Harvey vs Lexis+ AI (Protégé) in 2026: Research vs Platform

The Short Answer

Lexis+ AI no longer exists under that name: on February 24, 2026 LexisNexis replaced it with Lexis+ with Protégé, a research-database-native AI with Shepard's citation validation built into every answer. Harvey is a standalone legal AI platform that licenses LexisNexis content and Protégé answers through an alliance, and wraps them in drafting, document review, and Microsoft 365 workflows. If research is the center of your practice, Lexis+ with Protégé is the more direct buy and the one a small firm can actually purchase. If you want one platform for research, drafting, and review across a mid-size or large firm, Harvey is the broader product. Whether you need both is the question most firms are really asking, and we answer it below.

Verdict by Firm Type

• Solo and small firm (1–10 lawyers): Lexis+ with Protégé, if you need a research database at all. LexisNexis sells single-user plans (a base non-AI Lexis+ seat is listed around $171 per user per month on store.lexisnexis.com, with the AI layer quoted separately). Harvey's reported 25–50 seat minimums put it out of reach. Pair the research database with Claude for drafting and analysis (see Claude pricing for lawyers) rather than paying for a second AI platform.
• Mid-size firm (10–150 lawyers): Depends on the mix. Litigation-heavy firms often get more from Lexis+ with Protégé plus a general model. Firms with heavy transactional and drafting work get more from Harvey with the Lexis bundle, and may keep a smaller number of full Lexis seats for research specialists.
• Large firm: Usually both, with the overlap managed deliberately: Harvey firm-wide, full Lexis+ with Protégé seats for research-intensive lawyers, librarians, and litigation groups.

What Changed at LexisNexis in 2026

Three dated releases define the current product. In February 2026, Lexis+ with Protégé launched with more than 300 pre-built workflows, a no-code custom workflow builder, models from Anthropic, Google, and OpenAI, and a knowledge graph spanning more than 200 billion documents. In May 2026, LexisNexis added agentic skills, collaboration workrooms, and customer-held encryption keys. On August 24, 2026, it unveiled the "Legal Intelligence Engine": an orchestration layer that picks models, agents, and sources per task, returns review-ready output to Word, Excel, or PowerPoint, and extends Shepard's into drafting as "Shepard's Verify," which checks that cited cases are accurate, still good law, and relevant to the specific assertion.

The strategic posture is the opposite of Harvey's. When Anthropic launched its legal plugin on February 3, 2026, Legal IT Insider described LexisNexis as absorbing Anthropic's models into Protégé to reinforce its data advantage, while Harvey integrated with Anthropic's MCP program so Claude users could call Harvey workflows from inside Claude.

What Harvey Brings to Research

Harvey is a platform first: Assistant, Vault, Workflows, Agents, Contract Intelligence, and a Copilot agent (see our Harvey AI review for the full tour). Its research answer comes from the LexisNexis alliance announced June 18, 2025: customers can select the Protégé service inside Harvey and receive answers grounded in LexisNexis U.S. case law and statutes, validated through Shepard's, plus jointly developed motion-to-dismiss and summary-judgment workflows. Harvey's scale is also relevant to research quality, because workflows improve with usage: as of September 9, 2026 it reports 3,000+ organizations, 80 percent of the Am Law 100, and ARR above $400 million.

The practical difference: in Lexis+ with Protégé, the research database is the product and the AI is a layer on it. In Harvey, the legal workflow is the product and Lexis research is one grounded source inside it. Harvey answers a broader range of questions; Lexis answers research questions with the deeper content stack (secondary sources, Lex Machina analytics, Law360, Intelligize) behind it.

Comparison Table

Citation Grounding: The Real Difference

Both products aim at the same failure mode, a confident answer citing a case that does not say what the AI claims. Lexis+ with Protégé addresses it structurally: every answer draws from the LexisNexis corpus and is checked against Shepard's, and the August 2026 release pushes that check into drafting, so a brief's citations are validated as you write. Harvey addresses it by licensing that same LexisNexis grounding for research answers, while its non-research features (Vault review, drafting from your documents) are grounded in the documents you supply. The gap is not "grounded versus ungrounded." It is that Lexis grounds a wider range of research work in more content, while Harvey grounds a wider range of work types in your matter files.

For a small firm using Claude directly, the honest statement is different: Claude is excellent at analyzing authority you give it and at drafting, but it is not a citation-validated database. That is why our research and cite-check workflow ends with verification in Lexis, Westlaw, or a free citator, and why we treat every AI research answer as a lead rather than a conclusion.

Pricing: What Is Actually Public

Neither company publishes an AI price list, so the best you can do is triangulate from listed base products and buyer reporting, as of September 2026. For Lexis, a single-user non-AI Lexis+ plan is listed around $171 per user per month, and buyer reporting puts all-in Protégé seats at roughly $250–500 per seat per month depending on firm size and content packages, on annual auto-renewing contracts with no free trial. For Harvey, industry reporting compiles roughly $1,000–2,000 per user per month mid-market, roughly $100–200 at Am Law 100 scale, 25–50 seat minimums, and 12-month terms; a separate analysis reports a Lexis-bundled Harvey seat near $2,400 per month. Full detail in Harvey AI pricing. Ask both vendors, in writing, what the Lexis bundle inside Harvey includes versus a full Lexis+ with Protégé seat; that answer decides whether you are buying overlap.