What is the difference between producing extracted text versus OCR text, and when is each required in a production?

In e-discovery, “text” usually means the searchable, machine-readable layer that accompanies a produced document. Two methods generate that layer, and they are not interchangeable. Understanding the difference helps legal, records, and IT teams scope productions accurately and avoid disputes over completeness.

Extracted Text

Extracted text is pulled directly from the file’s underlying digital content. When a document is “born digital” — a word processing file, spreadsheet, email, or text-based PDF — the actual characters already exist inside the file. Extraction reads those embedded characters, so the result is highly accurate and typically captures metadata, comments, and hidden content as well.

Because it reflects the file’s true contents, extracted text is the preferred source whenever a native or near-native electronic original is available.

OCR Text

Optical Character Recognition (OCR) is used when no embedded text exists — for example, scanned paper, faxes, photographs of documents, or image-only PDFs. OCR analyzes the visual image and predicts the characters it sees. It makes otherwise unsearchable images searchable, but results depend on image quality and can contain recognition errors, especially with handwriting, poor scans, or unusual fonts.

When Each Is Required

The governing principle is that the producing party should provide text that reasonably reflects the document’s content and supports searching and review.

• Use extracted text for native electronic files, since it most accurately represents the source.
• Use OCR text for documents that exist only as images, where extraction is impossible.
• Many productions contain a mix, so a single set may legitimately combine both methods.

Specific obligations are driven by the form of production agreed upon by the parties or ordered by the court. In US federal civil matters these issues fall under the Federal Rules of Civil Procedure, but requirements vary by jurisdiction — state courts, regulators, and other countries may differ. Parties commonly address text, format, and metadata during early “meet and confer” discussions and memorialize them in an ESI protocol.

Documenting which method was applied, and why, supports defensibility if completeness is later challenged. For broader context, see e-discovery.