Law Firm / Legal AI GEO¶

When your client asks AI "which Chinese law firms specialize in IP litigation", is your name on the list?

How legal customers use AI (4 journeys)¶

1. Enterprise legal counsel pre-selecting outside counsel¶

Actual prompts: "Shanghai law firms with cross-border e-commerce compliance experience", "semiconductor patent litigation firms in China top 3", "listed-company stock-option review firms". AI returns: 3-5 firm names · specialties · contacts. Without GEO: not on the list.

2. Client using AI to draft contracts / terms¶

Client: "draft a cross-border agency agreement, referencing best practices recommended by Chinese firms". AI cites: public clause templates / case articles / lawyer commentary from firms. Value: client sees your name = free trust.

3. General Counsel picking co-counsel¶

Enterprise only uses AI to shortlist. 3-5 firms move to final round. No AI-citation memory = eliminated.

4. Lawyer using AI to find experts / precedents / statutes¶

Lawyers need cross-jurisdiction precedents · rare-clause judgments · latest statute interpretation. AI is an accelerator, and lawyers remember firms that give good answers.

Legal AI GEO key actions¶

1. Public case summaries — anonymized, preserve legal reasoning (Harvard Law Review style)
2. Clause-level commentary — actual civil code / company law / securities law clauses with practical takes
3. Vertical practice areas — healthcare / semiconductor / cross-border / IP / bankruptcy, each with a lead partner
4. Judgment analysis — worthwhile judgments, your firm's take, published
5. Compliance alerts — new statute → 24-48h client alert
6. Individual lawyer brand — partner profile + LinkedIn + Zhihu + WeChat unified identity

Legal AI delivery cases¶

Case 04 · Law Firm Case + Contract + Statute Tri-Corpus RAG¶

• Business problem: clause search P50 45 minutes, version chaos, citation errors
• Tech: bge-large + BGE-reranker · tri-corpus tiered · clause-level provenance · version-aware routing
• Boundary: client contracts on separate tenant · read-only auth chain
• KPI: P50 45min → 3min · citation accuracy >95% · lawyer active-usage >70%
• Full detail: Case 04

Case 18 · Group Contract Intelligent Review¶

• Business problem: conglomerate 15 subsidiaries · 200+ contract types · 1000+/month
• Tech: 4-layer review · clause-level risk · version alignment · group-unified template library
• KPI: review time -55% · missed key clauses -80%
• Full detail: Case 18

Case 08 · 4-Layer Government-SOE Contract Review Agent¶

• Business problem: SOE 20 subsidiaries, 600 contracts/month, missed clauses
• Tech: 4-layer (baseline / compliance / commercial / risk) · human final review
• KPI: legal review workload -60% · mean review time 3d → 4h · missed clauses -80%
• Full detail: Case 08

What legal projects won't do¶

• ❌ Put sensitive client data on "cloud GPT contract review" (data leaving firm = breach of privilege)
• ❌ Let AI directly generate final drafts (attorney must be final decider; agent outputs suggestions only)
• ❌ Answers without provenance (attorney citing = firm liability)
• ❌ Use statute libraries without version awareness (using stale law = malpractice)

Cost range¶

300k - 2M CNY typical.

Related¶

• Direction · RAG Knowledge Base
• Direction · Agent Workflow
• How-To · RAG Cold Start
• 中文详细版

Next: Portfolio PDF · 30-min alignment