📄 medicine

Association of AI Literacy and AI Trust With Subjective and Objective Healthcare-Seeking Delay Among Patients With Gynecologic Malignancies

This cross-sectional study of 300 patients with gynecologic malignancies reveals that higher AI literacy is significantly associated with reduced both subjective and objective healthcare-seeking delays, whereas higher AI trust is linked only to lower subjective delays.

Original authors: Yingying Cai, Chuanying Ding, Xiaohe Ren, Min Li

Published 2026-07-02

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Yingying Cai, Chuanying Ding, Xiaohe Ren, Min Li

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The Big Picture: The "AI GPS" for Health

Imagine you wake up with a strange ache in your body. Before you call a doctor, you might turn to your phone to ask, "What is this?" In the past, you might have Googled it. Today, you might ask an Artificial Intelligence (AI) chatbot.

This study looked at 300 women in China who had been diagnosed with gynecologic cancers (cancers of the female reproductive system). The researchers wanted to know: How does a woman's ability to understand AI (AI Literacy) and how much she trusts AI (AI Trust) affect how long she waits before seeing a doctor?

They measured two types of "waiting":

Objective Delay: The actual number of days between noticing a symptom and seeing a doctor. (If you wait 90 days or more, that's a "delay.")
Subjective Delay: How worried, confused, or hesitant the patient felt while making the decision to go to the doctor.

The Main Characters: Literacy vs. Trust

To understand the results, think of these two concepts like different tools in a toolbox:

AI Literacy is like having a map and a compass. It's the skill to know how to use the AI, how to check if the AI is making sense, and how to spot when the AI might be wrong. It's about competence.
AI Trust is like having faith in the driver. It's the feeling that the AI is reliable and won't lead you off a cliff. It's about confidence.

What the Study Found

1. The "Map and Compass" Effect (AI Literacy)

The study found that women who were good at using and understanding AI (high literacy) did two important things:

They felt less confused: They had fewer mental barriers and worries about making a decision.
They actually went to the doctor sooner: They were much less likely to wait 90 days or more.

The Analogy: Imagine you are lost in a forest. A person with a good map (high AI literacy) can quickly figure out which path leads to safety. They don't panic, and they don't wander in circles for weeks. They know how to use the tool to get help fast.

The researchers also noticed something interesting: The "map" was most helpful for people who started out with very little map-reading skill. Once you are already good at using the map, getting even better doesn't help as much as it did for the beginners.

2. The "Faith" Effect (AI Trust)

The study found that women who trusted AI felt less worried and had fewer mental barriers. However, trust alone did not make them go to the doctor sooner.

The Analogy: Imagine you have a driver you trust completely (high AI trust). You feel calm sitting in the car. But if you don't know how to read the map or if the driver is just guessing, you might still end up driving in circles for a long time. You feel good, but you aren't necessarily getting to the hospital faster.

Why This Matters

The researchers concluded that skills matter more than feelings when it comes to getting medical help on time.

Knowing how to use AI (Literacy) helps you cut through the noise, understand your symptoms better, and decide to see a doctor quickly.
Just trusting AI (Trust) makes you feel better about the process, but it doesn't guarantee you will take action quickly.

The Bottom Line

If you want to help people get diagnosed with cancer earlier, teaching them how to use AI tools critically is more effective than just trying to make them trust the tools. Being able to read the map is what gets you to the hospital; trusting the driver just makes the ride feel smoother.

Technical Summary: Association of AI Literacy and AI Trust With Subjective and Objective Healthcare-Seeking Delay Among Patients With Gynecologic Malignancies

Problem Statement
Gynecologic malignancies present a significant global health challenge, where early diagnosis is critical for survival. However, timely diagnosis is frequently compromised by healthcare-seeking delays. These delays are often driven by the nonspecific nature of early symptoms, leading patients to postpone consultation. Concurrently, the rapid integration of Artificial Intelligence (AI) into digital health environments has altered how patients seek information, interpret symptoms, and make decisions. While AI tools (e.g., symptom checkers, large language models) offer potential benefits, their impact on the timing of medical consultation remains unclear. Specifically, it is unknown how two distinct cognitive factors—AI literacy (the ability to understand, evaluate, and use AI) and AI trust (the belief in AI's reliability)—differentially influence both the subjective experience of decision-making barriers and the objective duration of delay between symptom onset and medical consultation.

Methodology
This cross-sectional study was conducted between September 2025 and February 2026 at a tertiary teaching hospital in Qingdao, China.

Participants: The study included 300 women (aged ≥18) with histopathologically confirmed primary gynecologic malignancies (cervical, ovarian, endometrial, vulvar, or vaginal). Exclusion criteria included recurrent malignancies, metastatic disease at presentation, severe cognitive impairment, or incomplete clinical records.
Measures:
- AI Literacy: Assessed via a 12-item adapted System for Measuring AI Literacy (SMIL) scale, covering navigation, evaluation, and understanding skills (7-point Likert scale).
- AI Trust: Measured using a 3-item scale adapted from Lai et al. (2024), focusing on perceived reliability and willingness to rely on generative AI for health issues.
- Subjective Delay: Quantified using the Perceived Barriers to Health Care-seeking Decision-Chinese (PBHSD-C) scale (10 items), where higher scores indicate greater perceived barriers.
- Objective Delay: Defined as a binary variable representing an interval of ≥90 days between self-recognized symptom onset and the first medical consultation.
Statistical Analysis: Linear regression models were used for subjective delay, and logistic regression models for objective delay. Analyses included unadjusted models and multivariable-adjusted models (controlling for age, marital status, education, income, employment, medical background, and disease history). Restricted cubic spline analyses were employed to detect nonlinear relationships. Subgroup analyses were conducted across age, education, medical background, and screening history.

Key Results
Among the 300 participants, 162 (54.0%) experienced objective healthcare-seeking delay.

AI Literacy: Higher AI literacy was significantly associated with:
- Lower Subjective Delay: A strong negative correlation ( $\beta = -8.72$ , 95% CI: $-10.00$ to $-7.44$, $P < .001$ ).
- Reduced Objective Delay: A significantly lower odds of experiencing a ≥90-day delay ($OR = 0.18$, 95% CI: $0.08–0.41$ , $P < .001$ ).
- Nonlinearity: A significant nonlinear association was observed between AI literacy and subjective delay, indicating a stronger effect at lower levels of literacy that plateaued at higher levels.
AI Trust: Higher AI trust was associated with:
- Lower Subjective Delay: A significant negative correlation ( $\beta = -1.61$ , 95% CI: $-2.47$ to $-0.75$, $P < .001$ ).
- No Significant Objective Delay Association: The association with objective delay was not statistically significant ($OR = 0.69$, 95% CI: $0.45–1.05$ , $P = .084$ ).
Subgroups: The protective association of AI literacy against objective delay was directionally consistent across all subgroups (age, education, medical background, screening history), with no significant interaction effects detected.

Key Contributions

Differentiation of Cognitive Factors: The study distinguishes between AI literacy and AI trust, demonstrating that while both reduce perceived barriers (subjective delay), only AI literacy translates into reduced actual delays in seeking care.
Dual-Dimensional Outcome Analysis: By simultaneously evaluating subjective perceptions and objective behavioral outcomes, the research highlights that attitudinal trust alone is insufficient to modify actual healthcare-seeking behavior in oncology contexts.
Contextual Evidence in Gynecologic Oncology: The findings provide specific evidence regarding AI's role in gynecologic malignancies, a field where symptom ambiguity often leads to prolonged diagnostic intervals.

Significance and Claims
The paper claims that AI-related cognitive skills, particularly AI literacy, play a more central role than attitudinal trust in shaping timely healthcare-seeking behavior among patients with gynecologic cancers. The authors suggest that the ability to critically evaluate and navigate AI-generated health information may empower patients to reduce uncertainty and hesitation, thereby accelerating the transition from symptom recognition to clinical consultation.

The study posits that enhancing patients' AI literacy could serve as a novel strategy to support earlier diagnosis. It suggests that clinical and public health interventions should focus not merely on increasing access to digital tools or fostering trust, but on integrating AI literacy education to ensure patients can effectively interpret AI outputs without falling into decisional paralysis or misplaced reassurance. The authors conclude that while AI trust influences perception, it is the competency to process AI information that drives behavioral change in this high-stakes medical context.

Limitations Noted by Authors
The authors acknowledge the study's cross-sectional design, which precludes causal inference and leaves open the possibility of reverse causality. They also note potential recall bias in self-reported symptom onset, the single-center nature of the study limiting generalizability, and the reliance on self-reported measures for AI literacy rather than performance-based assessments.