written forms and an amount has several, so a single search string finds only some of the pages that mention it.

Dates

Suppose the question is what happened on March 14, 2025. The file may print that day as:

March 14, 2025     Mar 14, 2025     Mar. 14
3/14/2025          03/14/25         3-14-25
2025-03-14         14 March 2025    3/14

Bank statements often print month and day with no year. Orders spell the month out. Payroll uses two-digit years. Search the short forms first, 3/14 and March 14, and accept more results than you need.

Amounts

An amount of 2,750.00 can appear with or without the comma, with or without the cents, with a dollar sign attached, or in parentheses when negative. Search 2,750 and 2750. Then think about how recognition fails: a comma read as a period gives 2.750, and a 5 read as a 6 gives 2,760. If the figure matters and you have not found it where you expected, look by eye on the pages where it ought to be.

Setting the file up so search reaches everything

Search can only reach files that have a text layer and sit where your search tool looks.

  1. Keep one folder per matter with every converted searchable PDF in it, including the ones you thought were unimportant.
  2. Use your PDF reader's search across a folder, or your operating system's file search, which indexes the text layer of a PDF.
  3. Keep the spreadsheets in the same folder. A file search reads their cells too.
  4. Record each useful result in a running list with the file name and page, so you find it once.

The file you are searching holds tax returns, account numbers and facts about children, much of it belonging to a person who is not your client. Decide where it may be uploaded before you convert it. Schema Law's commitments are set out in the security section of its homepage, and they are worth reading alongside any protective order in the case.

Nothing here is legal advice. Disclosure duties, what may be filed publicly and what must be redacted all vary by state and by court, so check the rules where your matter is pending.

All posts in Practice areas