Report finds AI tools unreliable for providing accurate immigration legal advice
A new study reveals that artificial intelligence models frequently provide incomplete or incorrect answers regarding complex immigration scenarios and recent policy changes.
A recent report from the startup Vals AI evaluated how well popular artificial intelligence models perform when answering questions about immigration law scenarios. The study assessed these tools against expert-provided rubrics to determine the accuracy and completeness of their responses compared to professional legal standards.
According to the report, even the best-performing model provided incomplete answers in the vast majority of situations tested. The findings concluded that AI tools “should not be relied on to independently advise clients on their specific situation. The gaps in applying the law to a client’s facts are particularly concerning.”
The analysis highlighted that AI models struggle significantly with navigating recent changes in immigration policy. When asked about new regulations and policies from 2025 and 2026, even models capable of searching the web for recent content provided less than a third of the information experts deemed necessary.
The category of questions where AI models were least likely to give complete and correct answers involved temporary or fragile status, including Temporary Protected Status and Deferred Action for Childhood Arrivals. This means individuals most vulnerable to losing their status are also the most likely to receive unreliable answers from these automated systems.
Experts note that AI models are limited by the text they are trained on, which poses problems in a field where laws change frequently. Furthermore, these models do not know what details a user should disclose, such as timeframes relevant to asylum applications, yet they often present answers with a confident tone that masks uncertainty.
While a 2025 survey indicated that only 6% of Americans would rely on AI for their immigration case, the integration of these tools into search results raises concerns. Users may encounter AI-generated text without realizing it, making it difficult to distinguish between reliable information and automated responses.