Tests such as Watson-Glaser give law firms a scalable way to assess how candidates interpret information, identify assumptions, evaluate arguments and reach logical conclusions.
But generative AI has introduced a problem that the traditional online testing format was never designed to handle.
As far back as 2022, Legal Cheek tested ChatGPT against Watson-Glaser-style questions. The chatbot correctly answered the examples covering deduction, recognition of assumptions, interpretation and evaluation of arguments, although it failed the inference question. The experiment followed a student's claim that they had achieved 70% on a mock test using the chatbot's responses.
The technology has moved considerably since then.
The problem isn't critical thinking testing
That doesn't mean law firms need to abandon critical thinking assessments.
The issue is more specific. A conventional text-based test completed remotely may give a candidate enough opportunity to transfer a passage and question into an AI tool and use its response to help select an answer.
Some employers are already responding. At LegalEdCon 2026, Linklaters described re-administering Watson-Glaser at its in-person assessment day, allowing recruiters to compare performance with candidates' earlier online results and identify significant discrepancies.
But there is another option: redesign the online assessment itself.
A middle ground for legal recruiters
Some newer assessments use interactive or game-based tasks that are inherently difficult to outsource to generative AI. However, not every law firm wants to move away from a traditional text-based critical thinking format.
Test Partnership's Signals Critical Thinking is designed as an alternative for those employers.
Candidates still read passages and answer critical reasoning questions. But instead of receiving a conventional block of questions with a broader time allowance, they receive questions one at a time, cannot preview future questions or revisit previous answers, and face a short timer for each individual question. The assessment also adaptively selects subsequent question sets according to estimated ability.
These measures don't make AI assistance impossible, but they make it considerably less practical to capture each question, obtain an external answer and submit it within the available window.
Signals Critical Thinking is also specifically designed for legal recruitment. Its passages use realistic legal scenarios while requiring no prior legal knowledge, so candidates are assessed on the reasoning demonstrated from the information provided rather than what they already know about law.
For legal employers, the choice therefore doesn't have to be between carrying on unchanged with traditional tests and moving entirely to game-based assessment. There is a middle ground: preserve the familiar text-based critical thinking format, while redesigning its delivery for an era in which every candidate potentially has an AI assistant sitting beside them.
