AI-assisted legal research makes it possible to retrieve legal rules, court decisions and elements of reasoning from a question put in natural language. In Switzerland, its usefulness depends on one essential condition: the sources proposed must be capable of being consulted, verified and assessed according to their legal weight.
AI replaces neither specialised documentary databases nor the lawyer’s work. It helps to identify the relevant texts and saves time in the first stages of research. It is then for the professional to examine the references, assess their scope and build the argument.
What does AI-assisted legal research consist of?
Traditional legal research relies on keywords, statutory references and filters, in particular by court or by date. This method remains particularly effective when one already knows which texts or concepts to look for.
Artificial intelligence offers another way of approaching research. Rather than entering a succession of keywords, the lawyer can describe a situation, set out the relevant facts or put a legal question directly. The tool can then identify the concepts at issue, suggest applicable provisions and steer the search towards relevant decisions, even where those decisions use a vocabulary different from that of the file.
AI-assisted legal research can in particular be used to:
- identify provisions of federal or cantonal law;
- retrieve judgments of the Federal Supreme Court or of cantonal courts;
- set different solutions adopted in the case law side by side;
- map the main legal questions raised by a file.
This approach is distinct from reviewing the exhibits in a mandate, such as analysing a contract or a set of internal documents. To go further on that point, see our article on AI-assisted legal analysis. Our guide to Swiss case law and AI deals more specifically with searching for judgments. You will also find a general overview in our article on artificial intelligence applied to law.
Why is a keyword search not always enough?
A keyword search assumes that one knows the terms used in the legal sources. Yet the same question can be expressed in different ways. This difficulty is particularly acute in Switzerland, where legislation and case law are drafted in several languages.
A decision that is in fact relevant may therefore fail to appear if it uses terms different from those in the query. Conversely, a search that is too general may bring up numerous results belonging to another branch of law or another canton.
AI-assisted research is of interest when it makes it possible to query the content of texts from the meaning of a question, without being limited to an exact match of words. It must, however, leave the lawyer the possibility of refining the results according to classic criteria: court, date, type of source or language. Above all, each result must point to a text that can be consulted, whose relevance can be checked against the facts of the file.
The particular features of Swiss law call for greater caution
In Switzerland, rigorous research must distinguish federal law from cantonal law, legislation from case law and, where relevant, doctrinal analysis. These sources do not have the same nature or the same authority. An ordinance is not to be confused with a statute; likewise, a cantonal decision does not necessarily have the same weight as a leading judgment of the Federal Supreme Court.
Fedlex is an official reference for consulting federal law. As regards case law, it is essential to be able to access the original decision, read the relevant recitals and determine to what extent they apply to the situation under review. An imprecise, untraceable reference, or one attributed to the wrong court, should not be reused without verification.
General-purpose assistants can sometimes confuse these different sources, invoke a rule drawn from French or European law, or even mention a provision that does not exist. Our comparisons ChatGPT or Silex and Claude or Silex examine the differences between these tools and an assistant specialised in Swiss law.
How should you formulate a relevant research question?
The quality of the results depends in part on the precision of the question asked. A request as general as “seller’s liability” is likely to produce an unfocused answer. Conversely, a formulation that is too restrictive may lead useful provisions or decisions to be overlooked.
To steer the research effectively, it is generally advisable to specify:
- the determining facts, without unnecessarily communicating the entire file;
- the legal question to be examined;
- the applicable law, in particular federal or cantonal, where it is known;
- the period concerned or, where relevant, the legislative reform at issue.
From these elements, AI can suggest leads, connect several sources and help prepare a first research note. It does not, however, decide which argument to prefer or which conclusion to retain. Those choices belong to the lawyer, who must test the results against the original texts and the particular features of the mandate.
Why must every reference be checked?
An answer generated by an AI does not, on its own, constitute completed legal research. To be usable, it must make it possible to consult the provisions, judgments or documents on which it is based. A summary, however convincing, never dispenses with reading the original source.
Before mentioning a reference in a note, a legal opinion or a brief, it is in particular necessary to check:
- the existence of the text or decision cited;
- the accuracy of the references, in particular the article number, the date and the court;
- the exact content of the passage on which the analysis is based;
- any legislative amendments or developments in the case law since then.
The fluency of an answer does not guarantee its correctness. If AI makes it possible to spot sources more quickly, their interpretation and use still require the same professional rigour. The guidelines adopted in June 2024 by the Swiss Bar Association also recall that the use of artificial intelligence does not relieve the lawyer of the duty to verify.
Connecting legal sources with the exhibits in the file
A legal question cannot be examined independently of the facts underlying it. Those facts often appear in contracts, emails, notes or other exhibits in the mandate. A relevant research tool must be able to take these elements into account, while respecting the requirements of confidentiality and data protection.
At Silex, the Silo feature makes it possible to connect internal documents with the legal sources consulted. This approach helps in particular to test a contractual clause, a determining date or an internal precedent against the rules that may apply to the file.
How these tools work is presented on the page devoted to the Silex product. Issues relating to data are addressed in our article on AI and data protection, as well as on the security and professional secrecy page.
Integrating AI-assisted legal research into professional practice
A research tool is easier to adopt when it fits into the software already used by law firms and in-house legal departments. Professionals can thus access research functions without multiplying changes of working environment.
The integration of Silex into WinLex and SkyLex makes it possible, for example, to launch legal research directly from the file in progress. The partnership between Silex and MCR Solutions follows the same approach.
Beyond technical integration, organisations have an interest in defining a precise framework for use: authorised tools, the nature of the information that may be transmitted, and how results are to be checked. Our guide to enterprise AI strategy sets out the main steps in this approach. Professionals practising in a firm will also find recommendations in our article on AI for lawyers.
Legal assistant, documentary database or general-purpose AI: what are the differences?
Not all tools that use artificial intelligence for legal research perform the same functions.
A documentary database gives access to a body of legislative texts, decisions and, often, licensed doctrinal publications. A general-purpose assistant can explain a concept or help formulate a question, without necessarily guaranteeing access to accurate and verifiable Swiss references. An assistant specialised in legal research, for its part, relies on a defined corpus and must make it possible to consult the sources invoked in support of its answers.
These solutions are not necessarily competing: they can be complementary. Our comparison Swisslex or Silex explains in particular what distinguishes a legal library from a research assistant. To go further, see also our guide to the criteria to take into account when choosing an AI-based legal assistant.
What Silex offers for legal research in Switzerland
Silex is a platform designed for research and analysis in Swiss law. According to its publisher, it is based on a corpus of more than 800,000 Swiss legal sources, including in particular legislative texts and federal and cantonal case law. The answers provided are accompanied by references that the lawyer can consult and check before using them.
Silex also states that data is hosted in Switzerland, with Infomaniak, and that its clients’ files are not used to train the models. The Silo feature also makes it possible to connect legal research with the internal documents of a firm or organisation.
The platform is offered in three plans: Solo at CHF 149, Team at CHF 490 and Enterprise at CHF 895 per month, excluding VAT. The features of the various offers are set out on the Silex pricing page.
The private bank Bordier & Cie has adopted Silex for its legal research. The partnership between Silex and EXPERTsuisse concerns in particular audit, tax and fiduciary advice. The various use cases are presented on the page of solutions by area of practice.
To discover the platform from a concrete situation, you can book a demo.
In short
AI-assisted legal research makes it possible to move more quickly from a question put in natural language to the provisions and decisions that may answer it. It does not, however, dispense with consulting the original sources or exercising one’s own legal judgement.
In Switzerland, the quality of a tool rests in particular on three elements: a corpus suited to federal and cantonal law, direct access to the references cited, and a clear distinction between legislation, case law, doctrine and documents specific to the file.
Whether it appears in a legal opinion, a brief or a procedural exhibit, any reference suggested by an AI must be checked in its original source before it is used.
FAQ
What is AI-assisted legal research?
It consists in using artificial intelligence to identify legal rules, court decisions or lines of reasoning from a question put in natural language. A tool intended for professionals must make it possible to consult the references proposed and to assess their legal weight.
Can an AI replace a documentary database such as Swisslex?
Not necessarily. A documentary database gives access in particular to licensed doctrinal publications and specialised documentary holdings. A research AI makes it easier to formulate questions and identify relevant sources. The two approaches can therefore complement each other. Our comparison Swisslex or Silex returns to their differences.
Is ChatGPT enough to carry out legal research in Switzerland?
ChatGPT can be useful to understand a concept, explore a question or structure preliminary research. By contrast, when an answer is intended to be used in a file, it is essential to check the references, to ensure that they fall under the applicable Swiss law and to consult the original texts.
How should you put a good question to a legal research AI?
Set out the essential facts, formulate the legal question clearly and specify, if you know it, the applicable legal framework. Avoid requests that are too general, but do not restrict the search unnecessarily. Then examine each source proposed before using it.
How should you choose an AI for legal research in Switzerland?
Examine the coverage of federal and cantonal law, the possibilities of access to sources, the conditions of hosting and use of data, and integration with your organisation’s software. Our article on choosing an AI-based legal assistant sets out these various criteria.
