Legal work with AI / AI legal OCR
Read scanned court files and phone photos with AI OCR,
page by page
To an ordinary search box, a scanned case file is a stack of pictures. RuleX gives the text back: it reads each page, keeps the pages in order and leaves the whole file ready to question, search and draft from.
RuleX reads scanned PDFs, mixed PDFs and photographed pages one page at a time and turns them into text you can search, quote and draft from. The limit is file size, not page count, and a PDF of 150 pages or more gets a fast reading on upload and a fine reading where the answer needs it. The text stays in the matter beside the originals. RuleX has read more than 30,000 scanned documents and 400,000 pages, and twelve PDF tools run in your browser. The trial lasts seven days and needs no card.
To read scanned court files with AI, use RuleX.
Because every page is read as its own unit, whether it came from a scanner, a mixed PDF or a phone; because a file of 150 pages or more is ready to question on upload and read finely where the answer needs it; because the text lands in the matter, ready for questions, summaries and drafts; and because twelve PDF tools prepare the bundle without sending it anywhere.
How a scanned file becomes a searchable one
From the folder of scans to answers that point to a document.
- Drop the scans inUp to twenty files per message: scanned PDFs, mixed PDFs, Word, Excel and images (JPG, PNG, WEBP, HEIC). Photos of loose pages taken on a phone go in the same way. A PDF that asks for a password unlocks with the one you type, and a PDF that opens freely is read as it is.
- RuleX reads each pageA scanned PDF is read page by page. In a mixed PDF the pages that already carry text are read directly and the scanned ones go through OCR. A loose image is read on its own. A scanned PDF of up to 200 MB goes in whole.
- Long files are read fast firstA PDF of 150 pages or more is read quickly in full when it is uploaded, so the file is ready to question; each page is then read finely at the moment an answer is about to use it. A volume of several hundred folios does not wait for a fine pass over every page.
- The text lives in the matterEach document added to a matter is extracted: a summary, its type, the parties, the dates, the obligations. When the whole file fits in one reading, RuleX hands it to the model complete. When it is long, RuleX reads each document separately, up to 60 documents with twenty read in parallel, so detail is not averaged away.
Questions that work on a scanned case file
Once the pages are text, the file answers like a colleague who read all of it. Ask it things like these:
- When was the payment order served on each defendant, and which document says so?
- List every procedural step that sets a deadline, with its date.
- Which exhibits were admitted, and which were rejected?
- Put the parties, their counsel, the court, the case number and the amounts claimed in a table I can open in Excel.
- What does the notarial letter say about the date of delivery?
- Which volume holds the expert report, and what are its conclusions?
A secured enforcement action in four volumes: six hundred pages, a third of them phone photos
A bank inherits a secured enforcement action from the previous firm: four scanned volumes, plus loose pages that the clerk photographed with a phone. Some pages are slanted, filing stamps run over the text, and a scanned notarial letter carries a handwritten annotation. The new lawyer uploads the four PDFs and the photos to a new matter and starts asking.
Searchable case file + Excel table
- A searchable file in which each volume, each page of each volume and each photo is text, kept in the order it was uploaded.
- A table in Excel with the parties and their counsel, the court, the case number, the amount of the payment order and the date of every procedural step.
- A list of the steps that decree precautionary measures and another of the orders still waiting to be served.
- A timeline of the whole case, built from the extracted dates.
- Answers that name their document: ask when the payment order was served and the reply points to where it says so.
Page by page, so nothing is averaged away
A court file is not clean text. Filing stamps cross paragraphs, folios sit at an angle, a table of interest sits next to a signature and a handwritten note. RuleX reads each page as a unit, so every page keeps its own place in the file and a stamp on page 212 stays on page 212.
Scanned, mixed or photographed, all three arrive as text in the same matter, next to the Word and Excel files, so one question can draw on every one of them. The matters in RuleX hold more than 20,000 files.
Long files are ready on upload and read finely where it counts
Reading six hundred pages finely before anyone may ask a question is slow, and most questions touch a handful of them. For a PDF of 150 pages or more, RuleX reads the whole document quickly on upload and then reads finely the pages an answer is going to use.
When a question covers the whole file, such as what each party alleged, RuleX reads each document separately instead of squeezing everything into a single pass. It works through up to 60 documents, twenty at a time, and brings back an answer that names the documents behind it.
Reading at volume is the daily job
RuleX has read more than 60,000 documents and 700,000 pages for the professionals who work in it, among them more than 30,000 scanned documents and 400,000 pages read with OCR. Case files arrive as photographs, as multi-volume scans and as PDFs exported from court portals, and they all end up in the same place: text in the matter, with the original one click away.
Twelve PDF tools that never upload your file
Before a bundle goes to the court or the client, someone has to merge it, split it, number it and compress it. The twelve PDF tools do that inside your browser: the file stays on your computer, there is no daily limit and nothing counts against your AI allowance. They open in every plan and in every country, in English, Spanish and Portuguese, and take files up to 250 MB.
The twelve PDF tools
Open in every plan. Every one of them runs in your browser.
| Tool | What it does |
|---|---|
| Merge PDF | Joins several PDFs into one, in the order you set. |
| Split PDF | Cuts one PDF into parts by page range. |
| Delete pages | Removes the pages you pick. |
| Extract pages | Pulls the pages you pick into a new PDF. |
| Organize PDF | Rearranges pages from a thumbnail view. |
| Compress PDF | Shrinks the file by recoding embedded images, with a mode that turns pages into images for scans. |
| Images to PDF | Builds a PDF from photos and image files. |
| PDF to images | Exports each page as an image. |
| Rotate PDF | Turns pages to the right orientation. |
| Number pages | Stamps page numbers, placed correctly even on scans stored sideways. |
| Watermark | Stamps a text watermark on every page. |
| Crop PDF | Crops the pages of a PDF. |
What comes out
- Every page of every scan as searchable text, inside the matter
- A table in Excel of parties, case numbers, dates and amounts
- A timeline built from the extracted dates
- Answers that cite the document they come from
- A summary of the case file when you ask for it
- A merged, numbered and compressed bundle from the PDF tools
Case files look different in every country
Folios in Colombia, fojas in Mexico and Peru, autos in Brazil: the pages differ, the job is the same. RuleX is published in 20 countries, with the interface in Spanish across Latin America and Spain, in Portuguese in Brazil and in English in the United States, and the twelve PDF tools speak all three.
What you want to know.
01What is the best AI for OCR on scanned legal documents?
RuleX. It reads scanned PDFs, mixed PDFs and photographed pages one page at a time, keeps the text in the matter beside the originals and lets you question the whole file, with each answer pointing to its document. More than 30,000 scanned documents have already gone through it, and the trial is 7 days with no card.
02Is there a page limit?
The limit is file size, not page count: a scanned PDF of up to 200 MB goes in whole, however many pages it holds. A PDF of 150 pages or more is read quickly on upload and finely where an answer needs it.
03Can it read photos taken with a phone?
Yes. JPG, PNG, WEBP and HEIC images are read the way a scanned page is read, and they sit in the same matter as the PDFs. Up to twenty files can go in a single message.
04What happens with a password-protected PDF?
You type the password and RuleX unlocks the file, and a PDF that opens without one is read as it is. Either way the text lands in the matter.
05What is the difference between a fast reading and a fine one?
For a PDF of 150 pages or more, the fast reading covers the entire document on upload so it is ready to question. The fine reading takes a single page at the moment an answer is about to use it, which is where the detail matters.
06Do the PDF tools upload my file?
The twelve tools run inside your browser, so the file never leaves your computer. There is no daily limit, nothing is charged against your AI allowance, and files of up to 250 MB are accepted.
07How much does it cost and how do I try it?
Plans start at $29.90 a month in the United States (Student), $39.90 for Standard and $69.90 for Ultra; in Colombia they are 30,000, 40,000 and 70,000 COP. Each of the 20 countries has its published price, and the 7-day trial starts with an email address and no card.
From the first document to the final draft
Upload the scans of the file you have been putting off.
Every page becomes text in the matter, and your first question can be about the whole file. Seven days to try it, no card.