The Legal Corpus
The 4CITE legal corpus draws from PACER filings, CourtListener, federal appellate records, and judicial opinions. Every document in the corpus is scored across four independent gates and stored with a longitudinal timestamp for drift analysis.
What 4CITE Analyzes in Law
4CITE for Law is built for the documents where structural failure has the highest stakes: filings submitted to courts, opinions that set precedent, and briefs that argue for or against someone's freedom, property, or rights.
Legal Briefs & Motions
Federal and state court briefs, motions to dismiss, summary judgment filings. Scores whether the argument structure is coherent with the cited authority — independent of whether the citations are real.
Judicial Opinions
Appellate and district court opinions scored for internal consistency. Surfaces reasoning gaps and contradictions between stated holdings and the underlying analysis.
Regulatory Filings
DOJ, FTC, SEC enforcement actions and responses. Detects accountability theater in institutional legal documents — the surface-coherent filing that contains no structural substance.
Contracts & Agreements
Material contracts where internal consistency of obligations and representations can be scored. Particularly effective for identifying drafting patterns common in AI-generated boilerplate.
The Mata v. Avianca Benchmark
The most extensively documented example of AI-generated legal failure is now part of the 4CITE validation corpus. The results illustrate what structural integrity analysis detects that citation checkers cannot.
Mata v. Avianca, Inc. (S.D.N.Y. 2023)
Attorneys Schwartz and LoDuca submitted a brief containing citations to nonexistent cases generated by ChatGPT. Judge P. Kevin Castel issued sanctions under Rule 11. The brief became the defining example of AI-generated legal accountability theater.
4CITE scored the fabricated brief at 7/100 — Tier 4, Fabricated — and scored authentic briefs from the same legal domain at 72–88/100. The 81-point gap between the fabricated brief and the authentic average is the largest single-document gap in the 4CITE validation corpus.
Critical finding: The fabricated brief's structural failure was detectable without knowing which citations were real. G4 (Paradox Resolution) scored the reasoning architecture itself — and the architecture was hollow. This is the gap citation checkers cannot close.
Full methodology and scoring breakdown: Evidence → WP7
Four Gates. Independent Evaluation.
Each document is scored across four gates in separate, isolated analysis passes. No gate sees another gate's output. This independence is what catches documents that score well on surface measures while failing at the foundation — the precise signature of AI-generated legal content and institutional accountability theater.
Paradox Resolution — Cannot be faked
Measures whether the foundational claims of the document actually support the conclusions. Detects the hollow architecture that underlies AI-generated briefs — arguments that read as coherent but contain no load-bearing reasoning. In the Mata brief, this gate detected near-zero structural depth.
Latent Intent
Probes the alignment between stated purpose and actual argumentative structure. Institutional documents designed to appear responsive without being substantive fail this gate at high rates — a characteristic pattern in accountability theater and AI-generated filings.
Argumentative Structure
Scores whether the document's internal logic is self-consistent under scrutiny. Authentic legal advocacy holds together under pressure because it comes from a single coherent reasoning source. AI-generated content develops internal contradictions because different sections were generated by different probability paths.
Rhetorical Architecture — Cannot be faked
Probes the rhetorical surface for performance vs. substance. In genuine legal writing, rhetoric is an expression of underlying reasoning depth. In AI-generated briefs, rhetoric is compensation for the absence of that depth. The gate detects the difference. Surface quality that scores high here while G4 is low is the classic Fabricated signature.
Integrity Tiers in Legal Context
Every 4CITE score maps to one of four Integrity Tiers. In the legal vertical, tier designation carries direct practical implications.
| Tier | Score Range | Designation | Legal Context |
|---|---|---|---|
| T1 | 67–100 | Integrated | Channel capacity matches or exceeds the claimed signal. Authentic advocacy with coherent foundation — surface and structure come from the same genuine source. Consistent with decades of precedent-setting briefs. Federalist No. 51 scores 91; genuine federal sanctions opinion: 88; Gettysburg Address: 89. |
| T2 | 50–66 | Functional | Channel capacity mostly sufficient. Genuine legal drafting with identifiable structural gaps — real substance, honest effort, some compression visible but non-critical. Typical range for standard litigation filings that have not been strategically constructed for reasoning integrity. |
| T3 | 30–49 | Incomplete | Channel capacity insufficient in identifiable dimensions — the gap is measurable and locatable. Surface holds in places, but specific structural areas cannot support the weight placed on them. Common in institutional filings designed to appear substantive. Warrants scrutiny before reliance. |
| T4 | 0–29 | Fabricated | Channel capacity catastrophically below what the document claims to transmit. Dressed-up surface, hollow underneath. Mata v. Avianca brief: 7. All AI-generated briefs in the validation corpus scored T4. In litigation, a T4 score on any filing you must rely on is material information. |
Who Uses 4CITE⁴law
Litigation Attorneys
Validate your own drafts before submission to reduce Rule 11 exposure. Score any filing you must rely on for structural integrity before you build on it.
Legal Research Teams
Bulk-score case law for structural integrity trends. Identify the corpus of opinions where the reasoning is strongest — and where reliance carries hidden risk.
In-House Counsel
Score outside counsel deliverables for accountability theater patterns. Know before you rely whether the analysis you received has structural substance or surfaces-only compliance.
Judicial Research Staff
Support chambers with structural integrity screening of submitted materials. The Mata problem is now a docket-management question; 4CITE gives chambers a structural read on what comes in.
Legal Tech Platforms
Integrate 4CITE scoring as a quality layer for AI-assisted legal drafting tools. Surface integrity score alongside AI-generated content at the moment of generation.
Law Schools & Clinics
Score student work for structural reasoning quality, independent of citation accuracy. Teach the difference between coherent argumentation and surface-level legal writing.
Legal Data Sources
The 4CITE legal corpus is built from public-record legal repositories. Document ingestion is ongoing; the corpus is scored as new filings become available.
PACER — Federal court electronic filing system. Source for district and appellate briefs, motions, and orders across all federal circuits.
CourtListener — Free Law Project's open-access database of federal and state opinions. Primary source for judicial opinion corpus.
Supreme Court opinions — Full corpus of published opinions from 1791 to present, scored for longitudinal integrity trend analysis.
Circuit court precedent — Published opinions from all 13 federal circuits, with emphasis on documents that are actively cited in current litigation.
Subscribe to 4CITE4law
4CITE for Law is in beta. Subscriptions are available now to law firms, legal research platforms, and in-house legal teams.
Choose your plan →Or buy a single 4CITE Cert ($15) — no account required