# OCR Court Filings: Turning Orders Into Editable Text

- Author: Schema Law (https://www.schemalaw.com/)
- Published: 2026-10-05T21:04:17Z
- Updated: 2026-10-11T19:22:20Z
- Category: Litigation support
- Canonical: https://www.schemalaw.com/blog/ocr-court-filings

> How to OCR court filings and orders into editable Word text, and how to deal with captions, line numbers, signature pages and docket headers.

To OCR court filings into editable text, upload the scanned order or motion, let the software read each page, and export Word for drafting or plain text for quoting. Expect to tidy the caption, the line numbers and the filing header by hand, and proofread every name, date and dollar figure against the page image before you reuse a word.

Court papers are harder to convert than they look. The body is plain typed prose, but it is wrapped in a caption laid out in boxes, numbered down the margin in some courts, stamped across the top by the clerk's system and finished with a signature. Each of those leaves marks in the output. The sections below take them one at a time.

## Why are court filings so often scans?

Court filings are often scans because paper still enters the record at several points. Judges sign orders by hand, self-represented parties file on paper, exhibits start life in filing cabinets, and older cases predate electronic filing. Each of those is scanned by someone, and a plain scan has no text behind the page.

The usual sources:

- **Signed orders.** An order is printed, signed in ink and scanned back onto the docket. The typed original had text. The signed copy that counts does not.
- **Paper filings scanned by the clerk.** Under [Rule 5(d)(3)(B) of the Federal Rules of Civil Procedure](https://www.law.cornell.edu/rules/frcp/rule_5), a person not represented by an attorney may file electronically only if a court order or local rule allows it. Their papers come in over the counter. One court's rule shows what happens next: the [Southern District of Florida bankruptcy court's Administrative Order 2022-03](https://www.flsb.uscourts.gov/sites/flsb/files/documents/general-orders/AO_2022-03_Amendment_of_Local_Rule_5005-1_to_Reflect_Requirement_for_Submission_of_Electronically_Filed_Documents_in_Text-Searchable_PDF_Format.pdf) describes documents "filed in conventional paper format by non-registered users and then converted to PDF by the clerk."
- **Court instructions to scan as images.** Courts do not agree on OCR. The [District of Arizona's scanning guidelines](https://www.azd.uscourts.gov/content/what-are-guidelines-scanning-documents) tell filers, "Do not use Textbridge or OCR software." A scanned attachment prepared under guidance like that is image-only by design. The page carries no date and refers to old software, so check that court's current rules before you rely on it.
- **Attachments and exhibits.** A motion drafted in a word processor is filed with twenty scanned exhibits behind it. Page one is searchable and page nine is not.
- **Old dockets.** Files from before electronic filing were scanned in bulk, if they were digitized at all, and state court records vary widely.
- **Copies served on you.** A lawyer prints a filing, signs it and sends you a scan, or forwards a conformed copy that has been faxed.

Whether a court wants filed PDFs to be searchable is a separate question with a different answer in each court. The post on [searchable PDFs for court filing](https://www.schemalaw.com/blog/searchable-pdf-for-court-filing) compares the rules of several named courts.

Check before you convert. Many filings on a modern docket were generated from a word processor and already have text. Try selecting a sentence in the body. If words highlight one at a time, copy what you need and skip OCR. 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) covers the cases where selection works but the pasted text is nonsense.

## Converting an order or motion to editable text

Convert the whole filing, export Word if you will edit the language or plain text if you only need to quote it, and proofread against the scan before reuse. The conversion takes minutes. The proofreading is the part that protects you.

1. **Get the best copy available.** A PDF downloaded from the court's own system is usually cleaner than a faxed or re-scanned service copy of the same paper. If you have both, use the docket copy.
2. **Upload it.** In Schema Law, sign in, upload the PDF in your browser and wait while the pages are read. PDF, PNG and JPEG files of up to 20 MB are accepted. The [how it works section](https://www.schemalaw.com/#how-it-works) shows the flow.
3. **Read the parsed pages on screen.** Glance at the caption, one page from the middle and the signature page. Those three show you how much cleanup to expect.
4. **Choose the export.** The table below matches each format to a job.
5. **Proofread before anything else.** Compare with the image, in this order: case number, party names, judge's name, every date, every dollar amount, every statute and rule citation, then the body.
6. **Save the converted file under a name that says what it is.** "Order_2026-03-14_OCR_unverified.docx" prevents anyone mistaking it for the court's document.

| You want to | Export | Why |
| --- | --- | --- |
| Draft a proposed order or a response using the language | Word | Paragraphs, headings and lists arrive as editable text, and each source page starts a new page |
| Paste a holding or a deadline into an email or brief | Plain text | No styles to strip out |
| Keep the order on file and be able to search it | Searchable PDF | The signed page image is kept, with text added behind it |
| Pull a schedule of dates or a fee table | Excel or CSV | Tables arrive as rows and columns |

Why that proofreading order? Errors in court papers concentrate in the short strings that matter most. "Rule 56(c)" read as "Rule 56(e)" still looks right. A case number with a transposed digit points to a different case. A deadline of the 18th read as the 13th is a calendar entry waiting to go wrong. Ordinary words that are misread tend to become non-words, which a spell checker catches. Numbers and citations do not, so check them by eye. The method is set out in the guide to [why OCR output is full of errors](https://www.schemalaw.com/blog/why-is-ocr-output-full-of-errors).

Proof the caption against the docket before you reuse a paragraph from the order or the motion. The filing names the parties, the judge, and any hearing date, and a wrong case number points at a different proceeding. An ex parte motion is often short, so check the proposed order line by line. If the clerk of court stamp did not come through, type it into your notes from the image.

The free plan handles one document a day of up to three pages and includes every export format, as all plans do, which suits a short order. Longer filings need a paid plan, and the [pricing page](https://www.schemalaw.com/pricing) lists the page limits.

## What do caption blocks and signature pages do to OCR?

Captions and signature pages break the simple top-to-bottom flow that the rest of a filing has. A caption sets parties, a column of brackets or a vertical rule, and the case number side by side. A signature page mixes ink, typed names, addresses and dates. Both come out of OCR in an order or shape you will want to fix.

### The caption

A typical caption has the court's name centered across the top, the parties stacked on the left, the case number and document title on the right, and a column of parentheses, "v." or a ruled line between them. Software has to decide whether that is two columns, a table or a single block. Common results:

- **Left and right sides interleaved.** "JANE ROE, Case No. 3:24-cv-01187 Plaintiff," where the case number has landed in the middle of the party block.
- **The caption becomes a table.** Sometimes this is the most faithful outcome, with parties in one cell and the case number in the next. It looks odd in Word but keeps things separated.
- **Decorative characters come through as text.** Rows of ")" or "§" down the middle appear as stray lines.
- **Spaced-out titles break apart.** "O R D E R" may be read as five separate letters, which no search for "order" will match.

The practical answer is not to repair the caption. For a new document in the same case, type the caption fresh from your own template and paste only the body from the conversion. Retyping a caption takes a minute and removes the risk of carrying a misread case number into a filing. How software decides the order of side-by-side blocks is explained in the post on [whether OCR can handle two-column documents](https://www.schemalaw.com/blog/can-ocr-handle-two-column-documents).

### The signature page

- **Ink signatures** are drawings. They are skipped or turn into a few random characters. Delete those.
- **"/s/" signatures** are typed text and are usually read correctly.
- **The typed name under the line** is the reliable source for who signed.
- **Handwritten dates** next to a judge's signature are the weak point. "Dated: March __, 2026" with a handwritten "14" may come through blank or wrong. Read the date from the image and type it.
- **Certificates of service** often follow the signature and contain lists of names and addresses in columns. Check them against the image if you plan to reuse the service list.

A signature page with a judge's stamp or a clerk's seal over the text has its own difficulties, covered in the post on [whether OCR can read stamps over text](https://www.schemalaw.com/blog/can-ocr-read-stamps-over-text).

## Line-numbered pleading paper

Line-numbered pleading paper puts a column of numbers down the left margin of every page. OCR reads those numbers as text, so they turn up at the start of lines, inside sentences or as a stack of digits. Plan to strip them from anything you edit, and keep them in mind when you cite.

The format is a requirement in some courts. California is a well-known example: [rule 2.108 of the California Rules of Court](https://courts.ca.gov/cms/rules/index/two/rule2_108) requires the lines on each page to be numbered consecutively, with the line numbers placed at the left margin and set apart from the text by a space or a vertical line. Other courts set their own formats, so what you meet depends on where the paper was filed.

What the numbers do to converted text:

- **Numbers glued to lines.** "14 The motion is GRANTED as to the first" at the start of each line. This is the easiest case to clean.
- **Numbers as a separate block.** The margin is read as its own column, giving "1 2 3 4" and so on in one lump before or after the body. Delete the lump.
- **Numbers that do not match the lines.** The numbers run down the page at a fixed spacing, but headings, block quotes and footnotes may be single-spaced. A number can sit between two lines of text, and the software attaches it to whichever is closer.
- **Numbers mistaken for list items.** A paragraph that happens to begin at line 1 may be turned into a numbered list in Word.
- **Margin rules read as characters.** The vertical line beside the numbers can appear as "|" or "l" or "1" at the start of lines.

To clean a Word export, use find and replace with wildcards or a pattern to remove one or two digits at the start of a paragraph, then read through for sentences that were split across lines. Work on a copy. Before you run any pattern, look for real numbers at the start of lines, such as numbered findings or "28 U.S.C.", so that you do not delete them.

For citation, the line numbers are the point of the format. A cite to "Order at 4:12-18" means page 4, lines 12 to 18. Converted text loses that link as soon as lines reflow. Take pin cites from the page image. Use the converted text for the words and the image for where they sit, as the guide to [quoting text from a scanned PDF](https://www.schemalaw.com/blog/how-to-quote-text-from-a-scanned-pdf) explains.

## Reusing language from an order in a proposed order

Reuse an order's language by pasting clean, unformatted text into your own template and checking it word for word against the signed page. Courts and opposing counsel read proposed orders closely where they track an earlier ruling. A recognition error in a quoted ruling is the kind of mistake that gets noticed.

A method that keeps you out of trouble:

1. **Start from your template, not the conversion.** Open a blank document with the right caption, fonts and margins for your court. Do not edit the converted file into shape.
2. **Paste as plain text.** Use "paste special" and keep text only, or copy from the plain-text export. This leaves behind line numbers formatted as lists, odd fonts and page breaks from the conversion.
3. **Repair line breaks.** Converted paragraphs sometimes break at the end of every scanned line. Join them, and check for words that were hyphenated across a line, such as "judg- ment".
4. **Compare with the image, aloud if it matters.** Read the pasted passage against the scan. For operative language such as amounts, deadlines and what is granted or denied, have a second person read one copy while you follow the other.
5. **Mark what you changed.** If you adapt the earlier language, make sure nothing presented as a quotation has been altered.
6. **Check how your court wants it.** Courts set their own requirements for proposed orders. As one example, [Administrative Order 2022-03 of the Southern District of Florida bankruptcy court](https://www.flsb.uscourts.gov/sites/flsb/files/documents/general-orders/AO_2022-03_Amendment_of_Local_Rule_5005-1_to_Reflect_Requirement_for_Submission_of_Electronically_Filed_Documents_in_Text-Searchable_PDF_Format.pdf) says proposed orders uploaded to its E-Orders program must be saved in a text-searchable format. Read your court's rules and your judge's instructions before you format anything.

The same routine works for other reuse: tracking a scheduling order into a calendar, lifting definitions from a protective order into a stipulation, or quoting a prior ruling in a brief. Scheduling orders deserve particular care. Every date in one is a deadline, and a table of dates is where a misread digit hides most easily. Export the table to Excel if you like, then check each cell against the image.

Two mistakes come up again and again. The first is quoting from the converted text without looking back at the scan, which puts the software's reading in the court's mouth. The second is leaving invisible leftovers in the document, such as a stray line number in the middle of a sentence or a docket header fragment at a page break. Turn on formatting marks in Word and read the pasted section once more before it goes out.

## Docket stamps and filing headers across the top of the page

Electronic filing systems stamp a line across the top of each page showing the case number, document number, filing date and page count. Clerks' offices add received and filed stamps. OCR reads all of it. The header repeats on every page of the output, and a stamp over text can hide the words beneath.

A federal filing typically carries a header along the lines of "Case 3:24-cv-01187 Document 42 Filed 03/14/26 Page 3 of 17". In converted text it has three effects.

**It repeats.** A seventeen-page order gives you seventeen header lines, each sitting between the end of one page's text and the start of the next. When a sentence runs across a page break, the header lands in the middle of it. Remove them with a find and replace on the constant part, such as the case number and "Document 42", and then rejoin the split sentences.

**It is small and sometimes overlaps.** Header text is often tiny, colored and printed over the top margin. If the original had its own header or a first line set high on the page, the two overlap, and both can be misread. Check the first line of each page against the image.

**It is useful.** The header gives the document number and filing date, which you need for a citation to the record. Record them once at the top of your working file before you delete the repeats. Two page numbers now exist for every page: the stamped "Page 3 of 17" and the page number the author printed at the bottom. They differ whenever a filing has a cover sheet, a table of contents or attached exhibits. Decide which one your court's citation practice calls for and use it consistently.

Clerk's stamps are less predictable. A "FILED" or "RECEIVED" stamp with a date may sit at an angle in the corner or across the caption. The date in it matters and is easy to misread, since stamps are often faint or partly inked. Read stamped dates from the image. If a stamp covers part of the caption or the first paragraph, expect gaps or wrong characters in exactly that spot.

Fax headers add one more layer on service copies, with a sender's number, a time and a page count along the top or bottom edge. They clean up like docket headers. The low resolution of a fax does more damage than the header itself, as the guide on [how to OCR a faxed document](https://www.schemalaw.com/blog/how-to-ocr-a-faxed-document) describes.

Whatever the clutter, the working rule is the same. Strip it from the text you edit, keep it in the searchable PDF you file away, and go back to the image for anything you cite.

## 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)