.svg)
One tree for the whole document
One structure from first page to last, not page-by-page output. An exhibit stays linked to its filing, a table stays whole across page breaks.
Unlocking unstructured data to power legal AI
Legal AI struggles with scans, bundles and complex layouts. LexSelect creates document-wide structure that produces more reliably, verifiable outputs with lower downstream compute and token costs.
One node tree for the entire record, from sub-document down to section, paragraph and line. Schedules, exhibits and attachments keep their own type and page boundaries. Reading end-to-end resolves what a page-by-page parser guesses at: whether the table on page 47 started on page 34.
Try in LexStudio
LexSelect identifies each element for what it is, including struck text, signature blocks, seals, footnotes, page numbers, handwriting, images, charts, tables and passage language. Non-body elements remain distinct, and struck language is not treated as operative text.
Try in LexStudio
Citations follow the document’s own reference system, not only page coordinates. Cite subsection 12.3(a) of Schedule A, paragraph 47 of Exhibit D within an affidavit, or deposition transcript p. 4, ll. 20-21.
Try in LexStudio
Send anything from a two-page letter to a 1,800-page minute book to the API. LexSelect returns a structured node tree with every element classified, nested and linked to its source. Build on that and your outputs stay consistent, stay traceable, and cost less to run.
Try in LexStudio
Clean text
Reading-order text from any layout, including multi-column pages.
Structure & hierarchy
Headings, clauses, numbering, and nesting: the document tree, not a text dump.
Tables
Rows, headers, and merged cells, stitched across page breaks.
Handwriting
Handwritten annotations, notes, and form entries on scanned documents.
Signatures & stamps
Signature and stamp detection with page locations.
Document types
Automatic classification: the engine detects what each document is on its own.
Document relationships
Exhibits linked to filings; sub-documents inside compiled bundles detected and separated.
Source citations
Exact page and position for every extracted element.
Affidavits
Sworn statements with their exhibits attached. Numbered paragraphs are preserved and each exhibit is detected as its own sub-document, linked to its parent.
Annotated elements
Header
2
Column
1
Heading
2
Paragraph
11
Line
31
Clause
3
Subdocument
17
Explore in Playground
Court filings
Complaints, motions, orders, petitions: anything filed with a court, typed or handwritten. Form fields come back as key-value pairs and docket stamps stay out of the text.
Annotated elements
Header
1
Paragraph
8
Line
24
Handwritten text
26
Table
1
Signature
1
Explore in Playground
Transcripts
Depositions, hearings, and examinations. Every line is numbered, every question paired with its answer, ready to cite by page and line.
Annotated elements
Header
1
Page number
1
Column
1
Paragraph
11
Transcript line
25
Question
6
Answer
5
Explore in Playground
Contracts
Agreements of any kind, from employment to commercial. Clause hierarchy survives the parse: 12.3(a) stays 12.3(a), and schedules are detected as sub-documents.
Annotated elements
Header
1
Heading
1
Column
1
Paragraph
11
Clause
4
Sub-clause
4
Explore in Playground
Legislation
Bills, statutes, and regulatory issues: deeply nested numbering, definitions, and cross-references, returned as the hierarchy the text was written in.
Annotated elements
Title
1
Heading
3
Page number
1
Column
3
Paragraph
12
Footnote
8
Subdocument
3
Explore in Playground
More
Notices, wills, corporate records, court forms, and everything in between. The classification vocabulary is open: the engine detects what each document is on its own.
Explore in Playground
Corrections you can audit
Every AI correction is recorded with the original, so changes can be audited and rolled back.
The document's own reference system
Paragraph numbers, clause 32.1(b), transcript lines. Output you can cite the way lawyers cite, not only by coordinates.
Parse only what you need
Request just the node types your workflow uses: tables, clean text, structure. Cost follows what you ask for.
LexStudio / API
Developer portal
Generate an API key, read the docs, and make your first parse in minutes.
Upload your own documents in the Playground and inspect every extracted class and structure visually.
Track usage against your plan and review logs, so you can see exactly what was processed and what it cost.
LexChat
AI legal assistant
Copy clean, perfectly formatted text from any document, even scanned or complex PDFs, and generate citations for any passage.
Ask questions across all your uploaded documents and get accurate, cited answers that link back to the original source.
Use it inside Microsoft Word with our add-in: review documents, run LexChat, and insert cited text without leaving your draft.
LexEnterprise
Enterprise pipelines
Custom intake, extraction, and review workflows, configured for your process.
Delivered through the same products your team already uses, with structured data flowing into your systems.
I’m a hard sell on software, much less legal tech—we build legal tech. We ripped out our own parsing solution for LexSelect because it delivered clear improvements in the accuracy of our results. I was thrilled. Theirs is simply the best at parsing legal documents.
Shlomo Klapper
Founder & CEO, Learned Hand
Thomas Martin
Business Operations Analyst, M Resolution
LexSelect is a complete game-changer. When it comes to drafting factums and submissions, it has literally cut my prep time by a third and cut-out the most tedious and unpleasant part of the work. LexSelect allows me and my junior associates to focus on what matters, drafting comprehensive and persuasive legal arguments.
Kevin Westell
Trial Lawyer and Co-Managing Principal,
Pender Litigation
Pender Litigation
I used LexSelect to generate an event chronology from opposing counsel’s motion brief package. What would normally take me four or five hours of careful review was done in just a few minutes. The efficiency gain is incredible and lets me focus on the higher-value parts of the case.
Scott Foster
Partner,
Lewis Silkin LLP
Lewis Silkin LLP
Which product should I start with?
How is this different from general-purpose document API?
How do I get an API key?
What file formats do you support?
What output formats can I get?
How is my data handled?
Who do I talk to about larger volumes or custom requirements?
SOC 2 Type 2
audited
Every document you process on LexSelect is protected by independently audited security controls designed to keep sensitive legal data secure.
