Research Summary

Patient-Facing AI Cancer Care Information Often Difficult to Read, Low Quality

Edited by:
Ashton L. Stahl

Key Highlights

  • 31% of screened webpages and 19% of screened videos met the inclusion criteria.
  • Webpage readability was at a college level across 3 validated indices.
  • Few webpages addressed hallucination or misinformation risk.
  • Only 33% of webpages and 38% of videos were rated high quality.

Publicly available patient-facing information about artificial intelligence (AI) in cancer care is limited, often difficult to read, and frequently low quality, according to a cross-sectional analysis published in the Journal of Clinical Oncology. The study evaluated online webpages and YouTube videos intended for patients and found gaps in both accessibility and discussion of key AI safety concepts, including hallucination and misinformation risks.

Researchers conducted a cross-sectional analysis of online patient-facing information about AI in cancer care. Using common cancer- and AI-related keywords identified through Google Trends, they searched Google and YouTube on August 6, 2025. The first 170 webpages and 150 videos were screened for relevance and patient-facing intent, with scientific or industry-facing content excluded. Eligible webpages and videos were independently reviewed by 2 reviewers, with discrepancies resolved by a third reviewer. Webpage readability was assessed using the Flesch-Kincaid, Gunning Fog, and SMOG indices. The content was evaluated for its discussion of AI safety concepts, including clinician oversight, transparency, bias, and the risk of hallucinations or misinformation. Quality of consumer health information was assessed using the DISCERN instrument, with scores of 4 or higher on a 5-point scale indicating high-quality information.

Study Findings

Of the 170 webpages screened, 52 met the inclusion criteria, representing 31% of those reviewed. Most eligible webpage content focused on breast cancer, reported for 30 webpages, or 58%, while 22 webpages, or 42%, addressed cancer more broadly across tumor types.

Readability scores indicated that webpage content was generally written at a college-level standard. Median grade levels were 12.8 by Flesch-Kincaid, with an interquartile range (IQR) of 11.6 to 14.1; 14.8 by Gunning Fog, with an IQR of 13.5 to 16.5; and 14.2 by SMOG, with an IQR of 13.6 to 15.6.

Most webpages discussed clinician oversight and transparency, each reported in 41 webpages (79%). More than half addressed bias, reported in 29 webpages (56%). However, only 8 webpages (15%) discussed hallucination or misinformation risk. Overall, 17 webpages (33%) met criteria for high-quality information based on the DISCERN scores.

Of the 150 videos screened, 29 met the inclusion criteria (19%) of those reviewed. Median view count was 127, with an IQR of 23-1000. Based on the DISCERN scores, 11 videos (38%) were classified as high quality.

Clinical Implications

According to the study authors, these findings highlight the need for accessible, high-quality educational resources that clearly explain AI’s clinical role in oncology and provide guidance for safe patient engagement. The authors noted that inadequate discussion of AI-related risks, particularly hallucinations or misinformation, may leave patients poorly informed as they encounter or independently use AI-based tools.

Expert Commentary

“Publicly available, patient-facing information about AI in cancer care is limited, difficult to read, and often of low quality,” the researchers concluded.

Reference

Subramanian P, Shyamsunder S, Eshaghi K, Mamtani R, Litt HK. Gaps in patient-facing information on artificial intelligence in cancer care: a cross-sectional analysis. J Clin Oncol. 2026;44(suppl 16):9000. doi:10.1200/JCO.2026.44.16_suppl.9000