A litigation fact graph is a structured map of the people, organizations, events, documents, claims, and relationships in a legal matter. Instead of treating each file as an isolated item, the graph connects a factual statement to its source, date, speaker, subject, and related evidence.
AI can help build and query this map across large document collections. Lawyers still determine relevance, credibility, privilege, legal significance, and what may properly be presented to a court.
How a Fact Graph Differs From a Timeline
A chronology places events in date order. A fact graph can include time, but it also records other relationships.
For example, an email may be authored by one employee, sent to another, discuss a contract obligation, contradict later testimony, and attach the document that triggered the dispute. A timeline may show when the email was sent. A fact graph can connect all of those elements.
The two tools work well together. The graph organizes relationships, while the timeline presents a selected sequence for analysis or advocacy.
What a Litigation Fact Graph Contains
The exact model depends on the matter. Common elements include:
• People, companies, departments, regulators, and third parties.
• Events, meetings, communications, payments, decisions, and deadlines.
• Documents, exhibits, transcripts, images, messages, and recordings.
• Allegations, defenses, admissions, denials, and disputed propositions.
• Contracts, policies, duties, and relevant legal standards.
• Source locations, custodians, dates, and confidentiality status.
• Relationships such as authored, received, approved, contradicted, amended, or supported.
Each important factual proposition should remain connected to the material that supports or challenges it.
How AI Helps Build the Graph
In a document-heavy matter, AI can assist with entity recognition, date extraction, event classification, document linking, name normalization, topic clustering, and identification of repeated or inconsistent statements.
The system may suggest that "A. Smith," "Alex Smith," and an email address refer to the same person. It may connect a meeting note to a later invoice or identify several versions of a representation made at different times.
These are proposed connections, not final findings. The legal team should validate material links against the original sources.
Why Source Links Are Essential
A graph without source references can become a persuasive-looking collection of unsupported assertions. Every important node or relationship should point back to the relevant document, page, paragraph, timestamp, or testimony.
The reviewer should be able to move from a factual statement to the source in one step. If the source does not support the statement, the graph should be corrected rather than rationalized.
This principle aligns with the broader requirement to authenticate and establish a proper foundation for evidence. The federal judiciary's Evidence Rules Committee continues to study AI-generated and machine-generated evidence, including reliability and authenticity issues. See the May 2026 committee report.
Use the Graph to Test Competing Case Theories
A useful graph should not contain only facts favorable to one side. It should allow the team to compare competing explanations.
For each material proposition, ask:
• Which sources support it?
• Which sources contradict or qualify it?
• Is the statement based on firsthand knowledge?
• Did the witness make a different statement earlier?
• Is a key date inferred rather than documented?
• What evidence would be expected but is missing?
This helps lawyers distinguish a strong theory from a story that survives only because contrary evidence was overlooked.
Identify Contradictions Without Overstating Them
AI can compare testimony, interviews, contracts, emails, and other records to surface inconsistent language. Not every difference is a legal contradiction.
A witness may use different words to describe the same event. Two statements may concern different time periods. A later document may reflect new information. Translation, context, and document versions can also explain an apparent inconsistency.
The graph should label the connection as a potential conflict until a lawyer reviews the source and context. This prevents an automated suggestion from becoming an unsupported accusation.
Map Evidence Gaps and Unanswered Questions
Fact graphs are valuable for what they do not contain. A missing approval record, unexplained date, absent attachment, or unsupported assumption can direct the next investigation step.
The legal team may use gaps to prepare discovery requests, deposition outlines, witness interviews, subpoenas, or document-preservation work. Each gap should be framed as a question rather than treated as proof of misconduct.
Protect Privilege and Sensitive Material
A litigation graph may expose relationships that are not obvious from individual documents. Access controls therefore need to apply to the graph, its search results, and derived summaries.
The system should preserve matter restrictions, ethical walls, protective-order designations, personal data controls, and privilege decisions. A user should not receive a prohibited fact merely because it was extracted into a shared graph.
Legal teams should also decide whether attorney notes, work product, and case theories belong in the same graph as discoverable factual material.
Keep Human Decisions Separate From Extracted Facts
The graph should distinguish among source text, extracted information, reviewer-approved facts, allegations, legal conclusions, and strategic assessments.
For example, "the contract was signed on March 1" may be extracted from a signature page. "The contract became enforceable on March 1" is a legal conclusion that may depend on other facts and governing law.
Clear labels reduce the risk that an AI inference is later repeated as an established fact.
Create a Reviewable Update Process
Litigation records change. New production, corrected testimony, amended pleadings, and court rulings can alter the graph.
Record when an item was added, which source supports it, who reviewed it, and whether later evidence changed its status. Preserve prior versions when they matter to the case history.
An update process is especially important when multiple lawyers, experts, investigators, and vendors contribute to the matter.
How Alloomi Supports Connected Review
Alloomi AI can help legal teams analyze multiple documents, identify relevant language, and keep findings connected to source material for human review. Those capabilities can support chronologies, contradiction checks, witness preparation, and issue maps.
See AI Document Review in Litigation for the broader document-review workflow. A fact graph is not a substitute for that review. It is a structured layer that helps the team see relationships across the collection.
Conclusion
A litigation fact graph turns a document collection into a connected map of evidence, events, people, and disputed propositions. Its value comes from traceability, not visual complexity.
AI can accelerate extraction and suggest connections, but lawyers should validate material facts, preserve privilege, distinguish inference from evidence, and keep every important conclusion linked to the underlying record.
Frequently Asked Questions
1. Is a litigation fact graph the same as a case chronology?
No. A chronology primarily organizes events by time. A fact graph also connects people, documents, claims, sources, and relationships. It can generate or support a chronology.
2. Can a fact graph prove that a witness lied?
No. It can surface potentially inconsistent statements, but lawyers must examine context, wording, timing, knowledge, and source reliability before drawing conclusions.
3. Is information in a fact graph discoverable?
That depends on what the graph contains, how it was created, the applicable law, privilege, work-product protection, court orders, and the circumstances of the matter. Legal teams should address those issues before deployment.
