Choosing an AI eDiscovery platform is not a feature-checklist exercise. The platforms in this space differ meaningfully in how their AI models are trained, how they handle privilege review, and what happens to your client data after a matter closes. A tool that works well for a 50GB insurance subrogation matter may be the wrong call for a 10-million-document antitrust production.
This comparison focuses on four platforms that have documented AI-assisted review capabilities and are actively marketed to legal teams: Relativity aiR, Reveal, Logikcull, and Everlaw. The task context is litigation document review: initial culling, technology-assisted review (TAR), privilege logging, and production preparation.
Comparison Criteria and Methodology
Each dimension below was selected because it directly affects a review team's decision about which platform to deploy on a given matter. Criteria values are drawn from vendor documentation, published feature pages, and third-party legal technology assessments. Where a claim could not be independently verified from a primary source, it is noted as unverified.
- AI review model — the underlying approach (continuous active learning, LLM-assisted, or hybrid) and whether the model is matter-specific or pre-trained on a shared corpus
- Predictive coding / TAR support — whether the platform supports TAR 1.0 (simple passive learning), TAR 2.0 (continuous active learning), or generative AI-assisted prioritization
- Data isolation and privacy — whether client data is processed in a dedicated tenant, and the vendor's stated zero-retention or data-deletion policy
- Privilege detection — automated privilege log generation, attorney-client privilege flagging, and work-product identification capabilities
- Pricing model — per-GB, per-seat, or enterprise contract; whether pricing is publicly disclosed
- Target matter size — documented performance range and where the platform is positioned by the vendor (small matters, large-scale productions, or both)
- On-premises or hybrid deployment — relevant for regulated industries and matters with cross-border data transfer restrictions
Side-by-Side Matrix
| Dimension | Relativity aiR | Reveal | Logikcull | Everlaw |
|---|---|---|---|---|
| AI review model | Generative AI (GPT-4 class) layered on Relativity's existing review infrastructure; matter-specific prompting | Continuous active learning (CAL) + neural network classification; Brainspace analytics engine | Automated upload processing with AI-assisted culling; simpler ML model, not CAL-based | AI-assisted issue tagging and predictive prioritization; generative AI features added from 2024 onward |
| TAR support | TAR 2.0 via existing Relativity workflows; aiR adds LLM-assisted document summarization and issue spotting on top | TAR 2.0 (CAL); strong track record in large-volume matters; Brainspace concept clustering | TAR 1.0 / basic predictive coding; not positioned as a TAR-first platform | Predictive coding with active learning; generative AI-assisted coding suggestions in newer releases |
| Data isolation | Dedicated tenant per organization; Azure-hosted; data processing stays within customer's Relativity environment | Multi-tenant cloud with customer-specific data partitions; on-premises option available via Reveal Server | Cloud-only; multi-tenant architecture; data deletion on matter close per stated policy | Cloud-hosted (AWS); dedicated environments available at enterprise tier; SOC 2 Type II certified |
| Privilege detection | AI-assisted privilege flagging with attorney name matching; privilege log drafting via aiR; requires attorney review | Automated privilege detection; concept-based privilege identification; privilege log generation | Basic keyword and custodian-based privilege flagging; less sophisticated than CAL-based platforms | AI-assisted privilege tagging; privilege log generation; attorney-client privilege and work product identification |
| Pricing model | Enterprise contract; per-GB and per-seat components; pricing not publicly listed | Enterprise contract; per-GB processing + hosting fees; pricing not publicly listed | Per-GB ingestion + per-user/month; pricing tiers published on website | Per-GB + per-seat; enterprise pricing for large matters; some pricing visible on website |
| Target matter size | Large-scale litigation and investigations; optimized for high-volume matters (1M+ documents) | Mid-to-large matters; strong in complex litigation and government investigations | Small-to-mid matters; positioned for in-house teams, regional firms, and straightforward productions | Mid-to-large matters; growing enterprise segment; strong in complex multi-party litigation |
| On-premises / hybrid | Relativity Server (on-premises) available separately from RelativityOne cloud; aiR features are cloud-only | Reveal Server provides on-premises deployment option; full feature parity not guaranteed vs. cloud | Cloud-only; no on-premises option | Cloud-only; no on-premises option as of May 2026 |
Platform Profiles
Relativity aiR
Relativity aiR is the generative AI layer built into RelativityOne, the cloud-hosted version of the platform that has dominated large-firm eDiscovery for over a decade. The practical implication is that aiR is not a standalone product — it extends a review environment that most large litigation teams already operate in.
The aiR for Review module uses an LLM to generate document summaries, suggest responsiveness determinations, and flag privilege issues. Critically, the model operates within the customer's existing Relativity workspace, so the AI outputs are integrated into the same review workflow rather than appearing in a separate interface. Reviewers see an AI-suggested coding decision alongside the document, with an explanation they can accept or override.
The limitation worth flagging: aiR's generative features are cloud-only. Firms running Relativity Server on-premises for data sovereignty reasons cannot access aiR without migrating matters to RelativityOne. That is a real constraint for regulated industries and for matters subject to cross-border data transfer restrictions.
Reveal
Reveal's core differentiator is its Brainspace analytics engine, which has been in production use in large-volume matters for several years. The continuous active learning model is well-documented and has been validated in court-approved TAR workflows, which matters for matters where opposing counsel or a court may scrutinize the review methodology.
Reveal added generative AI features from 2023 onward, including AI-assisted document summarization and issue identification. The platform's on-premises option (Reveal Server) is a genuine differentiator for firms that cannot use cloud-hosted processing — though the on-premises version may lag the cloud release in feature availability.
For government investigations and regulatory matters where the review methodology will face scrutiny, Reveal's established CAL track record is an argument in its favor over newer generative AI-first approaches.
Logikcull
Logikcull occupies a different market position than Relativity or Reveal. It is priced transparently (per-GB ingestion plus per-user monthly fees, with tiers visible on the website), designed for self-service operation, and targets in-house legal teams and regional firms handling straightforward productions.
The AI capabilities are less sophisticated — basic predictive coding rather than continuous active learning — but for a 50,000-document employment matter or a standard commercial dispute, the simpler model is often sufficient and the lower cost and faster onboarding are real advantages.
Logikcull is not the right choice for matters requiring defensible TAR methodology documentation, multi-party complex litigation, or very large document volumes. The platform does not publish a CAL workflow, which limits its defensibility in contested review methodology disputes.
Everlaw
Everlaw has positioned itself as a collaboration-first eDiscovery platform, with strong features for multi-party litigation where multiple law firms need access to the same document set. The interface is generally considered more modern than Relativity's, and the onboarding curve is lower.
The generative AI additions from 2024 onward include AI-assisted issue tagging, document summarization, and coding suggestions. Everlaw is AWS-hosted with SOC 2 Type II certification and dedicated environments available at enterprise tier — a reasonable security posture for most commercial matters, though not equivalent to the on-premises option Reveal offers.
For in-house teams managing recurring litigation with external counsel, Everlaw's collaboration features and more approachable pricing transparency make it worth evaluating alongside Logikcull.
Decision Framework by Team Type
The right platform depends heavily on matter volume, team structure, and whether defensible TAR methodology is a requirement. Below is a practical mapping.
| Team / Scenario | Recommended Starting Point | Key Reason | Avoid |
|---|---|---|---|
| Large firm, complex litigation (1M+ docs) | Relativity aiR or Reveal | Established TAR infrastructure, enterprise data controls, court-accepted review methodologies | Logikcull (not built for this scale) |
| In-house team, recurring commercial disputes | Logikcull or Everlaw | Self-service operation, transparent pricing, faster matter setup | Relativity aiR (cost and complexity overhead) |
| Government investigation / regulatory matter | Reveal | Documented CAL methodology, on-premises option, track record in government matters | Logikcull (limited TAR defensibility) |
| Multi-party litigation, multiple outside counsel | Everlaw | Collaboration features, shared workspace model, modern interface | Reveal Server on-premises (collaboration features weaker) |
| Regulated industry, data sovereignty required | Reveal (on-premises) or Relativity Server | On-premises deployment; aiR features unavailable on Relativity Server | Logikcull, Everlaw (cloud-only) |
AI Accuracy and Hallucination Risk in eDiscovery
The hallucination risk profile in eDiscovery AI is different from legal research tools. A research platform hallucinating a case citation is a direct professional responsibility problem. An eDiscovery platform making a wrong responsiveness determination is a quality control problem that human review is designed to catch — but the risk is not zero.
Generative AI-assisted coding (as in aiR and Everlaw's newer features) introduces a different failure mode than traditional TAR: the model may generate a plausible-sounding summary or responsiveness rationale that is factually incorrect about the document's contents. Reviewers relying on AI summaries without reading the underlying document are exposed to this risk.
Traditional CAL-based platforms (Reveal, and Relativity's pre-aiR TAR workflow) have a more established body of case law supporting their use in court-supervised review. Generative AI-assisted review is newer and has less documented judicial acceptance, though this is changing as courts become more familiar with the technology.
Data Privacy and Confidentiality Considerations
Every platform in this comparison processes client documents — often highly sensitive materials subject to attorney-client privilege. The data privacy questions that matter most are not the same as the ones that matter for a legal research tool.
- Does the vendor use client documents to train or fine-tune its AI models? All four platforms state they do not use client matter data for model training, but review the current DPA before relying on this.
- Where is data processed geographically? For matters with EU data subjects, GDPR obligations apply to where document processing occurs. Relativity and Everlaw offer EU data residency options; confirm current availability for your specific matter.
- What happens to data after matter close? Logikcull's published policy includes deletion on matter close. Relativity and Reveal deletion timelines depend on the enterprise contract.
- Are AI model outputs (summaries, coding suggestions) retained separately from the documents? This is relevant for privilege log defensibility — some platforms retain AI outputs as part of the audit trail.
Pricing: What You Can Actually Compare
Logikcull is the only platform in this comparison with publicly listed pricing tiers. As of the last verified date, Logikcull's per-GB ingestion fee combined with per-user monthly fees makes cost modeling straightforward for matters under ~500GB.
Everlaw publishes some pricing information for smaller matters but moves to enterprise contracts at higher volumes. Relativity aiR and Reveal are both enterprise-contract-only, with no publicly listed rates. For procurement purposes, this means budget estimates for these platforms require vendor engagement.
A common cost comparison mistake: comparing per-GB ingestion fees across platforms without accounting for hosting fees, processing fees, and user seat costs. The total cost of a matter on Logikcull at $X/GB may be lower or higher than Everlaw depending on the number of reviewers and the matter duration.
What This Comparison Does Not Cover
This matrix also does not evaluate customer support quality, implementation timelines, or integration with specific case management systems — all of which affect the real-world experience of deploying these platforms on live matters. Those dimensions require direct reference checks with firms of comparable size and matter type.
Comments
Join the discussion with an anonymous comment.