Moderate evidence
More
In a survey, 216 students rated a sexual health chatbot well yet had weak plans to use it
Study title: Trust Over Ease: Extending the UTAUT2 Model to Explain Young Adults’ Intentions to Use Sexual Health Chatbots in the Generative AI Era
- Evidence type
- Survey
- Sample
- n = 216
- Sources
- 1 sourcePeer-reviewed
- Published
- 2 min read
- Truvace record
- TRV-2026-2972cite or check it
Study title: Trust Over Ease: Extending the UTAUT2 Model to Explain Young Adults’ Intentions to Use Sexual Health Chatbots in the Generative AI Era
A survey of 216 college students aged 18 to 24 at one US university asked what shapes their intention to use a purpose-built sexual health chatbot. Each student first interacted with Planned Parenthood's chatbot Roo, then completed a questionnaire. A regression model explained 40.3% of the variance in behavioral intention.
Read the study abstract
Abstract Purpose-built sexual health chatbots offer young adults anonymous, clinically vetted information, yet sustained adoption remains low, and general-purpose large language models (LLMs) now compete for the same information-seeking moments without comparable safety assurances. This study extended the Unified Theory of Acceptance and Use of Technology 2 (UTAUT2) with three domain-specific constructs (trust, privacy concerns, and eHealth literacy) to explain young adults’ behavioral intentions to use sexual h…
The problem
In a survey of 216 US college students aged 18 to 24 who tried Planned Parenthood's Roo chatbot, participants rated it favorably but reported weak intentions of routine use.
In brief
Trust in the information the chatbot gave was the strongest predictor of intention, followed by how useful students thought it was. Students judged Roo favorably but reported weak intentions of routine use, a gap the authors link to episodic, need-driven engagement.
The work matters because purpose-built sexual health chatbots offer anonymous, clinically vetted information, yet sustained adoption remains low, while general-purpose large language models now compete for the same moments without comparable safety assurances. The study did not examine those general-purpose models directly.
Main points
- The sample was 216 college students aged 18 to 24 at a single US university, who used Planned Parenthood's Roo chatbot before answering a survey.
- The regression model explained 40.3% of the variance in behavioral intention to use the chatbot.
- Trust in chatbot-provided information was the strongest predictor of intention, followed by performance expectancy.
- Facilitating conditions, privacy concerns and eHealth literacy were not significant predictors, though exploratory analysis suggested eHealth literacy may strengthen the link between perceived usefulness and intention.
- Ease of use correlated positively with intention on its own but had a small, borderline negative coefficient in the full model, which the authors read as suppression because ease of use was near ceiling.
- Results held after adjusting for demographic covariates.
The full story
The researchers extended the Unified Theory of Acceptance and Use of Technology 2 (UTAUT2) with three added constructs: trust, privacy concerns and eHealth literacy. They tested how well the extended model explains young adults' intentions to use sexual health chatbots, using multiple regression on survey responses collected after students used Roo.
Trust and perceived usefulness carried the most weight. Ease of use added little once credibility and usefulness were accounted for, since students already found the chatbot very easy to use. Privacy concerns did not significantly predict intention, even though anonymity is a selling point of such tools.
Despite favorable evaluations, intentions of routine use were weak. The authors describe this evaluation-adoption gap as consistent with students turning to the chatbot only when a need arises, and they offer design implications for purpose-built chatbots in an information environment increasingly dominated by large language models.
Sources
1 on file- Journal of Technology in Behavioral Sciencedoi:10.1007/s41347-026-00762-9
Talk it over
Students rated Planned Parenthood's Roo favorably but reported weak intentions of routine use, and they answered right after trying it. What would that survey design need to add to settle why they would or would not return?
In a survey, 216 students rated a sexual health chatbot well yet had weak plans to use ithttps://truvace.com/item/trust-over-ease-extending-the-utaut2-model-to-explain-young-adults-intentions-to
Students rated Planned Parenthood's Roo favorably but reported weak intentions of routine use, and they answered right after trying it. What would that survey design need to add to settle why they would or would not return?
Answer in the threadAll 216 students were 18 to 24 and came from one US university, and they used only Roo. Who is missing from that group, and how could that change which factors predict intention?
Answer in the threadTrust in the information was the strongest predictor of intention, followed by usefulness. If you were designing a sexual health chatbot, which of those would you work on first, and how?
Answer in the thread
If you have more time
Students answered the questionnaire right after trying Roo once. What would you look up or ask to learn whether their actual use later matched their stated intentions?
Answer in the threadGeneral-purpose large language models now compete for the same moments without comparable safety assurances. How might a school or clinic you know compare them with a purpose-built chatbot like Roo?
Answer in the thread
ace
What readers say