How to Build a Privilege Log From Scanned Documents

To build a privilege log from scanned documents, give every withheld document a control number, run OCR so the pages can be searched, and fill each log entry by reading the page: date, author, recipients, document type and a neutral description. OCR text helps you find and draft. A lawyer still makes every privilege call and reads every entry.

The work is slower than logging email, because paper carries no fields to export. This post sets out what the federal rules ask a log to say, what a scan can and cannot supply, and how to keep the log consistent with the documents you produce. It explains document handling and is not legal advice. Log requirements vary by court, judge and agreed protocol, so read the ones that bind your case.

What does a privilege log have to say about each document?

In federal court, a log has to make the claim expressly and describe each withheld document well enough for the other side to assess the claim, without giving away the privileged content. The rule names no columns. Local rules, standing orders and the parties' protocol may, and those are what you build the spreadsheet from.

The source is Rule 26(b)(5)(A) of the Federal Rules of Civil Procedure. A party that withholds otherwise discoverable information as privileged or as trial-preparation material must "expressly make the claim" and must describe the nature of what was withheld in a manner that "will enable other parties to assess the claim."

The advisory committee's note to the 1993 amendment, printed with the rule, explains how much detail that means, and it is frank that the answer changes with volume: "Details concerning time, persons, general subject matter, etc., may be appropriate if only a few items are withheld, but may be unduly burdensome when voluminous documents are claimed to be privileged or protected, particularly if the items can be described by categories." The same note warns that withholding without notice "may be viewed as a waiver of the privilege or protection."

Some courts spell out the fields. In the Southern and Eastern Districts of New York, Local Civil Rule 26.2 requires, for a withheld document, the type of document, its general subject matter, its date, and its author, addressees and other recipients, with their relationship to each other where that is not apparent. The person asserting the claim must also identify the nature of the privilege. That is one court's list. Yours may ask for more, less or something arranged differently.

A request for production adds one more requirement on the response itself. Under Rule 34(b)(2)(C), "an objection must state whether any responsive materials are being withheld on the basis of that objection." The log is where the other side then looks to see what those materials are.

Why are scanned documents harder to log than email?

Scanned documents are harder to log because nothing about them is recorded except the picture. An email carries its sender, recipients, date and subject as data, and review software can pour those into a draft log. A scanned letter carries the same facts only as ink on the page, and someone has to read them off.

The gap shows up field by field:

Three consequences follow for a paper set.

First, document boundaries are yours to decide. A scanner produces one long file per box or per stack. Where one memo ends and the next begins, and whether a fax cover sheet belongs with the pages behind it, has to be settled before anything can be logged, because a log lists documents, not pages.

Second, the page may not say who wrote it. Unsigned drafts, handwritten notes and file memos with initials only turn up in old files. The log needs a name or an honest "author not identified on the document", and a lawyer may need to ask the client.

Third, you cannot sort or filter until OCR has run. A text layer lets you search the withheld set for a lawyer's surname or a firm's letterhead, which is how you find the documents and group them. It does not fill the log. The broader difference between text pulled from electronic files and text recognized from images is explained in the overview of extracted text versus OCR text in e-discovery.

Which withheld scanned documents belong on the log at all?

A document belongs on the log when it is responsive, would otherwise have to be produced, and is being withheld or redacted on a claim of privilege or work product protection. Documents that are not responsive stay off. The parties' protocol or a court order may exclude whole categories, so read it before logging anything.

The committee note to Rule 26(b)(5) makes the first point itself: the obligation "applies only to items 'otherwise discoverable.'" A privileged letter about an unrelated transaction, sitting in the same box, is not withheld from anything.

A protocol may also carve out categories by agreement. One might, for example, excuse the parties from logging communications with litigation counsel dated after the complaint was filed. Do not assume any such exclusion. It exists only if your order, protocol or local rules say so. The federal rules expect the subject to be discussed early: the discovery plan under Rule 26(f)(3)(D) must state the parties' views on privilege claims, "including the timing and method for complying with Rule 26(b)(5)(A)."

The form of the log is also open to agreement in some courts. Local Civil Rule 26.2 in the Southern and Eastern Districts of New York says that, unless the parties agree or the court orders otherwise, where privilege is asserted on the same basis for multiple documents, "it is presumptively proper to provide the information required by this rule by group or category." A category entry suits paper well when a file holds, say, forty near-identical cover letters from the same lawyer to the same client. Whether you may use one is for your court and the other side.

Paper raises scope questions that email does not:

• Redacted documents. A page produced with a privileged paragraph blacked out is partly withheld. Check whether your protocol wants it logged, with its Bates number, and marked as redacted.
• Attachments and enclosures. A privileged cover letter may enclose a contract that is not privileged in itself. Decide with the lawyer whether the enclosure is produced, and log only what is held back.
• Duplicates. A box may hold the signed original, a file copy and a copy with someone's notes on it. The annotated copy is a different document and may need its own entry.
• Handwritten notes on a produced page. A lawyer's margin note on an otherwise ordinary invoice may lead to a redaction and a log entry for that page.

Privilege log columns a scanned letter or memo can fill

A scanned letter or memo can usually fill the date, author, recipients, copyees, document type, page count and a subject description, all read from the face of the page. It cannot fill sent times, file names, file paths or blind copies. Custodian comes from your scanning records, not from the image.

Where to look on the page for each column:

A worked entry

Suppose the file of a fictional company, Harlow Freight, holds a two-page letter on a law firm's letterhead. An entry might read:

Every fact in that entry is visible on the letter except the two that matter most. That Ms. Whitcombe was acting as the company's lawyer, and that the letter is privileged, are conclusions a lawyer reaches. The roles in brackets come from a list of names the legal team maintains.

Enter one row per document on the log, and give that row the page range for the document. For each item withheld in discovery, record the date, the author, the recipients, and the privilege or work-product doctrine label counsel gave you. If a set will be reviewed in camera, keep the image with your text so the description matches the scan. Take the label from the lawyer's instruction.

Using OCR text to draft faster

Converted text speeds the drafting in three ways. You can search the set for known lawyers' names, firm names and words such as "privileged" to find candidates. You can copy a dateline or address block into the spreadsheet in place of retyping it. And a table export gives you a starting grid. In Schema Law, Excel and CSV exports are included on every plan, and the export formats section of the site lists all eight. Nothing in the product decides whether a document is privileged, and no search result does either.

Checking a privilege log drafted from OCR text

Check the log by having a lawyer read every entry against the page image, never against the OCR text alone. The check has two halves: that the facts in each row match the document, and that the description supports the claim without disclosing the advice. The lawyer who signs the discovery response owns both.

OCR errors land in exactly the cells a log depends on. A date read as 2016 where the page says 2018 can move a letter outside the period in dispute. A misspelled surname stops the entry matching the name list. A dropped "cc:" line hides a third party whose presence may bear on the claim. The general routine for checking OCR output against the original applies, with the image on one side of the screen and the row on the other.

A review pass in order:

1. Count. The number of entries equals the number of withheld and redacted documents. Sort by control number and look for repeats and skips.
2. Dates. Read each date from the image. Flag undated documents as "undated" and do not guess.
3. Names. Check spelling against the name list and confirm each person's role. Make every name appear one way throughout.
4. Copyees and handwriting. Look at the foot and margins of each page for routing marks the OCR missed.
5. Basis. The lawyer confirms the claim for each entry and whether it covers the whole document or part.
6. Description. Read it as the other side would. It should say what kind of communication this is and its general subject, and stop.
7. Families. Enclosures are either logged, produced or noted, and the entry says which.

Before you upload privileged pages to any conversion service, settle the confidentiality question. The duties involved, and what to look for in a vendor's terms, are covered in the post on attorney-client privilege and AI tools. Schema Law's stated commitments on every plan are listed under security on the homepage; read them as written and decide with the lawyer responsible.

Keep the drafts. A log is often revised after a meet-and-confer, and a dated copy of each version shows what changed and when.

Withheld pages, Bates gaps and slip sheets in the produced set

Every withheld document should leave a visible, explained mark in the production, and the log should use the same reference the production does. Two methods are a slip sheet that takes the document's place in the numbering, or a separate privilege number series with no gap in the produced range. Follow your protocol if it chooses one.

Slip sheets. A slip sheet is a single placeholder page, stamped with a Bates number, that says something like "Document withheld as privileged". It sits where the document would have been. The log entry cites that number. The receiving side sees an unbroken range and knows where to look on the log.

A separate series. Withheld documents get their own control numbers, such as the HF-PRIV numbers in the example, and never enter the production range. The produced set has no holes. The log is then the only place the withheld documents appear, so its source column matters more.

Gaps. If documents were numbered first and pulled afterwards, the produced range has missing numbers. That works only if the log accounts for every one. A receiving party that tests the range will find each hole, by the arithmetic described in the section on checking for gaps and duplicates in the numbering, and will ask about any the log does not explain.

A reconciliation to run before the production leaves:

• Every Bates number in the range is a produced page, a slip sheet or a logged gap.
• Every log entry points to a slip sheet, a gap or a privilege control number that exists.
• Every redacted page in the production appears on the log, and every log entry marked "redacted" matches a page with a redaction on it.
• The text delivered for redacted pages was made after the redaction, so it does not contain the withheld words.
• No withheld document's text, file name or folder name went out in a text folder or load file by mistake.

The last two points belong to the outgoing production as much as to the log. The full pre-delivery routine, with the privilege screen and the cover letter, is in the guide on how to produce scanned documents in discovery.