Can LLMs Estimate Lexile Measures?

Key Findings

LLM-estimated Lexile® measures are not a reliable substitute for official measures.

• Only 3.03% of the Lexile measures estimated by the four LLMs matched the official Lexile measures.
• In more than 60% of cases, LLM-estimated Lexile measures fell outside ranges that would be optimal for students.
• The LLMs also differed substantially from one another, reinforcing the limitations of using their estimates as proxies for official Lexile measures.

Background Summary

Through field research, we discovered that educators use large language models (LLMs) to identify classroom texts appropriate for their students’ Lexile measures. LLMs are popular, widely used, and readily provide Lexile measures when prompted, sometimes with confident but inaccurate justifications. However, the accuracy of these LLM-generated estimates is not well established.

Read the full report