To obtain available education materials, a Google search was performed using a browser without prior search history for the terms “dorsal root ganglion stimulation patient education”. 

To create new materials, the publicly available research preview of ChatGPT was prompted with “Write a patient education document for dorsal root ganglion stimulation. Make it simple enough that a third/sixth/ninth grader could read it”. 

Each prompt was provided five times, to which generated fifteen unique outputs. 

Readability was assessed using five previously validated algorithms: Flesch-Kincaid Grade Level, Coleman-Liau Index, Simple Measure of Gobbledygook Index, Automated Readability Index, and Linsear-Write Formula. 

Comprehensiveness of each article was rated from 0-5 points predetermined metrics, with one point given if the article contained information on indications, contraindications, anatomic targets, procedure description, and complications.

Human-authored patient education materials were written at a mean reading level of 12.6 (mean [SD], 12.6 [1.9]; range, 6.2-16.01), with only one of fifty resources (2%) meeting the target reading level. 

We found that ChatGPT wrote at a higher grade level than prompted. When asked to write at a 6th grade reading level, ChatGPT-authored documents were within the target reading level 50% of the time, with a mean 8th grade reading level (7.98 [1.4]). 

Readability scores were significantly different between ChatGPT and human-authored documents (p<0.005). 

When writing at a 3rd grade level, ChatGPT produced documents within the target reading level 90% of the time, with a mean reading level of 7th grade (6.8 [0.7]). 

There was no statistical difference in comprehensiveness between human and ChatGPT authored texts. Human-authored documents had mean comprehensiveness of 2.76 (0.91), while ChatGPT authored texts at the 3rd grade level had comprehensiveness scores of 2.8 (1.095).

Over the past decade, the use of DRG for managing chronic axial and appendicular pain has increased dramatically. 

Online information explaining the risks and benefits of DRG needs to be accessible, accurate, and comprehensive. 

All search results, had calculated reading levels greater than 12th grade - significantly above the 6th grade reading level recommended by the AMA.

The gap between recommended reading level and actual reading level of the given materials represents a limitation in information accessibility to patients. 

We found that ChatGPT writes at a higher grade level than prompted. Considering current GPT relies heavily on extant human authored literature, 

it is no surprise that the generative model—like the human text it is based on—both provide higher reading levels than desired. The authors recommend use of a deliberately lower reading level to compensate for this observed effect.