Powered by Google Gemini, constrained by our curated legal knowledge base. From witness statements to dispute responses, every output is verified, every citation is cross-referenced.
Personal identifiers are masked on our server before anything reaches the AI. Your documents have their text extracted in your browser, and only that extracted text is sent to us. The AI does not train on your data. Encrypted at rest and in transit.
210 UK authorities across five practice areas including Hussain, Lagden, Stevens, Dimond, and more. Each case includes citations, key principles, and BAILII links.
The Witness Statement Generator runs 15 automated format checks before export — statement of truth wording, formatting, required declarations, preparation-method disclosure (Wiltshire v Aioi guard), source attribution for every fact, and the CJC-aligned AI-use declaration block above the statement of truth. The checks cover format and structure; substantive compliance rests on your fee-earner’s verification with the witness.
Custom AI prompts developed by experienced claimant practitioners ensure responses are claimant-focused, well-structured, and practically useful. Separate instruction sets sit behind the witness statement, HDR, clinical negligence, EL/PL, BHR and TPI tools.
Visual indicators show when responses are cross-referenced against our case law database. Limitations are clearly stated when authority cannot be confirmed.
Integrated Highway Code provisions, Road Traffic Act 1988, and Consumer Contracts Regulations 2013 for comprehensive legal reference.
Designed and maintained by people who work in claimant litigation every day, across RTA, EL/PL, housing disrepair and clinical negligence, not generic AI developers.
Every AI output passes through three independent safety layers before it reaches you. Whether it's a witness statement, a TPI response, or a liability assessment, the same verification process applies.
1. Prompt Rules
The AI is strictly prohibited from citing any case not in our knowledge base.
2. Citation Validator
Every case name is cross-referenced against our verified database. Unverified citations are stripped.
3. Text-Body Scrubber
A regex-based scanner catches any remaining "X v Y" patterns in prose and removes them if unverified.
The same verified knowledge base drives every feature - here's how the AI works across the platform.
The system extracts facts from your uploaded case documents (call notes, accident report, engineer report, payment pack), assembles a first-person draft from the witness’s own account, suggests relevant case law where appropriate, and runs 15 CPR Part 32 format checks — including the CJC-aligned AI-use declaration — before export for fee-earner verification.
The AI identifies every argument in the insurer's letter, checks every citation they've used, and flags any unsupported claims.
The AI drafts a structured, case-law-backed response to each argument identified, citing only authorities held in the library.
The AI runs 40+ interrogation checks on the BHR evidence, identifying vehicle mismatches, locality errors, and unsupported rate reductions.
The AI analyses competing accounts and produces a structured liability assessment with recovery percentages and relevant case law.
The AI provides strategic guidance grounded in real authority. If it cannot find supporting case law, it tells you.
We follow the Civil Justice Council’s position on AI-assisted witness evidence in three concrete ways. First, every generated witness statement carries an AI-use declaration block immediately above the statement of truth, confirming that the statement was prepared from the witness’s own account and that the facts have been checked by the witness. Firms and jurisdictions can override the precise wording.
Second, for trial statements the system is disciplined to assemble, not generate: it quotes the witness’s own supplied wording verbatim where it exists, records the bare fact of a communication where it does not, and flags “wording to be confirmed with the witness” rather than inventing first-person speech. Coverage Check surfaces gaps as questions to put to the witness \u2014 never as suggested answers.
Third, we operate a clear fee-earner responsibility model. LegalDocs Assist is a decision-support and assembly tool. Substantive compliance with CPR Part 32, PD 32 and the professional duties owed to the court rests on the qualified fee-earner who verifies the draft with the witness, confirms the facts, and signs the statement of truth.
What happens when the knowledge base has no answer. The system refuses to fabricate authority and says so visibly in the output. The prompt rule requires the AI to state: “No supporting authority is present in the verified knowledge base for this proposition.” Two independent server-side checks enforce this even if the AI misbehaves: the citation validator strips unverified array entries, and the text-body scrubber replaces any case-name pattern not in the database with “[case reference removed — not in verified knowledge base]”.
These layers are what sit underneath our witness statement drafting software for UK law firms, where the same curated authority libraries and automated CPR Part 32 compliance checks are applied to every draft.
Get in touch to discuss how LegalDocs Assist can support your team.