Every eDiscovery platform will review your documents. Almost none of them will quote what the privilege log costs. Here is what each pricing model does to the same 2,000-document privilege set — and what the log has to survive once it is filed.
The short answer: privilege logging is billed three ways in 2026 — $5 to $15 per entry for the drafting step alone, or $8 to $30 effective once you price the 15 to 30 minutes a contract reviewer actually spends on each one; $4 to $8 per document for the privilege review that has to happen before any entry can be written; or nothing extra, under a flat per-gigabyte platform fee that bundles both. On a 2,000-document privilege set, that is the difference between roughly $32,000 and $0 in marginal cost.
The reason that comparison is hard to find is that almost nobody publishes it. Vendor review sites rank platforms on star ratings and feature checkboxes. Vendor comparison pages are written by one of the two vendors. And privilege logging sits in the awkward middle of the buying process: it is not a headline feature, so it does not appear on the pricing page, but it is frequently the single largest line item on the invoice. This piece puts the pricing models side by side, says plainly which vendors do not publish rates, and then walks through what the log has to survive once it is filed — because the cheapest log that gets struck is not cheap.
Scroll table horizontally →
| Platform | Pricing model | How the privilege log gets billed |
|---|---|---|
| DecoverAI | $60/GB/month flat, all-in. Unlimited users, no seat fees, no contract minimum. Optional managed LegalOps at $150/hr. | AI-drafted entries with attorney QC, bundled into the platform fee. No per-entry charge. |
| Relativity / RelativityOne | Roughly $75–$150/GB/month, plus separate per-user and per-document charges. aiR is an add-on. | aiR suggests privilege flags for attorney confirmation; log drafting is typically a separate services or labour line. |
| Everlaw | Roughly $95–$150/seat/month plus data hosting. Self-serve pricing from $250/month at the small end, enterprise contracts at volume. | AI-assisted issue tagging and privilege identification; entries are typically attorney-drafted. |
| Reveal | Enterprise contract only. No public rate card. | Continuous active learning via Brainspace with automated privilege detection; logging remains labour-intensive. |
| Logikcull | Roughly $25–$60/GB/month plus per-user tiers. Publishes the most detail of the legacy set. | Basic predictive coding; privilege flagged by keyword and custodian matching, log drafted manually. |
| CS Disco | Roughly $2,000–$8,000/month per matter at the enterprise tier. | AI review assist; privilege logging largely manual. |
| Exterro | Custom enterprise pricing. | Governance-first deployment; privilege workflow varies by configuration. |
On the gaps in that table: Reveal, Exterro, and CS Disco at the enterprise tier do not publish rate cards, and the ranges shown for Relativity, Everlaw, and Logikcull are assembled from published tiers and market reporting rather than from a quote for your matter. Treat every figure here as a starting point for the conversation, not a price. That includes ours — the $60/GB/month rate is published on our pricing page and is subject to change like anyone else’s.
Privilege logging is the last thing to happen and the first thing to blow the budget, and the reason is structural: it is priced per unit of human output at exactly the point in the matter where the units are most expensive. A reviewer has to read the document, decide whether the privilege claim holds, and then write a description specific enough to let opposing counsel assess the claim without disclosing the thing being protected. That is 5 to 15 minutes of attorney or contract-attorney attention per document, every time.
Run the arithmetic on a 2,000-document privilege set — a realistic number for a mid-sized commercial matter:
That is the number to hold in your head when you read a rate card. We break the per-entry economics down further in privilege log fees: why you’re paying $10–$15 per entry, and the same pattern shows up across every other unbundled charge in discovery — processing surcharges, hourly project management, egress fees on the way out.
Nearly every vendor in the market combines at least two of the following three, which is why a quoted headline rate so rarely predicts an invoice.
Per gigabyte. You pay for the volume of data hosted. Cost is knowable at intake, because you know roughly how much you collected. The failure mode is when per-GB is only the base and the review work is billed separately on top — which is exactly how the privilege log escapes the quote. Our note on what per-GB hosting actually buys covers where the line usually falls.
Per seat. You pay for each person with access, typically $50 to $500-plus per user per month. A 15-person review team on a six-month matter is $18,000 to $36,000 in access fees before anyone opens a document. The worse effect is behavioural: per-seat pricing gives you a direct financial reason not to add the outside counsel, second-chair associate, or subject-matter expert who would resolve the hard privilege calls fastest. We argue that at length in the fee that penalizes you for adding outside counsel.
Per document and per entry. You pay for each unit of review work. This is the model that makes privilege logs expensive, because a privilege set is small in gigabytes and large in units. It is also the only one of the three where the vendor’s revenue rises with the number of hours your reviewers spend — a point worth sitting with when a vendor is advising you on review scope. See the per-document review fee for the full version.
Everlaw is the most transparent of the legacy set at the small end, with published self-serve pricing starting at $250/month, and it rates well with reviewers: 4.9 out of 5 across 87 reviews on Software Advice, with 94% recommending. Its own comparison material leans on generative AI features — coding suggestions, writing assistance, review support — predictive coding and clustering, and processing throughput of a million documents an hour, alongside a claimed number-one ranking on G2’s Summer 2026 Grid Report. What that material does not do is state a Reveal price or a privilege-log price, because it is a marketing page and was never meant to.
Logikcull publishes per-GB rates and per-user tiers, which makes it the easiest legacy platform to model in a spreadsheet, and it rates 4.6 out of 5 across 269 reviews with 87% recommending. Its privilege handling is the oldest pattern in the comparison: keyword and custodian matching to flag candidates, then manual drafting. If you are weighing it against the field, we compared it directly in seven Logikcull alternatives.
Relativity holds the enterprise position and its aiR add-on applies frontier-class models to suggest privilege flags, with attorney review required on every suggestion. The pricing structure is the complication: a per-GB base plus per-user and per-document charges, with logging typically landing as a separate services line. It is a capable stack that is genuinely hard to budget in advance without a negotiated contract.
Reveal pairs continuous active learning through Brainspace with automated privilege detection and is frequently recommended for government and regulatory work. It publishes no rate card at all, so any per-entry comparison against it is guesswork until you have a quote in hand.
CS Disco and Exterro sit at the enterprise end with per-matter and custom pricing respectively. Both can do the work; neither will tell you the price on a web page.
The honest summary is that only Logikcull and DecoverAI publish enough detail to build a genuine side-by-side, with Everlaw and Relativity partially there. Anyone presenting a complete seven-vendor per-entry price table is inventing at least three of the numbers. For the broader field comparison beyond privilege logging, see best eDiscovery software tools in 2026.
Pricing models are downstream of architecture. A platform that runs relevance and privilege as one classification pass over the same triaged pool can bundle the log; a stack that hands the matter from a processing vendor to a review platform to a privilege reviewer to a production vendor has to bill each hand-off, because each one is a different cost centre.
Scroll table horizontally →
| Workflow stage | Legacy stack | Mid-market self-serve | AI-native |
|---|---|---|---|
| Ingestion | Separate one-time fee, often $75–$150/GB; days to complete | Self-serve upload, per-GB fee | Automated upload-to-review pipeline, minutes to hours |
| Responsiveness | Keyword plus predictive coding or CAL | Basic predictive coding | Multi-model classification handling relevance and privilege in one pass |
| Privilege review | Manual attorney read, $4–$8/doc | Keyword and custodian flagging, largely manual | Recall-first AI detection, multi-model consensus, confidence scoring |
| Log drafting | Manual, $5–$15/entry ($8–$30 effective) | Manual, sometimes templated | AI-drafted from source text, attorney QC, bundled |
| Redaction & production | Production services fee plus PM overhead at $200–$300/hr | Simpler in-platform tools | Bates numbering, redaction, and export in one pass |
| Access | Per-user licensing, $50–$500+/user/month | Per-user tiers | Unlimited concurrent users, no seat fee |
“AI reduces review cost” is a claim every vendor now makes, and most of the time it means a model scores documents faster. That is not where the money is. Across every cost breakdown in this comparison, attorney and reviewer labour — not software licensing — is the dominant expense. So the only lever that matters is the one that removes attorney hours. Five mechanics do that, and they compound:
The mechanism matters more than the headline percentage, which is why we published the full accounting in where AI privilege review cost reduction actually comes from rather than asking anyone to take the number on faith.
Our standing benchmark is a 100 GB matter: roughly 250,000 documents, 15,000 responsive, 2,000 requiring privilege review and a log entry. Under a traditional vendor stack it comes to about $460,000. First-pass responsiveness review alone accounts for $375,000 of that — 250,000 documents at $1.50 each, 81% of the bill — with privilege review and logging adding the $32,000 described above on top.
The same matter under a flat all-in rate lands near $36,000. Roughly 92% of the gap traces to eliminating per-document and per-entry attorney billing, not to any hosting or storage discount, which is the part that gets misread most often: this is not a story about cheaper servers. A tax credit investigation we ran processed 30,000 documents end to end, complete privilege log included, in three days. The full arithmetic, line by line, is in the $460,000 vs. $36,000 benchmark.
A cheap log that gets struck is the most expensive outcome available, so cost comparison is only half the evaluation. Three authorities set the floor.
Rule 26(g) certification. The attorney of record signs the log, certifying after reasonable inquiry that it is complete and correct. That signature is what makes a boilerplate entry a problem: descriptions reading “attorney-client communication” and nothing else are routinely rejected, because they give the requesting party no basis on which to assess the claim. Whether a person or a model drafted the text is not the question the court asks — whether the description is grounded in the actual document is.
FRE 502(d) orders. A court-entered 502(d) order provides that disclosure does not waive privilege, regardless of the precautions taken, and it is widely described as the most powerful and least-used cost-control tool in discovery. Get one and the review can be calibrated to proportionality rather than to zero-defect fear. Without one you fall back on the five-factor reasonable-steps test in 502(b) — which is exactly where the producing party in Mt. Hawley Insurance Co. v. Felman Production, Inc., 271 F.R.D. 125, 133–36 (S.D. W. Va. 2010), ended up when its ESI stipulation with clawback language was held not to be a 502(d) order, and lost privilege over the disputed material as a result.
Categorical logging under Thrasher. SEC v. Thrasher, 1996 WL 125661, at *1 (S.D.N.Y. Mar. 20, 1996), permits summaries by category where document-by-document listing would be unduly burdensome and the extra detail would be of no material benefit to the requesting party. Both prongs, and the burden is on the producer — so the practical move is to agree the categorical approach in the ESI protocol before production rather than argue for it after a challenge. Privilege review at scale walks through the drafting.
Where AI helps rather than hurts: an AI-assisted workflow strengthens the Rule 26(g) and 502(b) reasonableness showing when descriptions are grounded in document content, required fields are validated before the log is finalised, and the audit trail distinguishes AI-drafted from attorney-edited entries. That last item is the one to insist on in a demo — it is what lets you show a magistrate what the process actually was. More on the standard in is AI-assisted review defensible.
What is a privilege log in eDiscovery, and when is one required?
A privilege log is the itemised list of documents a producing party is withholding or redacting on privilege grounds, with enough detail for the requesting party to assess each claim. Federal Rule of Civil Procedure 26(b)(5) requires it whenever responsive material is withheld: the party must describe the nature of the documents without revealing the privileged content itself. In practice the log is produced alongside the production set, and its adequacy is what gets litigated when privilege is challenged.
How much does it cost to create a privilege log manually versus with AI?
Manual logging runs $5 to $15 per entry for drafting alone, and $8 to $30 effective once you price the 15 to 30 minutes a contract reviewer at $35–$60/hour actually spends on each one. The privilege review that precedes logging adds another $4 to $8 per document. On a 2,000-document privilege set, that stacks to roughly $32,000. AI-drafted entries with attorney QC cut the per-entry attorney time to 3 to 15 minutes, and under a flat per-gigabyte platform fee the marginal cost of the log itself is zero.
What is the difference between per-GB, per-seat, and per-document eDiscovery pricing?
Per-GB bills the volume of data hosted, so cost tracks collection size and is knowable at intake. Per-seat bills each user with access, typically $50 to $500+ a month, so cost tracks how many people you let into the matter. Per-document and per-entry bill each unit of review work, so cost tracks reviewer labour and is the model that makes privilege logs expensive. Most vendors combine at least two of the three, which is why a quoted headline rate rarely predicts the invoice.
Can an AI-generated privilege log survive a court challenge?
Yes, when the workflow is built for it. What a court examines is whether the descriptions are grounded in the actual document rather than templated, whether required fields are complete and consistent, and whether an attorney of record reviewed the entries before certifying under Rule 26(g). An audit trail distinguishing AI-drafted text from attorney-edited text strengthens that showing. Boilerplate entries reading “attorney-client communication” get rejected regardless of whether a human or a model wrote them.
What is an FRE 502(d) order, and how does it reduce privilege review cost and risk?
Federal Rule of Evidence 502(d) lets a court order that disclosure of privileged material in the case does not waive privilege, in that proceeding or any other. With one entered, an inadvertent production is clawed back without a waiver fight, and the review can be calibrated to proportionality rather than to zero-defect fear. Without one, the producing party falls back on the 502(b) reasonable-steps test — the position the producer in Mt. Hawley Insurance Co. v. Felman Production, Inc. ended up in when its clawback stipulation was held not to be a 502(d) order.
How is DecoverAI’s pricing different from Relativity, Everlaw, Logikcull, and Reveal?
DecoverAI charges a flat $60/GB/month that covers classification, privilege review, log generation, redaction, Bates numbering, and production, with unlimited users and no contract minimum. Relativity layers per-user and per-document charges on a per-GB base; Everlaw is primarily per-seat plus hosting; Logikcull combines per-GB with per-user tiers; Reveal is enterprise-contract-only with no public rate card. The structural difference is that the line items that make privilege logs expensive — per-entry drafting, per-document review, per-seat access — do not exist in a flat data-based rate.
What is a categorical privilege log, and when can I use one instead of entry-by-entry logging?
A categorical log groups withheld documents into described categories rather than listing each one. SEC v. Thrasher, 1996 WL 125661 (S.D.N.Y. Mar. 20, 1996), set the two-prong test still applied today: a document-by-document listing must be unduly burdensome, and the added detail must be of no material benefit to the requesting party in assessing the claim. Both prongs have to be met and the producing party bears the burden, so the practical move is to agree the categorical approach in the ESI protocol before production rather than argue for it afterwards.
This article is general information about legal technology and discovery practice, not legal advice for any particular matter, and citing a case here is not a representation about how any court would rule on your facts. Third-party pricing reflects publicly available vendor and review-site information at the date of writing and may change; several vendors named here publish no rate card at all, and every figure — including ours — should be confirmed directly with the vendor for a specific matter. Fee ranges are typical market figures rather than quotes.