Methodology
Effective Date: October 20, 2026
Last Updated: November 20, 2026
DeeLexy supports legal research and document analysis across multiple jurisdictions. Our approach combines a curated, non-public legal knowledge base, specialised web research tools and the documents and facts supplied by users. We help legal professionals, businesses and individuals examine applicable law, assess legal positions and prepare informed next steps.
Our Code of Ethics sets out the principles that guide this work. Use of the service is governed by our Terms of Service.
1.Curated legal knowledge base
We develop and maintain a multi-jurisdictional legal knowledge base containing legislation, codes, regulatory instruments, judicial decisions, judicial interpretations and authoritative guidance. Coverage varies by jurisdiction and subject. The collection is a non-public research resource: access to its texts and extracted data requires authentication within the service. We do not publish the collection through public pages or an open database-export API.
Materials are processed with attention to their legal structure and source identity. We preserve relevant divisions, provisions, numbered paragraphs and source references, together with jurisdiction, territorial scope and available edition and effective-date information. Automated checks examine source identity, structural integrity and completeness of processing before materials enter the searchable collection. These checks support faithful representation of the source; legal analysis then considers its relevance and application.
2.Jurisdiction, territory and time
Research considers the applicable legal system, territorial scope, hierarchy of authorities, relevant dates and facts. National and territorial materials are distinguished, and an authority from one jurisdiction is not treated as automatically governing another. Legislative requirements, judicial interpretations and non-binding guidance are assessed according to their respective legal roles.
Where the collection holds edition information, research can inspect the provision and its recorded amendments or application notes. We distinguish current law from the law relevant to past events where the available materials permit. Historical coverage varies by source; a record of amendments is not a complete archive of every previous text.
3.Specialised web research and cross-checking
Our specialised web research tools supplement the knowledge base with targeted searches for relevant legislation, judicial practice, regulatory guidance and recent developments. Searches are formulated around the legal issue and directed towards official and authoritative publications. The materials found are assessed for their authority, relevance, timing and relationship to the question.
The AI identifies the substance of the question and the investigation it requires. Its legal knowledge helps interpret the issue, develop hypotheses and decide what to look for. Research then examines the relevant provisions, judicial interpretations, competing positions and exceptions that could change the conclusion. Retrieved materials are assessed for relevance and legal weight before they are used in the answer. Supplementary web research helps address gaps in the collection and examine information against additional sources.
For a material legal assertion, the normal research path leads from a source obtained from the knowledge base, web research or user documents to a verifiable citation. The model's internal knowledge supports reasoning; it does not substitute for source evidence for that assertion. Answers explain the applicable rules and relevant judicial practice in relation to the question, rather than simply assembling search results or lists of provisions.
Reasoning depth lets users balance speed against the extent of investigation. Accurate representation of sources and attention to legal qualifications remain requirements at every depth.
4.Analysis of user documents
DeeLexy considers user-provided facts and documents within the relevant chat or matter. Supported documents are processed to preserve meaningful structure, including headings, numbered provisions and tables; scanned pages and images can undergo text recognition. Research distinguishes the contents of these materials from external legal authorities and from the system's analysis.
User materials remain bound to their case context. They are kept separate from other users' materials, other cases and the shared legal knowledge base. Users can control which available documents participate in the analysis. Submitting a document does not independently establish the truth of its contents.
5.Sources and verifiability
Numbered citations connect an answer to sources obtained and cited during research, whether from the legal knowledge base, the user's documents or web research. The service links these references to the corresponding source records; a marker without an acquired source is excluded from the numbered citations.
Depending on the material, users can examine the relevant provision or passage, open a cited user document, follow a link to the official publication, or view an available saved copy of a legal source. This enables review of the wording, context and qualifications underlying the analysis. A web reference opens the linked publication rather than presenting the answer itself as a quotation.
We distinguish source statements, user-provided information and analytical interpretation. AI-generated analysis is not a judicial decision or an official statement of a court, regulator or legislature. Errors remain possible, including errors of interpretation or application.
6.Currency and quality assurance
We work to keep the knowledge base current by incorporating updated materials and addressing corrections identified at source or reported to us. Recorded effective dates and edition boundaries guide selection of the materials used in corpus research. Targeted web research also supports examination of recent developments, source status and gaps in coverage.
Quality assurance includes automated source-processing checks, retrieval evaluation against reference questions and separate evaluation of legal reasoning and document analysis. We use these assessments and reported issues to identify weaknesses and improve the service. They do not establish exhaustive coverage of every legal system or guarantee correctness in every case.
Legal collections are selective. The absence of an authority from the collection or an answer does not establish that no such authority exists. Access to historical, corrected or withdrawn materials varies by publisher.
7.Human-in-the-loop review
DeeLexy supports legal research, drafting and decision-making through a human-in-the-loop approach. In professional work, the practitioner is the human-in-the-loop: examining the cited materials, evaluating the reasoning and factual assumptions, and exercising independent judgment about advice, documents or proceedings. Source access supports that review.
Users can review the cited provisions and documents alongside the analysis, examine factual assumptions and identify information that may change the conclusion. Our Code of Ethics sets out the principles governing responsible use of information, and our Privacy Policy explains personal-data processing and safeguards.
8.Feedback and corrections
Questions or concerns about an answer, source accuracy, outdated material or privacy may be sent to support@deelexy.com. Identifying the response, the disputed statement and any relevant authority helps us investigate and determine whether source material, analysis or presentation needs correction.