AI e-discovery cost per GB compared: Relativity (aiR), Everlaw, Disco, Logikcull (Reveal), Reveal, and Casepoint — what litigation teams actually pay in 2026

**Relativity** is still the gravitational center of e-discovery. Its desktop product RelativityOne is the cloud-native version every Am Law 200 firm has standardized on for complex matters, and the aiR add-on launched in late 2023 — aiR for Review, aiR for Privilege, and aiR for Case Strategy — has matured into a genuine GenAI review tool rather than the predictive-coding wrapper it started as. Relativity does not sell direct to law firms below a certain size; you buy through a Relativity Certified Partner (a service-bureau like Lighthouse, Consilio, or Epiq), which means the headline per-GB number on https://www.relativity.com/pricing/ is rarely what you actually pay. The partner marks it up and bundles project management on top.

**Everlaw** is the cleanest browser-native challenger. Founded in 2011, it skipped the Windows-thick-client legacy and built a true web platform with collaborative review, predictive coding, and the Everlaw AI assistant included in every seat. It sells direct, prices per seat per year, and includes hosting and processing in the seat cost up to reasonable thresholds. The pricing page at https://www.everlaw.com/pricing/ does not list rates publicly — you submit a form — but verified RFP responses in 2026 land in the $3,000 to $12,000 per seat per year range depending on tier and volume commitment.

**Disco** (CS Disco, ticker LAW) pioneered the bundled per-matter model: you do not pay separately for hosting, processing, and review software. You pay a blended per-GB rate per matter, with Cecilia AI features included. That is excellent for unpredictable matter loads — a litigation boutique that takes on a sudden 800 GB matter does not need to pre-purchase anything — but it also means Disco does not have a $20/GB headline rate; the all-in rate per matter typically runs $10 to $30 per processed GB depending on review volume, with hosting calculated separately on a monthly basis after the first 90 days. https://www.csdisco.com/pricing has the model laid out.

**Logikcull**, acquired by **Reveal** in early 2023, kept its self-service identity. It is the only major platform where you can swipe a credit card and start uploading a 25 GB custodian collection at 11pm on a Sunday. Flat per-GB pricing, no seat fees on the entry tier, and a UX designed for in-house counsel and solo litigators rather than dedicated litigation-support teams. The published rate of $20 to $50 per GB at https://www.logikcull.com/pricing depends on volume tier and contract length.

**Reveal** the parent platform is a different animal: enterprise-only, AI-first (Reveal AI was originally NexLP, one of the strongest content-analytics engines in the market), and sold as a platform deal to corporate legal departments and large firms. Reveal does not publish per-GB pricing. RFP responses in 2026 typically start around $50,000 per year for a small enterprise deployment and scale into seven figures. **Casepoint** is the closest analog to Reveal on the enterprise side, with the distinguishing feature that it is FedRAMP authorized and has been the platform of choice for the DOJ, FTC, and several civil agencies. Casepoint is also enterprise-only with $50,000+ annual floors per https://www.casepoint.com/.

Every e-discovery platform charges for three distinct things, and the trap is that vendors collapse them differently. **Relativity** keeps them maximally itemized: hosting is a monthly per-GB fee on whatever data is currently sitting in the workspace, processing is a one-time per-GB fee on data as it is ingested, and aiR review is a third per-GB add-on on documents you actually push through the AI workflow. A 500 GB matter that lasts six months on Relativity with full aiR review is roughly $75,000 in processing ($150/GB midpoint), plus $52,500 in hosting ($17.50/GB/mo midpoint times six months times 500 GB), plus $16,250 in aiR ($32.50/GB midpoint on the half of the corpus that gets AI review). That is $143,750 before partner markup, project management, or expert services.

**Everlaw** collapses all three into the per-seat annual cost up to a per-seat data quota, which is typically 25 to 100 GB of hosted data depending on tier. A ten-person review team at $7,500 per seat per year is $75,000 with up to 1 TB of included hosting — for the same 500 GB matter, that is roughly half the Relativity bill, but only if your review headcount actually fits a ten-seat license. If you need 40 reviewers for two months, the per-seat model becomes punitive unless Everlaw offers a flex license, which they do but rarely publicize. https://www.everlaw.com/pricing/ confirms the seat-based foundation.

**Disco** is the cleanest math: a blended per-GB matter rate, typically in the $10 to $30 range per processed GB, with hosting prorated after the first 90 days at a few dollars per GB per month. A 500 GB matter at $20 per processed GB is $10,000 in processing, plus three months of hosting at $5/GB/mo on 500 GB is $7,500, total $17,500 — which sounds suspiciously cheap until you read the fine print about review-time billable hours, which Disco does charge separately if you use its bundled review services. The platform itself is genuinely cheaper than Relativity; the total project cost is closer once you add labor.

**Logikcull** at the small end is the cheapest credible option. At $30/GB flat for the mid-tier subscription, a 25 GB custodian collection runs $750 in upload-and-process fees with hosting included for the duration of the active subscription month. That is the price point where solo litigators and small in-house teams operate, and it is below every other platform's economic floor. The catch: Logikcull's AI features, while genuinely useful, are not at parity with aiR or Everlaw AI for complex privilege review, so it works best for investigations and pre-litigation matters rather than bet-the-company cases.

**Reveal** and **Casepoint** do not negotiate on a per-GB basis at all. They sell platform deals: you commit to a multi-year contract with a fixed annual fee, and within that envelope you get unlimited matters up to a total-data ceiling (typically 5 TB to 25 TB depending on tier). A mid-size corporate legal department with $200,000 to $400,000 in annual e-discovery spend across multiple matters often comes out ahead on a Reveal or Casepoint deal versus paying per-matter on Relativity or Disco. Casepoint additionally bundles FedRAMP-required controls, which is non-trivially valuable if you have any government work.

The AI add-ons are where 2026 pricing diverges most sharply from 2023 pricing. **Relativity aiR** is the only major platform that charges separately for AI review. The aiR for Review module runs $15 to $50 per GB on top of standard processing and hosting per Relativity's published guidance at https://www.relativity.com/pricing/, with the lower end of the range for partner-volume deals and the upper end for one-off large matters. aiR for Privilege and aiR for Case Strategy are separate SKUs again. On a 500 GB matter where you push 250 GB through aiR, that is $3,750 to $12,500 added to the bill — meaningful, but the time savings against linear review at $50/hour contract attorney rates pay for it on essentially any matter over 100 GB.

**Everlaw AI** is included in every seat license at no incremental cost — that is the entire pricing strategy. The bet is that AI features become table stakes within 18 months and Everlaw would rather lock in seat revenue than try to charge for what will soon be commoditized. In practice, Everlaw AI's assisted-review and summarization features are strong but slightly behind aiR on privilege classification, which still favors Relativity for big-document privilege reviews. For mid-market matters, the included AI is a major budget advantage.

**Disco's Cecilia AI** is similarly included in the per-matter blended rate. Cecilia's strength is rapid issue tagging and pattern detection across custodian sets; its summarization and document-level analysis are competitive with Everlaw AI. Because Disco prices per matter, you do not face the same per-GB AI surcharge that Relativity imposes, and small-matter economics on Disco with Cecilia AI included are genuinely best-in-class for litigation teams without an existing Relativity contract.

**Reveal AI** (the former NexLP technology) is arguably the strongest content-analytics engine in the market for sentiment analysis and concept clustering at very large scale — it was originally built for fraud and corporate-investigation use cases where the relevant signal is buried in thousands of communications. It is included in the Reveal platform contract, which is one reason corporate legal departments running multiple investigations standardize on Reveal: per-matter, the AI costs are effectively zero marginal. **Logikcull's Reveal-derived AI** features are a lighter-weight subset focused on self-service workflows. **Casepoint's AI** is a recent addition focused on government investigations; it is included in the platform contract.

The cynical read on AI pricing in 2026: Relativity is the only vendor still trying to monetize AI separately, and that strategy works only because Relativity has the most defensible incumbency in the largest matters. Everyone else has accepted that AI is now a baseline feature and is competing on platform-level value. For matters under 1 TB without an existing Relativity contract, the per-GB AI surcharge on aiR is the single biggest reason to seriously evaluate Everlaw or Disco instead.

If you are an **Am Law 200 firm** with an existing Relativity certification, dedicated litigation-support staff, and a steady stream of bet-the-company matters over 1 TB, the answer in 2026 is still **Relativity** with aiR. The pricing premium is real but the ecosystem — third-party tools, project managers who know the platform, certified review teams, court-recognized defensibility — is unmatched. Your fully-loaded cost on a 2 TB matter will land north of $400,000, but it would have been that anyway, and aiR materially reduces the review-attorney hours.

If you are a **mid-market litigation boutique** doing matters in the 50 GB to 500 GB range with five to twenty reviewers, **Everlaw** is the strongest fit. The per-seat model aligns with your headcount, AI is included, browser-native means no IT overhead, and the total cost on a typical matter is 40 to 60 percent below Relativity. The exception is if you have a specific need for FedRAMP or aiR-class privilege review.

If you are a **plaintiffs' firm or a litigation team with unpredictable matter inflow**, **Disco** is built for you. The per-matter pay-as-you-go model means you do not pre-commit. When a $200 million case lands on Tuesday, you can spin up a matter on Disco that afternoon with Cecilia AI included, and you only pay for what you process. The trade-off is that Disco's enterprise features and integration depth are lighter than Relativity's, and very large matters above 2 TB sometimes get steered toward Relativity or Reveal.

If you are **in-house counsel at a mid-size company** running internal investigations, employment matters, and pre-litigation reviews — typically under 100 GB at a time — **Logikcull** is the most economical and the lowest IT burden. The self-service UX means a single legal-ops person can run discovery without a service bureau. For occasional matters under 25 GB, you can stay on the pay-as-you-go entry tier and never sign an annual contract.

If you are a **Fortune 500 legal department** with multi-million-dollar annual e-discovery spend across dozens of matters, **Reveal** or **Casepoint** become economically rational because the platform deal converts variable per-matter costs to a fixed annual line item. Casepoint specifically wins if any of those matters involve government investigations or regulated-industry data residency requirements, because of its FedRAMP authorization. Reveal wins if you want the strongest content-analytics AI across long-running corporate investigations.

All six platforms hold SOC 2 Type II certification, ISO 27001, and offer SSO/SAML and customer-managed encryption keys on enterprise tiers. The differentiation is at the edges. **Relativity** with RelativityOne offers the broadest data-residency footprint — US, EU, UK, Canada, Australia — and has a separately certified RelativityOne Government environment for federal work. On-prem Relativity Server, while declining in market share, is still available for organizations that cannot use cloud, which matters in some defense and intelligence contexts.

**Everlaw** holds SOC 2 Type II and ISO 27001 and operates Everlaw for Government as a separately authorized environment with FedRAMP Moderate authorization, which is sufficient for most civil agency work but not for the IL5/IL6 tier of defense workloads. Data residency is available in US, EU, Canada, and Australia. https://www.everlaw.com/pricing/ touches on the government tier; the security documentation is gated.

**Disco** holds SOC 2 Type II and operates in US, EU, and Australia. Disco does not have FedRAMP authorization as of June 2026, which is a meaningful exclusion if your matter portfolio touches federal agencies. **Logikcull's** security posture is SOC 2 Type II, US and EU residency, no FedRAMP. For self-service matters under standard commercial requirements, Logikcull is fine; for anything touching CUI or federal data, it is a non-starter.

**Reveal** holds SOC 2 Type II, ISO 27001, and HIPAA attestation, with residency in US, EU, and Canada. **Casepoint** is the security leader in this group: SOC 2 Type II, ISO 27001, HIPAA, and FedRAMP Moderate authorization, with a path to High in process per their compliance page. If your CISO's checklist includes FedRAMP, Casepoint and (in the cloud tier) Relativity are the only two viable options. Everlaw for Government is a strong third, but with a narrower module set than the commercial Everlaw product.

On defensibility — the question of whether court orders and opposing counsel will accept your AI-assisted review workflow — all six platforms are court-tested at this point in 2026. Relativity has the deepest case-law trail going back to predictive coding in Da Silva Moore (2012). Everlaw and Disco have several years of TAR-2.0 case law behind their workflows. Reveal AI's content-analytics approach has been more frequently challenged but ultimately accepted in major federal matters since 2020. The newest aiR generative-AI workflows from Relativity have had fewer than a dozen reported court decisions specifically on the GenAI workflow as of mid-2026, which is the only meaningful defensibility caveat in the market.

The published per-GB rates are not the full picture. **Storage after matter close** is the most-overlooked line item: every platform charges to keep data live after the active review wraps, and rates of $5 to $15 per GB per month for warm storage add up fast on a 500 GB matter that needs to stay accessible for an active appeal or follow-on regulatory matter. Relativity and Casepoint have the most flexible archive tiers; Disco's after-matter storage costs have surprised more than one litigation manager. Always negotiate post-active-review hosting at signing, not after.

**Egress fees** when you leave a platform are a separate trap. Logikcull is the cleanest here — they explicitly do not charge to export your data and have built brand reputation around that policy. Relativity partners vary widely; some charge $50/GB to export at matter close, which on a 1 TB matter is a $50,000 lock-in fee. Everlaw, Disco, and Reveal have moderate egress charges in the $5 to $15 per GB range. Always read the export-and-termination clauses before signing.

**Project management** is the labor cost wrapper around platform fees. A Relativity partner typically bills $150 to $250 per hour for project management on top of platform fees, and a complex 1 TB matter consumes 200 to 600 PM hours over its lifetime, which is $30,000 to $150,000 of additional labor. Everlaw and Disco are usable by litigation-support staff without a service bureau, which can eliminate that cost entirely if you have in-house capability. Logikcull is designed for self-service. Reveal and Casepoint enterprise deals typically include a customer success manager but bill professional services separately at $200 to $350 per hour.

**Contract attorney review** is the largest single line item on most matters, and it does not change by platform — but the platform's AI features change how many hours you need. A traditional linear review at 50 documents per hour at a $55 per hour contract rate costs $1,100 per 1,000 documents. aiR or Everlaw AI assisted review at 200 to 400 documents per hour effectively cuts that to $138 to $275 per 1,000 documents. On a 500,000-document matter, that is $431,000 to $481,000 in saved review costs — which dwarfs any platform pricing difference between Relativity and Everlaw.

**Expert witness fees** for TAR validation and AI-workflow defense are now standard on contested matters. Budget $25,000 to $100,000 per matter for a TAR consultant or AI-workflow expert if opposing counsel challenges the protocol. This cost is platform-independent but worth modeling because it shifts the rational platform choice toward tools with the deepest case-law trail (Relativity, Everlaw) and away from tools whose AI workflows have less reported precedent.

Only **Relativity** still offers a true self-hosted option in 2026 — Relativity Server, the on-prem version of the platform. Market share for Relativity Server has been declining year over year as RelativityOne (cloud) becomes the default, but for organizations with regulatory or contractual requirements that prohibit cloud hosting of certain data — typically defense, intelligence, and some financial-services use cases — Relativity Server remains the only credible option in the market. On-prem deployments add infrastructure, licensing, and certified-admin costs that typically inflate total cost of ownership by 2x to 4x versus RelativityOne for the same data volume.

**Everlaw**, **Disco**, **Logikcull**, **Reveal**, and **Casepoint** are SaaS-only. None of them offer an on-premise deployment, and that is not changing. The strategic implication is that any organization with a hard no-cloud requirement is effectively locked into Relativity Server. Every other organization should evaluate the cloud platforms on their merits, because the operational overhead of running on-prem e-discovery infrastructure in 2026 is meaningful and rarely cost-justified.

**Data residency** matters separately from self-hosting. Even on cloud platforms, you can typically specify the region your data lives in: US, EU (Frankfurt or Dublin most commonly), UK, Canada, and Australia are the standard options. Relativity has the broadest residency footprint. Everlaw covers US, EU, Canada, Australia. Disco covers US, EU, Australia. Logikcull covers US and EU. Reveal covers US, EU, Canada. Casepoint focuses on US with FedRAMP boundaries. If your matter involves EU custodians and GDPR considerations, EU residency is non-negotiable and narrows the practical short list.

**Cross-border data transfers** are a 2026 minefield. The EU-US Data Privacy Framework remains in effect but is under active challenge, and several US platforms have responded by adding contractual standard contractual clauses (SCCs) and supplementary measures to their EU deployments. Relativity, Everlaw, and Disco all have current SCCs and DPF certification. Verify the specific version of SCCs in the customer agreement, because the 2021 version is the operative one for new contracts and older boilerplate has appeared in some platform agreements.

**Customer-managed encryption keys (CMEK)** are available on enterprise tiers of all six platforms in 2026. CMEK adds a meaningful security posture upgrade — your platform vendor cannot decrypt your data without your keys — and is worth the typical $10,000 to $25,000 annual surcharge for any matter involving trade secrets, attorney work product, or sensitive M&A diligence materials. CMEK is included in Casepoint's FedRAMP environment and in Relativity Government.

The AI features in 2026 e-discovery platforms break into four functional categories: (1) assisted document review and coding, (2) privilege classification, (3) summarization and narrative generation, and (4) communications analytics for investigations. **Relativity aiR** is strongest in categories 2 and 3, with privilege classification accuracy improvements over the previous TAR generation in the 15 to 25 percent range on benchmark datasets per Relativity's published research. **Everlaw AI** is strongest in category 1, with the cleanest reviewer-experience integration and the lowest training overhead.

**Disco Cecilia** is strongest in category 1 with built-in issue tagging that does not require pre-training, and in category 3 with deposition-prep summarization. **Reveal AI** dominates category 4 — communications analytics, sentiment, and concept clustering across custodian sets. If your matter is a corporate investigation rather than litigation review, Reveal's advantage in this category is meaningful enough to change the platform decision.

Where AI consistently falls down across all six platforms in 2026: foreign-language privilege review, highly technical document sets (drug development, semiconductor IP, complex financial derivatives), and matters with significant audio or video content that requires transcription before AI review. None of the platforms have solved multimodal review well; you still need separate transcription tools and human review for AV. Foreign-language privilege classification accuracy degrades materially below English, and high-stakes matters in EU jurisdictions still require human-attorney privilege review regardless of platform.

**Hallucination risk** in summarization features remains the biggest defensibility concern. All six platforms have implemented citation-required summarization that links every claim to a specific document Bates range, which materially reduces hallucination risk versus a generic LLM. Relativity, Everlaw, and Disco have published their hallucination-rate benchmarks; Reveal and Casepoint have not. Best practice in 2026 is to require any AI-generated summary used in deposition prep or court filings to be independently verified by a reviewing attorney, regardless of vendor.

**Cost of getting AI wrong** is the line item nobody puts on an RFP but everyone pays. An AI-misclassified privileged document produced to opposing counsel is potentially a sanctions exposure, a privilege-waiver argument, and a malpractice claim. The platforms with the deepest court-tested workflows (Relativity, Everlaw) carry less marginal risk than newer or less precedent-rich workflows. This is the one area where paying the Relativity premium is genuinely defensible on risk-adjusted economics for matters with high privilege stakes.

Vendors will not give you their best price until you make them compete. The single highest-leverage RFP question is: "Provide your fully-loaded per-GB cost for a 500 GB matter over six months, including processing, hosting, AI review on 50 percent of documents, and post-active-review storage for an additional 12 months." This question forces every vendor onto comparable footing and reveals the hidden line items they would rather you not notice. Relativity partners typically come back with two or three different scenarios; Everlaw and Disco give you a single blended number; Reveal and Casepoint give you a platform-deal number.

Second-highest leverage question: "What is your egress fee if we terminate the contract or migrate at matter close?" Get this in writing in the master agreement, not in a side letter. The answer ranges from zero (Logikcull) to $50 per GB (some Relativity partners), and on a 1 TB matter that difference is $50,000.

Third question: "Provide three references from active customers running matters of comparable size and complexity to ours." Talk to them about platform stability during major review pushes, support responsiveness, and total cost relative to original quote. Vendors will only give you references who will say good things, but specific questions about overruns and support escalation will still surface real signal.

Negotiate a price hold for the contract term. SaaS pricing in 2026 is moving up — Relativity raised RelativityOne base hosting roughly 8 percent in early 2026 per partner channel communications — and a multi-year contract without a price cap is a budget time bomb. Standard ask is 5 percent annual cap; vendors will often agree to a CPI-linked cap if you push.

Finally, do not buy the AI features without doing a paid pilot on representative data. Every vendor will say their AI delivers 60 to 80 percent review-time reduction. Run a 10 GB pilot on a closed matter with known coding decisions, measure precision and recall against the human-reviewed gold standard, and use the actual numbers in your TCO model. Anything less, and you are buying marketing claims rather than verified performance.