Canonical text (the exact bytes fingerprinted)

TRUVACE RECORD VERSION
record: TRV-2026-2009
version: 1
kind: certified
reason: Certified into the record
timestamp: 2026-10-20T10:42:55.860448Z
status: published
lens: p_space
sector: education
headline: Generative AI vs. Human Sources for Sexual Health Information: Credibility, Trust, and Behavioral Intentions in a U.S. College Sample
dek: Limited access to comprehensive sexual health education has prompted a need to consider the utility of alternative strategies, such as generative artificial intelligence (genAI), for delivering accurate and accessible sexual health information. Although genAI tools can generate health information, less is known about how perceptions of the source of information shape evaluations of content (e.g., human vs. genAI). The present study examined whether college students' perceptions of sexual health information diffe…
gain_title: (none)
problem_title: Students who believed generative AI wrote the sexual health passages judged identical content as less accurate, trustworthy and actionable than content attributed to human sources, which may limit AI's value for sexual health education.
trace_subject: (none)
gain_reading: (none)
gain_evidence: (none)
problem_reading: Students who believed generative AI wrote the sexual health passages judged identical content as less accurate, trustworthy and actionable than content attributed to human sources, which may limit AI's value for sexual health education.
problem_evidence: genAI-attributed information was evaluated less favorably than identical information attributed to human sources
quick_read: Researchers randomly assigned 266 students at one southeastern U.S. university in 2025 to read three identical sexual health passages. Each group was led to believe the author was generative AI, an academic expert, a community health professional, or a peer or social media influencer.

The result suggests that how credible people think a source is can matter as much as the content itself when health information is delivered. It is unclear how far this applies beyond a single-campus student sample, but the authors suggest AI tools may work best when paired with human guidance.
limitation: The sample came from a single university in the southeastern USA, so the findings may not generalize to other student or non-student populations.
tag: Evidence-backed problem
key_points: The sample was 266 students from a single university in the southeastern USA, recruited in 2025. | Passages covered sexual communication, body image, and sexual pleasure and safety, and were identical across conditions. | AI-attributed text scored lower on accuracy, clarity, warmth, helpfulness, trustworthiness, self-efficacy and willingness to act. | Source effects on willingness to act were strongest for sexual pleasure/safety and weakest for sexual communication. | Results held after controlling for familiarity with and use of genAI and attitudes toward it.
rundown: The study was motivated by limited access to comprehensive sexual health education, which has led to interest in generative AI as an accessible way to deliver accurate information. Less was known about whether the perceived source changes how people evaluate that information.

Participants were randomly assigned to one of four source-attribution conditions: generative AI, an academic expert, a community health professional, or a peer or social media influencer. The text was the same in every condition, so differences reflect the attribution rather than the content.

The authors conclude that evaluations depend on perceived source legitimacy as well as informational content. They suggest AI-based tools may be most effective when combined with human guidance to support trust and actionability.
sources:
- peer_reviewed | Archives of Sexual Behavior | https://doi.org/10.1007/s10508-026-03594-z | 2026-10-19
prev: 0000000000000000000000000000000000000000000000000000000000000000

sha256

6e49931c6defc7b24d4dbb109f673878c1c85181d33f81d701b448c02d1276c8

previous

0000000000000000000000000000000000000000000000000000000000000000