Influence of visual design on young users’ attention to digital museum homepages: an eye-tracking study Authors: Hao Fang, Hongyun Guo, Mian Wu, Yuanyan Hu, Minjian Hong, Feng Gan Published: 2026 ABSTRACT Digital museum websites are often the first point of contact with collections, but their homepage designs vary in effectiveness at attracting and guiding attention. We examined how visual design elements affect the perceived appeal and attention distribution on digital museum homepages for young users. Mixed methods were used: interviews and a focus group identified candidate elements, a self-report study (n=70) rated museum homepages, and an eye-tracking experiment (n=40) tested six virtual homepages representing the most and least appealing sites. Paired t-tests and a generalized linear mixed model (GLMM) were used to analyze appeal ratings, and the eye-tracking heat maps with fixation-order metrics characterized attention patterns. Homepages with more hypothesized elements were rated as more appealing. Effect on appeal, while concise layouts (effect. Eye-tracking showed that pages with large images (especially with simple animations) attracted early, focused fixations, whereas pages lacking these features produced more dispersed gaze patterns. These findings suggest that homepages with more appealing images and less distracting elements are more likely to attract attention and encourage engagement. Introduction Digital museum websites have proliferated, extending cultural access and offering personalized, interactive experiences that appeal to young users1–4. These online museum platforms allow users to discover and engage with collections, and the homepage is a critical entry point that shapes first impressions and influences subsequent navigation. However, many digital museum homepages lack visual appeal and fail to guide users’ attention effectively1,5, hindering information exploration—especially for young visitors prone to rapid task switching. This highlights the need to examine how specific visual design elements influence perceived visual appeal and early attention distribution on museum websites. Young adults represent an important audience segment for digital museums; however, their information practices are shaped by choice-rich online environments in which attention is constrained and task switching is routine6,7. Research on media multitasking suggests that heavier media multitasking is often associated with heightened susceptibility to distraction and, in many studies, poorer performance on certain cognitive-control measures, although findings are mixed8,9. Moreover, screen-based reading is frequently characterised by browsing and scanning, keyword spotting, and non-linear navigation rather than sustained linear reading10. In information-saturated contexts, information overload has been discussed as a deleterious condition that can generate processing difficulties and negative affect such as anxiety, thereby impairing effective engagement with information tasks11. For digital museum websites, these behavioural tendencies underscore the need for interface cues that facilitate rapid orientation, efficient evaluation, and sustained engagement with cultural information. Accordingly, examining how young users allocate visual attention and respond to specific page elements is critical for improving information findability and overall experience in digital museums. Website quality is recognized as a crucial factor in enhancing users’ perceived value and revisit intention in digital museums12. Visual appeal, in particular, is a key determinant of quality in knowledge-based web design, such as digital museums13–15. This study treats visual appeal as a composite construct, the core of which lies in users’ judgments regarding the visual appeal of the homepage’s visible design–that is, their cognitive evaluation of its visual appearance16. At the same time, existing research indicates that such evaluations may be accompanied by emotional responses and may be related to other user judgments (such as perceived usability and credibility). Therefore, visual appeal in this study refers to participants’ overall evaluative judgments of the homepage’s visual presentation. Such judgments form first impressions and may further influence interaction willingness and subsequent attention orientation. It can guide observers’ attention, making it easier for them to focus on specific objects or areas17. Furthermore, visual appeal also contributes to prolonging the observers’ sustained attention on specific content, as attractive elements can capture their gaze and reduce attentional dispersion18. From an attention perspective, models and reviews indicate that salient image features can guide early selection, while goal-directed factors can bias selection over time18,19.However, many studies of digital museums evaluate aesthetics or visual appeal at an aggregate level (e.g., overall ratings or quality dimensions), which offers limited guidance on which specific layout schemes and visual elements drive attention and evaluation during actual use2. This limitation is also theoretically important, because HCI research has long suggested that users form rapid aesthetic impressions of webpages and that these impressions can shape broader judgments of interface quality and subsequent interaction tendencies19. Building on this perspective, the study extends the holistic focus on website quality in digital museum research to the specific design features of the homepage, thereby enriching the existing body of research in this field regarding the visual design characteristics of homepages and their significance in terms of user perception.This gap motivates a more element-level analysis of visual appeal in digital museum webpages. To address this gap, we used a mixed-methods approach combining qualitative interviews and a focus group to identify appealing elements and formulate hypotheses. We then tested web pages with these elements using self-report ratings and found their impact on perceived appeal. Finally, we conducted eye-tracking experiments to objectively observe how these design elements attract attention, isolating their effects on user perception of appeal. Research methodology Research 1: a qualitative study of the visual elements of digital museums Exploring appealing visual elements through user interviews Eight university students (4 males, 4 females, age 20–27) participated in the interviews, and each received a small reward. Participants browsed 30 domestic and international museum homepages, then were interviewed about their demographics and visual expectations. To reduce subjective bias in sample selection and to maintain relative consistency of webpage features for comparison, the homepages were randomly sampled within the same website type. Science-communication-oriented digital museum websites were selected as the focal context because prior research suggests that science and technology museum websites tend to perform strongly in presentation, usability, interactivity, and e-services15, these characteristics make such websites a suitable context for examining salient visual design features on digital museum homepages. The interviews focused on participants’ experiences with the homepages. From the interviews, we identified participants’ most and least preferred pages and six key elements influencing visual appeal: bright colors, large images, concise text, simple animations, clean layouts, and 3D effects (Table 1).Table 1Characteristics of visual elements.Visual ElementDescriptionLarge imagesA large image attracts more attention than text and a small image.Short textsFewer texts help enhance readability and aesthetics. Too many texts increase cognitive load.Simple animation effectsAnimation elements can rapidly attract a user’s attention. Conversely, complex animation is distracting.Concise template layoutsTemplate complexity affects a user’s first impression of the site. Template density should be moderate.Bright colorsCompared with darker colors, bright ones are more attractive.3D effects3D effects provide a sense of immersion for users. Table 1Characteristics of visual elements.Visual ElementDescriptionLarge imagesA large image attracts more attention than text and a small image.Short textsFewer texts help enhance readability and aesthetics. Too many texts increase cognitive load.Simple animation effectsAnimation elements can rapidly attract a user’s attention. Conversely, complex animation is distracting.Concise template layoutsTemplate complexity affects a user’s first impression of the site. Template density should be moderate.Bright colorsCompared with darker colors, bright ones are more attractive.3D effects3D effects provide a sense of immersion for users. Characteristics of visual elements. Further filtering of visual elements through focus groups A focus group of four experienced website design experts reviewed the 30 museum homepages and discussed the interview findings to identify the most important visual elements. We aimed to reduce subjective influences and focus on design features that could be clearly identified and compared across webpages. When selecting candidate elements, we prioritized elements that could be defined relatively consistently across the sampled homepages and translated into observable coding features for subsequent quantitative analysis. Due to their variability and difficulty in quantifying, bright colors and 3D effects were excluded20,21. Color can evoke different psychological responses, but its perceived appeal is often closely tied to the semantic tone and communicative intent of webpage content22,23; for this reason, it was not treated as an independent visual factor in the present study Reactions to 3D effects were mixed–some found them novel24, while others felt they added complexity25. Because such divergent aesthetic responses made it difficult to treat 3D effects as a stable evaluative feature across the sampled homepages, they were not included in the final set of visual elements. The analysis identified four key visual elements that attract users: large images, short text, simple animations, and concise template layouts. These were chosen for further quantitative analysis: Large images Large images in web design refer to visually significant graphical elements that occupy a substantial amount of space on a webpage. These images are often utilized to enhance visual appeal, provide a richer visual experience, and emphasize important content or products25. Djamasbi et al. (2010) found that young users showed a stronger preference for webpages featuring a dominant main image, and operationally defined a ”large image” as one occupying at least 40% of the webpage26. Following this study, a large image was operationally defined here as a complete image displayed on a computer screen (1366 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\times$$\end{document} 768) that occupied at least 40% of the visible page area.Short text Based on participants’ perceived text density in previous user interviews and expert evaluations from focus groups, we classified the total text length of the 30 selected museum homepage samples into three categories: high, medium, and low. The average length of the high category was 1325.57 Chinese characters, the medium category had an average of 635.21 Chinese characters, and the low category had an average of 197 Chinese characters. Consequently, we defined three scenarios on the homepage based on the total number of words: long text, medium text, and short text. Short text on webpages refers to concise and compact textual content that conveys information effectively. The average length of short texts in this study (197 Chinese characters) is largely consistent with the typical content length on international and Chinese social media platforms (such as Twitter, Weibo, and WeChat Moments), which supports the ecological validity of this classification. We performed a t-test on the long- and short-text groups and found a significant difference (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$t=9.374$$\end{document}, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$p < 0.005$$\end{document}). Because 377 Chinese characters was the upper bound observed in the low-text group, we used this value as an empirical cut-off for operationalizing short text in the subsequent coding and quantitative analyses.Simple animation effects Simple animation effects in web design refer to subtle and lightweight visual movements applied to elements on a webpage. These effects involve basic and uncomplicated transitions, fades, rotations, or scaling of elements, creating a sense of motion or dynamic change27. Complex animation effects are not the most appealing to users. Rather, increasing users’ pleasant feelings might distract them. In this study, simple animation effects were defined as a few images with scrolling effects (e.g., a simple switching animation effect).Concise template layouts Template layouts in web design refer to pre-designed structures or frameworks that provide a consistent and organized arrangement of elements on a webpage. These layouts serve as a foundation or starting point for designing webpages, offering predefined sections and positioning guidelines for content, navigation menus, headers, footers, and other components28. Concise template layouts often employ a clean grid structure, balanced whitespace, and strategic placement of key elements to guide the user’s attention and facilitate intuitive navigation. The low page rate and its relatively focused visual presentation can improve the overall visual impact of a web page29,30. Having large spacing between lines, text, and images on the page can improve overall website aesthetics. In this study, we defined concise template layouts as consisting of fewer than three-page columns, with large spacing between each visual element. Large images Large images in web design refer to visually significant graphical elements that occupy a substantial amount of space on a webpage. These images are often utilized to enhance visual appeal, provide a richer visual experience, and emphasize important content or products25. Djamasbi et al. (2010) found that young users showed a stronger preference for webpages featuring a dominant main image, and operationally defined a ”large image” as one occupying at least 40% of the webpage26. Following this study, a large image was operationally defined here as a complete image displayed on a computer screen (1366 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\times$$\end{document} 768) that occupied at least 40% of the visible page area. Large images Large images in web design refer to visually significant graphical elements that occupy a substantial amount of space on a webpage. These images are often utilized to enhance visual appeal, provide a richer visual experience, and emphasize important content or products25. Djamasbi et al. (2010) found that young users showed a stronger preference for webpages featuring a dominant main image, and operationally defined a ”large image” as one occupying at least 40% of the webpage26. Following this study, a large image was operationally defined here as a complete image displayed on a computer screen (1366 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\times$$\end{document} 768) that occupied at least 40% of the visible page area. Short text Based on participants’ perceived text density in previous user interviews and expert evaluations from focus groups, we classified the total text length of the 30 selected museum homepage samples into three categories: high, medium, and low. The average length of the high category was 1325.57 Chinese characters, the medium category had an average of 635.21 Chinese characters, and the low category had an average of 197 Chinese characters. Consequently, we defined three scenarios on the homepage based on the total number of words: long text, medium text, and short text. Short text on webpages refers to concise and compact textual content that conveys information effectively. The average length of short texts in this study (197 Chinese characters) is largely consistent with the typical content length on international and Chinese social media platforms (such as Twitter, Weibo, and WeChat Moments), which supports the ecological validity of this classification. We performed a t-test on the long- and short-text groups and found a significant difference (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$t=9.374$$\end{document}, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$p < 0.005$$\end{document}). Because 377 Chinese characters was the upper bound observed in the low-text group, we used this value as an empirical cut-off for operationalizing short text in the subsequent coding and quantitative analyses. Short text Based on participants’ perceived text density in previous user interviews and expert evaluations from focus groups, we classified the total text length of the 30 selected museum homepage samples into three categories: high, medium, and low. The average length of the high category was 1325.57 Chinese characters, the medium category had an average of 635.21 Chinese characters, and the low category had an average of 197 Chinese characters. Consequently, we defined three scenarios on the homepage based on the total number of words: long text, medium text, and short text. Short text on webpages refers to concise and compact textual content that conveys information effectively. The average length of short texts in this study (197 Chinese characters) is largely consistent with the typical content length on international and Chinese social media platforms (such as Twitter, Weibo, and WeChat Moments), which supports the ecological validity of this classification. We performed a t-test on the long- and short-text groups and found a significant difference (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$t=9.374$$\end{document}, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$p < 0.005$$\end{document}). Because 377 Chinese characters was the upper bound observed in the low-text group, we used this value as an empirical cut-off for operationalizing short text in the subsequent coding and quantitative analyses. Simple animation effects Simple animation effects in web design refer to subtle and lightweight visual movements applied to elements on a webpage. These effects involve basic and uncomplicated transitions, fades, rotations, or scaling of elements, creating a sense of motion or dynamic change27. Complex animation effects are not the most appealing to users. Rather, increasing users’ pleasant feelings might distract them. In this study, simple animation effects were defined as a few images with scrolling effects (e.g., a simple switching animation effect). Simple animation effects Simple animation effects in web design refer to subtle and lightweight visual movements applied to elements on a webpage. These effects involve basic and uncomplicated transitions, fades, rotations, or scaling of elements, creating a sense of motion or dynamic change27. Complex animation effects are not the most appealing to users. Rather, increasing users’ pleasant feelings might distract them. In this study, simple animation effects were defined as a few images with scrolling effects (e.g., a simple switching animation effect). Concise template layouts Template layouts in web design refer to pre-designed structures or frameworks that provide a consistent and organized arrangement of elements on a webpage. These layouts serve as a foundation or starting point for designing webpages, offering predefined sections and positioning guidelines for content, navigation menus, headers, footers, and other components28. Concise template layouts often employ a clean grid structure, balanced whitespace, and strategic placement of key elements to guide the user’s attention and facilitate intuitive navigation. The low page rate and its relatively focused visual presentation can improve the overall visual impact of a web page29,30. Having large spacing between lines, text, and images on the page can improve overall website aesthetics. In this study, we defined concise template layouts as consisting of fewer than three-page columns, with large spacing between each visual element. Concise template layouts Template layouts in web design refer to pre-designed structures or frameworks that provide a consistent and organized arrangement of elements on a webpage. These layouts serve as a foundation or starting point for designing webpages, offering predefined sections and positioning guidelines for content, navigation menus, headers, footers, and other components28. Concise template layouts often employ a clean grid structure, balanced whitespace, and strategic placement of key elements to guide the user’s attention and facilitate intuitive navigation. The low page rate and its relatively focused visual presentation can improve the overall visual impact of a web page29,30. Having large spacing between lines, text, and images on the page can improve overall website aesthetics. In this study, we defined concise template layouts as consisting of fewer than three-page columns, with large spacing between each visual element. Hypotheses Through focus group discussions and subsequent visual element screening, we identified four key visual elements in digital museum websites that can attract the attention of young people: large images, short text, simple animation effects, and concise template layouts. Further quantitative research hypotheses are proposed according to the focus group results: Generally speaking, websites that consist of images look better and are more likely to arouse users’ pleasant visual experience31, previous studies indicated that the size of images is closely related to website appeal32, and the large images play an active role in the visual appeal of users during the web page information transmission26,33. Based on this, the research hypothesis is proposed: H1: Websites of digital museums with large images are more appealing than those without large images. Prior research shows that on-screen reading is often characterized by browsing and scanning rather than sustained linear reading34. In addition, higher visual complexity tends to lead to more negative aesthetic first impressions of websites35. Accordingly, the study hypothesizes: H2: Homepages of digital museums with short text gain more appeal than those with longer text. Animation effects not only attract users’ attention with interesting visual effects but also provides a smooth user experience by subtly transforming a static web page into a lively and interactive interface36. However, too many animation effects not only increase the complexity of the page but also increase the cognitive load of the user37. Therefore, we can propose the following hypothesis: H3: Homepages of digital museums with simple animation effects are more attractive than those with no animation effects. According to the study by Parush et al.38, Visual layout has a strong impact on website aesthetics and visual effect and is a key factor in human-computer interface and website design. Excessively high information density diminishes users’ search efficiency. From the above, we hypothesize that: H4: Websites of digital museums with concise template layouts are more appealing than those without. Studies have suggested that the aesthetic response from a single stimulus is not positive enough39. In the study of the shopping website page, it is also mentioned that the combination of large pictures and short text has a high visual appeal to young people40. Therefore, this study speculates that the impact of large images, short text, simple animation effects, and concise template layouts on the web page can interact to further enhance the visual appeal of the page. On these grounds, the study hypothesizes: H5: Websites of a digital museum with a higher number of matching hypothetical feature elements has a higher overall appeal. Research 2: a quantitative study of the visual elements of digital museums General background In the second study, we grouped the 30 homepages according to a predefined coding rule based on the four hypothesized visual elements identified in Research 1, and participants then rated their visual appeal on a 5-point scale. We used paired t-tests to compare the high- and low-feature groups, and a generalized linear mixed model (GLMM) to examine the effects of the individual visual elements while accounting for repeated ratings across both participants and webpages. We used paired t-tests to compare high- and low-feature groups and then conducted eye-tracking experiments to identify which elements attracted users’ attention. Participants Through campus recruitment, 70 participants from the China University of Geosciences were selected from different grades, and majors (32 males and 38 females). Aged between 18 and 28 (23.24±2.05); The web pages were displayed in random order though the whole process. Materials and procedures The experimental materials used in this study consisted of a selection of 30 museum homepages, including the Exploratorium in San Francisco, the Science Centre Singapore, the China Digital Science Museum, the China Science Popularization Museum, and others. These materials were drawn from the same homepage pool described in Research 1. In the experiment, participants freely browsed each of the 30 pages and then rated its appeal on a 5-point unipolar scale ranging from 0 to 4 (0 = not appealing at all, 4 = very appealing), focusing only on the page’s visual layout and elements. A unipolar format was adopted because the purpose of the measure was to capture how much visual appeal was perceived, not to assess a continuum between two opposing qualities. Accordingly, 0 was used as the lowest anchor to indicate that no visual appeal was perceived, and the scale increased with the strength of perceived appeal. Data analysis All statistical analyses were conducted using IBM SPSS Statistics 22.0. Before conducting the paired-samples t-test, we examined the normality of the within-participant difference scores using the Shapiro–Wilk test and Q–Q plots. No substantial deviation from normality was observed. We first classified the 30 experimental materials based on the presence or absence of the hypothesized visual elements identified in Research 1. Each homepage received a feature-count score from 0 to 4 according to how many of the four predefined elements were present. This categorization was specified on the basis of the coding scheme and hypothesis structure. Materials scoring 0 or 1 were categorized as the low-feature group, 2 as the moderate-feature group, and 3 or 4 as the high-feature group. A paired samples t-test of the experimental material for the high and low feature groups found that the results were consistent with the study hypothesis, with a significant difference in visual appeal scores between the two groups (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$t(69) = 10.565$$\end{document}, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$p < 0.001$$\end{document}, Cohen’s \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$d=1.26$$\end{document}). Specifically, the high feature group (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$2.63 \pm 0.421$$\end{document}) had a higher visual appeal score than the low feature group (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$1.68 \pm 0.680$$\end{document}). We then used a Generalized Linear Mixed Model (GLMM) to examine each element’s effect on visual appeal scores, analyzing pages with versus without each feature. In the GLMM, the four visual elements were treated as fixed effects, while participants and the 30 homepage stimuli were treated as crossed random intercepts to account for the variance resulting from repeated measures at both the rater and stimulus levels. It is important to note that interactions between these features were not within the scope of our study. The GLMM analysis results are as follows (Table 2):Table 2Characteristics of visual elements.Visual ElementsNBeta95% CI1p-valueLarge images142.7172.179, 3.256\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$<0.001$$\end{document}Short text12−0.127−0.539, 0.2850.545Simple animation effects17−0.113−0.419, 0.1930.469Concise template layouts12−0.795−1.206, −0.384\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$<0.001$$\end{document}N = The number of webpages containing visual elements; CI = Confidence Interval. Table 2Characteristics of visual elements.Visual ElementsNBeta95% CI1p-valueLarge images142.7172.179, 3.256\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$<0.001$$\end{document}Short text12−0.127−0.539, 0.2850.545Simple animation effects17−0.113−0.419, 0.1930.469Concise template layouts12−0.795−1.206, −0.384\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$<0.001$$\end{document}N = The number of webpages containing visual elements; CI = Confidence Interval. Characteristics of visual elements. N = The number of webpages containing visual elements; CI = Confidence Interval. The coefficients (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\beta$$\end{document}) in the table represent the direction and magnitude of the association between each visual element and the homepage’s visual appeal scores. We observed that webpages with the element of a large image demonstrated a significantly positive effect on visual appeal scores (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\beta = 2.717$$\end{document}, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$p < 0.001$$\end{document}), suggesting that webpages containing large images are more effective at capturing users’ attention. However, other elements, including short text and simple animation effects, did not have a significant impact on the visual appeal scores of the homepage (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$p> 0.05$$\end{document}). Additionally, Concise template layouts showed a negative correlation, indicating that a tidy layout may lead to lower visual appeal scores (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\beta = -0.795$$\end{document}, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$p < 0.001$$\end{document}). Research 3: eye-tracking experiment General background Our self-report study showed that pages with a richer set of hypothesized visual features were generally rated as more appealing. Among the individual elements, only large images were positively associated with appeal, whereas short text and simple animations were not significantly related to appeal. In contrast, concise layouts were associated with lower appeal ratings in the present sample. Despite yielding intriguing results, our user-reported study is subject to the inherent subjectivity of participant reports. We conducted a controlled eye-tracking experiment with standardized virtual homepages (same content, original layouts). This allowed us to objectively observe visual attention patterns and confirm which elements drive appeal. Participants Through campus recruitment, 40 participants from the China University of Geosciences were selected from different types, grades, and specialties (15 males and 25 females). Aged between 20 and 28 (21.85±2.27).All methods were carried out in accordance with relevant guidelines and regulations. The study protocol (including the interviews/focus group, the questionnaire-based rating study, and the eye-tracking experiment) was approved by the Ethics Committee of Wuhan Institute of Technology (approval no. A2025003). Informed consent was obtained from all participants prior to participation. Materials and instrument For this section, we selected the top 3 and bottom 3 digital museum homepages based on visual appeal scores from the participants in the self-report experiment conducted in Section Discussion. These chosen homepages were used as the experimental materials in this section. Interestingly, the museum homepages with the highest mean self-reported scores by the participants were all from the high-feature group with a mean score of 3.22; the museum homepages with the lowest mean scores were all from the low-feature group with a mean score of 1. Taking the copyright into account, we replaced all the contents of six homepages and only kept their layouts. We created six virtual homepages named “geological explorer”, and all the contents belonged to geological science, adopted from http://www.china-shj.org.cn/archive/list-11. html. Doing so can help us to control the variable due to different participants. We used Tobii X2-60 eye tracking equipment with a gaze accuracy of 0.4, gaze precision of 0.2, and a sampling rate of 60 Hz. The websites in this study were displayed on a 15.6-inch screen with a \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$1366 \times 768$$\end{document} resolution. The distance between the participant’s eyes and the screen was set at 60 cm. Procedures First, the participant filled out their information such as name, gender, age, contact information, etc. Before the experiment started, the participant performed a five-point eye calibration. After calibrating to the standard level, the document was set to record the eye tracking test and save the file. To begin the experiment, a fixation point ”+” was displayed randomly on the screen for 1800–2000 ms. A digital museum homepage was then presented, and participants were asked to freely view the page for at least 10 s within the first visible screen area, without scrolling. This setting was adopted to ensure comparability across different homepages in terms of the amount of information visible on the first screen and the proportion of visual elements, and also because homepage viewing typically begins with an initial scan of the visible area, during which users form first impressions and orient themselves to the page. After that, an evaluation table related to the appeal of the homepage (the same table adopted in Self-report method) was displayed The participant was required to score the visual appeal of the homepage. Then a blank screen would display for 2000ms to allow the participant to rest briefly. Each participant needed to complete a total of six trials. To avoid order effects during the experiment, the presentation order of the six museum homepages was randomized. Because the homepage layouts were visually distinguishable, the experimenter was not blinded to condition during data collection. However, the eye-tracking outputs were processed using coded stimulus labels during analysis. During preprocessing, brief signal losses caused by blinks or momentary tracking loss were coded as missing data and were not treated as valid gaze samples. These missing segments were not interpolated, because the subsequent analyses focused on descriptive heat-map and fixation-order patterns rather than on continuous time-series reconstruction. Trials were excluded only when tracking loss was substantial enough that the remaining gaze data no longer allowed a reliable visual interpretation of the overall heat-map distribution or fixation sequence. Research results 1) Visual appeal scores After preprocessing, trials with excessive blink-related signal loss or prolonged temporary tracking loss that substantially reduced usable gaze samples were excluded prior to the descriptive eye-tracking analyses. In the eye-movement experimental study, six web pages from Self-report method - the three most attractive (high feature group) and the three least attractive (low feature group) web pages - were used for visual appeal rating. The results of the paired samples t-test indicated a significant difference in scores between the two groups (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$t(39) = -7.119$$\end{document}, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$p < 0.001$$\end{document}, Cohen’s \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$d=1.13$$\end{document}). Specifically, the high features group (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$2.83 \pm 0.403$$\end{document}) scored higher than the low features group (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$1.86 \pm 0.574$$\end{document}). (2) Heat map The heat map data, generated from eye gaze data collected via an eye-tracking system, visualizes attention patterns with warmer colors indicating higher gaze density. These heat maps were used as descriptive visualizations of gaze concentration and were not intended as substitutes for AOI-based inferential statistics. We collected gaze data for all durations from 40 participants in each museum. Heat maps were used as qualitative visualizations of gaze concentration across the six homepage stimuli. The heat map results indicate (see Figs. 1 and 2) that the hotspots on the high grouping web pages are more easily focused, creating bright red hotspots. In the lower grouped pages, the hotspots are more scattered, which indicates that participants’ attention is more dispersed and all elements seem to be equally important on these pages, and there is no specific attraction element. Specifically, the red part of the heat map appears in the high feature score grouping on the main large images and simple animation effects elements. For example, in the high grouping A (see Figure 1A), there are two red highlighted areas in the interface, and a comparison with the real web page shows that this area has a large image that occupies 40% of the displayed page, and there is a simple image switching motion at this large image; in the high grouping B interface (see Fig. 1B), the bright red area is concentrated at the top of the interface, which contains a simple switching motion of the large image. These observations suggest that large-image regions, including some with simple animation cues, were visually focal in the high-feature pages. In the pages with low feature scores (see Fig. 2), the B and C heat map interfaces were compared with their real pages. The red areas of the heat map appear scattered across different locations on the webpage, indicating a lack of discernible pattern or regularity.Fig. 1Heat map of the high-feature digital museum homepage stimuli A, B, and C. Heat maps were generated from Tobii X2-60 gaze data using Tobii Studio, version 3.3.1, Tobii Technology AB, Sweden, https://www.tobii.com/. Warmer colors indicate higher gaze density.Fig. 2Heat map of the low-feature digital museum homepage stimuli A, B, and C. Heat maps were generated from Tobii X2-60 gaze data using Tobii Studio, version 3.3.1, Tobii Technology AB, Sweden, https://www.tobii.com/. Warmer colors indicate higher gaze density. Heat map of the high-feature digital museum homepage stimuli A, B, and C. Heat maps were generated from Tobii X2-60 gaze data using Tobii Studio, version 3.3.1, Tobii Technology AB, Sweden, https://www.tobii.com/. Warmer colors indicate higher gaze density. Heat map of the low-feature digital museum homepage stimuli A, B, and C. Heat maps were generated from Tobii X2-60 gaze data using Tobii Studio, version 3.3.1, Tobii Technology AB, Sweden, https://www.tobii.com/. Warmer colors indicate higher gaze density. (3) Fixation order We first classified website areas of interest based on two criteria: hypothesized visual elements and heat maps. The first five seconds were selected as the observation window to capture the initial scanning stage of homepage viewing, during which users orient their attention toward salient page regions and form early evaluative impressions. This time window was used to focus the analysis on early attentional organization while minimizing the influence of later, more deliberate exploration41,42. To do this, we used heat maps to categorize areas of interest, facilitating the measurement of user responses to different parts of a website. The gaze order reflects changes in participants’ interest and attention. We identified regions of interest that received fixations within the first five seconds and calculated the average time it took for participants to first fixate on them. Shorter times indicate higher participant interest26. Based on this order, we arranged the fixation order in Figs. 3 and 4, where gray boxes represent areas of interest, and the numbers denote their relative fixation order. Regardless of whether the homepage had high or low visual appeal, the top three in the fixation order concentrated on large images, while the latter ones basically concentrated on text. These descriptive findings provide an initial account of early attentional order and may serve as a basis for future studies using more fine-grained AOI-based statistical analyses.Fig. 3Fixation-order diagram of the high-feature digital museum homepage stimuli A, B, and C. Grey boxes indicate areas of interest (AOIs), and numbers indicate relative fixation order.Fig. 4Fixation-order diagram of the low-feature digital museum homepage stimuli A, B, and C. Grey boxes indicate areas of interest (AOIs), and numbers indicate relative fixation order. Non-essential visual details have been obscured. Fixation-order diagram of the high-feature digital museum homepage stimuli A, B, and C. Grey boxes indicate areas of interest (AOIs), and numbers indicate relative fixation order. Fixation-order diagram of the low-feature digital museum homepage stimuli A, B, and C. Grey boxes indicate areas of interest (AOIs), and numbers indicate relative fixation order. Non-essential visual details have been obscured. (4) AOI-based descriptive statistics of fixation duration To provide quantitative validation of the visual patterns observed in the heat maps and fixation sequences, AOI-based statistical analysis was conducted. AOIs were defined at the interface-element level and grouped into eight categories: large images, animations, navigation bars, search bars, text blocks, small images, generic images, and combination areas. These AOI categories were established for supplementary eye-tracking analysis and do not completely overlap with the four hypothesized design features examined in the self-report study. Mean fixation duration was used as the primary metric because it reflects the average duration of visual fixation within each AOI and allows quantitative comparison across interface elements. Table 3 summarizes the descriptive statistics for each AOI category.Table 3Mean fixation duration for each AOI category in the eye-tracking experiment.AOI categorynMean fixation duration (s)SDLarge images401.190.83Animations402.831.80Navigation bars353.282.25Search bars334.193.64Text blocks404.521.87Small images405.911.31Combination areas406.141.07Generic images386.572.89AOI = area of interest. Mean fixation duration refers to the average duration of fixation within each AOI category across participants. The effective sample size varies because fixation duration was calculated only for participants who generated valid fixations within a given AOI. Table 3Mean fixation duration for each AOI category in the eye-tracking experiment.AOI categorynMean fixation duration (s)SDLarge images401.190.83Animations402.831.80Navigation bars353.282.25Search bars334.193.64Text blocks404.521.87Small images405.911.31Combination areas406.141.07Generic images386.572.89AOI = area of interest. Mean fixation duration refers to the average duration of fixation within each AOI category across participants. The effective sample size varies because fixation duration was calculated only for participants who generated valid fixations within a given AOI. Mean fixation duration for each AOI category in the eye-tracking experiment. AOI = area of interest. Mean fixation duration refers to the average duration of fixation within each AOI category across participants. The effective sample size varies because fixation duration was calculated only for participants who generated valid fixations within a given AOI. Large-image AOIs showed the shortest mean fixation duration (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$M = 1.19$$\end{document} s, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$SD = 0.83$$\end{document}), whereas generic-image AOIs (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$M = 6.57$$\end{document} s, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$SD = 2.89$$\end{document}), combination AOIs (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$M = 6.14$$\end{document} s, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$SD = 1.07$$\end{document}), and small-image AOIs (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$M = 5.91$$\end{document} s, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$SD = 1.31$$\end{document}) showed longer fixation durations. Text AOIs also showed relatively long fixation durations (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$M = 4.52$$\end{document} s, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$SD = 1.87$$\end{document}). These results suggest that although large images effectively attracted initial attention, participants subsequently spent more time processing information-rich AOIs. To further test these differences, paired comparisons were conducted using large-image AOIs as the reference category. As shown in Table 4, fixation duration was significantly longer for animations, navigation bars, search bars, text blocks, small images, generic images, and combination areas than for large-image AOIs (all \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$p < 0.001$$\end{document}). The largest differences were observed for combination areas (mean difference = 4.96 s, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$d = 3.75$$\end{document}), small images (mean difference = 4.72 s, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$d = 2.83$$\end{document}), and generic images (mean difference = 5.38 s, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$d = 1.70$$\end{document}). These statistical results are consistent with the heat-map and fixation-order findings, confirming that large images mainly functioned as early attention attractors, whereas longer fixation duration occurred in information-rich AOIs.Table 4Paired comparisons of mean fixation duration between large-image AOIs and other AOI categories.ComparisonnMean difference (s)Cohen’s dtpAnimations vs. large images401.640.794.98\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$<0.001$$\end{document}Navigation bars vs. large images352.070.925.45\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$<0.001$$\end{document}Search bars vs. large images333.010.864.95\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$<0.001$$\end{document}Text blocks vs. large images403.331.7310.95\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$<0.001$$\end{document}Small images vs. large images404.722.8317.88\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$<0.001$$\end{document}Combination areas vs. large images404.963.7523.74\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$<0.001$$\end{document}Generic images vs. large images385.381.7010.46\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$<0.001$$\end{document}Mean difference = comparison AOI minus large-image AOI. Table 4Paired comparisons of mean fixation duration between large-image AOIs and other AOI categories.ComparisonnMean difference (s)Cohen’s dtpAnimations vs. large images401.640.794.98\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$<0.001$$\end{document}Navigation bars vs. large images352.070.925.45\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$<0.001$$\end{document}Search bars vs. large images333.010.864.95\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$<0.001$$\end{document}Text blocks vs. large images403.331.7310.95\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$<0.001$$\end{document}Small images vs. large images404.722.8317.88\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$<0.001$$\end{document}Combination areas vs. large images404.963.7523.74\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$<0.001$$\end{document}Generic images vs. large images385.381.7010.46\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$<0.001$$\end{document}Mean difference = comparison AOI minus large-image AOI. Paired comparisons of mean fixation duration between large-image AOIs and other AOI categories. Mean difference = comparison AOI minus large-image AOI. Discussion This study aimed to explore the impact of specific visual design elements on the visual appeal of digital museum homepages among young Chinese users. Utilizing a mixed-methods approach that combined qualitative methods (interviews and focus groups) to identify key visual elements, and quantitative methods (user reports and eye-tracking experiments) to assess their impact, the study provided comprehensive insights into user preferences and behaviors. Regarding the first research question, findings from both user reports and eye-tracking experiments revealed a broadly consistent pattern: homepage designs incorporating a richer set of hypothesized visual features were generally evaluated more favorably in terms of visual appeal. This outcome aligns with previous studies highlighting the importance of visual complexity and diversity in enhancing user engagement and emotional responses43,44. Rather than being attributable to the mere accumulation of features, this pattern may reflect the way multiple interface cues jointly contribute to first impressions of a webpage. Visual elements such as large images and animations can enhance perceptual salience, support information presentation, reinforce visual hierarchy, and facilitate attentional orientation45,46. Analyzing individual visual elements yielded several key findings. Firstly, large images emerged as the most robust visual feature in the present analyses and were consistently associated with more favorable evaluations of webpage appeal, consistent with prior research suggesting that prominent images create engaging experiences by capturing users’ attention and interest43,47. From the perspective of interface perception, large images may serve as immediate visual anchors, helping users rapidly identify the thematic focus of a homepage while also strengthening the legibility of the overall layout. However, AOI-based analysis showed that large-image regions did not produce the longest fixation duration. Instead, longer mean fixation durations were observed in text blocks, small images, generic images, and combination areas. This suggests that large images effectively attract early visual attention, whereas sustained visual processing occurs in information-dense interface regions. Secondly, contrary to expectations, short text did not significantly affect the visual appeal of digital museum homepages. One possible explanation is that text length may be less salient during the early stages of homepage viewing, when users are more likely to rely on global layout and prominent visual cues than on detailed reading. This could be attributed to factors such as visual complexity and prototypicality43. It’s possible that short text lacks sufficient visual prominence or that users prefer more detailed information on digital museum platforms. Additionally, in the eye-tracking experiment, participants had limited time for interaction, making it challenging to assess the impact of text length on subsequent exploration and engagement. Accordingly, the role of text length may become more apparent in later stages of interaction, such as information seeking, content comparison, or deeper browsing. Thirdly, while user reports did not highlight the impact of simple animations on visual appeal, eye-tracking heat map analysis suggested that, in high-feature pages, simple animations were often located within the visually focal regions occupied by large main images. This pattern indicates that simple animations were less likely to function as independently salient elements and were more often embedded within regions that already possessed relatively high visual prominence44. In the early stages of page viewing, these regions tended to correspond to more concentrated distributions of attention, suggesting that the perceptual presence of simple animation was closely tied to the salience structure of the surrounding visual field. Simple animations alone may not command sufficient visual prominence, whereas their appearance within already salient image-based regions points to a page-level attentional pattern. This suggests that the contribution of animation may be better understood in relation to the broader attentional composition of the interface, rather than as an isolated design cue. Regarding the second research question, for the participants in this study, a visually simplified homepage layout was not necessarily perceived as more appealing. Previous cross-cultural web design research has noted that Western websites tend to employ simpler layouts and lower display densities, whereas Chinese websites tend to include more elements and adopt denser page layouts48. This difference may be related to the visual and spatial characteristics of Chinese characters, such as the absence of spaces between characters and their relatively compact arrangement49. Against this backdrop, the findings of this study suggest that when designing digital museum homepages for Chinese users, moderate page density and layout complexity may be more effective in enhancing visual appeal. Moreover, the prevalence of mobile internet usage in China has cultivated user expectations for content-dense designs that are easily navigable on smartphones47. Users may perceive minimalist designs as lacking substance or requiring additional navigation to access desired information, which could detract from the perceived utility and appeal of the homepage. For digital museum platforms in particular, this finding indicates that homepage design should not simply replicate minimalist conventions derived from Western interface aesthetics, but should instead balance visual clarity with sufficient informational richness to meet the expectations of local user groups. Conclusions and implications This study concludes that specific visual design elements significantly influence the visual appeal of digital museum homepages among young Chinese users. Large images were consistently associated with more favorable visual evaluations and earlier attentional focus. Eye-tracking results further showed that large-image regions containing simple animations tended to attract concentrated early fixations, highlighting the value of combining multiple visual cues within the overall page composition. Contrary to expectations, short text and clean layouts did not significantly enhance visual appeal, which may be influenced by cultural preferences for information-rich content and layouts. These findings underscore the necessity for digital museum designers to consider cultural factors and user preferences when developing interfaces to ensure they effectively engage the target audience. Based on these findings, several recommendations can be made for digital museum design: (1) Incorporate Prominent Visual Elements: Utilize large images and consider integrating simple animations to capture and retain users’ attention. (2) Embrace Information-Rich Layouts: Design layouts that accommodate a wealth of information without overwhelming the user, aligning with cultural preferences for content richness. (3) Cultural Sensitivity: Understand and integrate cultural nuances into design practices to enhance user satisfaction and engagement. By addressing these considerations, digital museums can create more engaging and effective platforms for science education, better meeting the cultural needs and preferences of young audiences in the digital age. Despite the valuable insights provided by this study, there are limitations that should be acknowledged. The exploration of young users’ interactions with webpages was relatively basic, primarily focusing on initial visual appeal rather than in-depth engagement metrics such as click-through rates or time spent on subsequent pages. Future research should design experiments to investigate users’ subsequent clicking and browsing behaviors to gain a more comprehensive understanding of engagement. In addition, this study used actual museum websites as stimulus materials to replicate real-world browsing conditions as closely as possible. Since it is difficult to ensure complete consistency in page characteristics across such materials, relevant design cues from different stimuli may coexist; consequently, there are certain limitations to interpreting the effects of individual visual elements. Future research could systematically manipulate specific visual variables under more controlled conditions to further examine their independent effects. Moreover, the eye-tracking analysis relied mainly on heat maps and fixation-order patterns, which were useful for showing early attention but remained descriptive. Future studies could incorporate AOI-based measures such as fixation duration, fixation count, and dwell time, together with inferential statistical testing.