We're building the tools lawyers need to practice with confidence.
Lawyers can't trust AI with their citations. Large language models hallucinate — they invent case names, fabricate volume numbers, and generate citations that look correct but don't exist. For a profession where citing bad law can mean sanctions, malpractice, or losing a case, that's not a minor inconvenience. It's a dealbreaker.
Citations.law takes a fundamentally different approach. Our flagship service, CiteGuard, doesn't use AI to verify your citations. It uses deterministic search — the same kind of mechanical lookup you'd use to find a word in a document.
We built a database of over 21 million indexed citation references from published court decisions across the United States. When you submit a document, CiteGuard detects supported case-reporter citations and checks them against the current corpus. Results use transparent statuses rather than probabilistic confidence scores.
CiteGuard checks supported case-reporter citations against a disclosed, current corpus. A match does not establish positive treatment or complete legal research; absence is not proof of invalidity.
A natural question: if AI hallucinates, how do we know the database itself is accurate?
The answer: AI didn't build the database. AI built the tools that built the database. The actual corpus was constructed through deterministic processes at every level — downloading published case law from authoritative public sources, parsing structured data, and indexing citations through mechanical extraction. No AI was involved in deciding what's in the corpus.
Citations.law draws from the most comprehensive public sources of American case law:
Together, these sources contain 16.8 million source case records and 21 million+ citation references across federal and state jurisdictions.
We designed Citations.law to minimize document data and support lawyers' confidentiality obligations. Documents are processed on Citations.law servers for citation extraction and verification. OCR, when needed, is sent to the service operator's configured Ollama endpoint. The standard deployment is intended to use service-controlled infrastructure, but configuration must be verified before relying on that deployment assumption. Raw uploaded documents are not retained after processing unless explicit raw-document retention applies in a separate documented workflow. Verification report HTML and audit metadata may be retained in the file cabinet when service-level report saving is enabled. Account or matter settings do not currently control /verify report saving.
OCR processing follows the configured Ollama endpoint described above. The standard deployment is intended to keep that endpoint on service-controlled infrastructure, but operators and lawyers must verify the actual deployment configuration.
CiteGuard is the first service on the Citations.law platform. We're building toward a comprehensive legal research and practice tool that provides the value lawyers need — citation verification, case research, authority ranking, and practice-specific workflows — without the cost and complexity of legacy legal research platforms.
Questions, feedback, or interested in early access? Reach us at info@citations.law.