Raydorf Benchmarks RDF-BMK-2026-1 September 2026
Measuring legal AI
Raydorf Benchmarks evaluates the AI tools lawyers rely on every day against a method announced in advance, under identical conditions, and published together with its limitations. Our first report examines Turkish legal AI research platforms.
§ IWhy we measure
In Turkish legal practice, AI-assisted research platforms are no longer a curiosity; they have settled into daily work. In-house counsel and attorneys now entrust part of their work to these platforms, from case-law searches to the preparation of legal opinions. What they entrust is not small: the accuracy of a citation, whether a precedent is still current, whether an exception has been missed. These are questions at the very centre of professional responsibility.
Yet there is very little reliable information on how these tools should be chosen. The platforms’ own promotional materials, by their nature, offer no basis for comparison. Individual trial and error is not enough either: no single practitioner can realistically test a question set spanning several areas of law across multiple platforms under identical conditions, verify the results document by document, and repeat the exercise every time the platforms are updated.
At the root of this gap lies a single absence: there is no agreed common standard in Türkiye for measuring the accuracy of legal AI tools. A lawyer’s confidence in a tool should rest not on impressions, habit or marketing language, but on verifiable findings produced by a disclosed method. The tradition of conformity assessment has done exactly this in other sectors for more than a century: announce the criteria in advance, apply the measurement to everyone in the same way, and publish the result together with its limitations.
§ IIThe first report
Turkish Legal AI Research Platforms — Pilot Evaluation Report is the first step on that path. It examines Leagle, Apilex and Dejure.ai against 14 common research questions prepared by the Raydorf Institute, spanning associations, environmental, intellectual property, tax, corporate, tenancy, employment, banking and enforcement, criminal and family law, and fundamental rights.
The platforms were evaluated with the same questions, under the same conditions, against metrics fixed in advance and by three separate judge models. Measurement has two layers: the RAGAS framework examines how the answer relates to the sources retrieved, while a law-specific rubric measures citation accuracy, currency of case law, comprehensiveness and caution. The method, data-collection dates and the study’s limitations are declared in the report.
As a natural consequence of its pilot scale, the report makes no claim to a ranking. Our aim is to test the method in the field and to give practitioners early findings they can read against their own use cases. For the same reason, the selection guide does not declare a single winner.
§ IIIThe next phase
In the next phase we are widening the scope, moving to an evaluation cycle that brings the market’s leading legal AI companies into the same framework. Performance criteria will be set jointly with participating companies and practising lawyers, and finalised by the Raydorf Institute; measurement and analysis will be carried out by Raydorf alone. Taking part in setting the criteria will not mean taking part in setting the result.
§ IVIndependence
No fee was taken from any platform in exchange for inclusion or ranking in this evaluation. The Raydorf Institute holds no commercial interest in any of the platforms evaluated. The methodology, metrics and judge guidelines were fixed before answers were collected, and no platform had editorial influence over the content of the results.
The Raydorf Institute’s mission is to make the use of AI in professional services measurable and auditable. We regard impartial evaluation of the tools lawyers use every day as a natural part of that mission, and we commit to doing it on a sustainable, recurring cycle. Every critique of and contribution to our method will make the next cycle better. Our door is open: contact@raydorf.com.
Dr. Taylan Yıldız
Chairman, The Raydorf Institute