Factors influencing user adoption of AI painting tools based on the AIDUA model

Abstract

Artificial intelligence painting tools (AIPT) are increasingly integrated into creative practices, yet designers’ willingness to adopt them remains underexplored. This study applies the artificial intelligence device use acceptance (AIDUA) model, grounded in cognitive appraisal theory, to investigate factors influencing designers’ acceptance of AIPT. Using survey data from 480 respondents and structural equation modeling (SEM), we examined the interplay of supportive factors (perceived usefulness (PU), hedonic motivation (HM), social influence (SI), trust (TRU)) and opposing factors (perceived risk (PR)). Results indicate that hedonic motivation significantly reduces PR, while social consensus enhances trust, both of which shape attitude (ATT) and intentions toward AIPT. Theoretically, this study extends AIDUA to the creative domain, highlighting its value for explaining ambivalent adoption behavior. Practically, the findings provide actionable insights for developers and educators to enhance AIPT acceptance by addressing risk perceptions, improving transparency, and fostering supportive user communities.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-48130-0.

Introduction

Artificial intelligence has become part of everyday life and every business process¹, and its footprint in artistic creation is expanding just as rapidly². Generative imaging systems such as Stable Diffusion, Midjourney and DALL-E are redrawing the boundaries of design and creative practice³ by translating text or parameters into rich digital visuals, offering unprecedented efficiency and ideation opportunities. The shift is no longer hypothetical: at the 2022 Colorado State Fair an AI-generated piece won the fine art division⁴, prompting wide public and academic debate about creative machines.

Scholarly debates persist around fundamental questions: Can AI-generated works be considered art⁵? Do they diminish human creativity⁶? Can designers distinguish between human and AI works⁷? These questions reveal the psychological tension and cognitive conflict that designers experience when confronted with AI-powered art tools. Even when the object is perceptually almost indistinguishable from the original, this response still underscores the central role of the creator in conferring artistic authenticity⁸. Their willingness to embrace such tools directly affects platform sustainability, user engagement and the deeper integration of AI into the creative industries.

Currently, most research on technology acceptance relies on traditional theoretical models, such as the technology acceptance model (TAM)⁹ and the unified theory of acceptance and use of technology (UTAUT)¹⁰. Although these models offer some explanatory power for non-intelligent technology adoption, they fail to fully account for the human-like intelligence, creative capabilities and elevated risk perceptions that characterize AI tools. Consequently, traditional technology acceptance models are theoretically limited in explaining AI tool acceptance in creative domains and cannot reveal the complex psychological mechanisms and cognitive conflicts that designers experience when confronted with AI-powered art tools.

To fill the research gap mentioned above, this paper introduces the artificial intelligence device use acceptance (AIDUA) model based on cognitive appraisal theory (CAT)¹¹, aiming to explore how designers evaluate and decide whether to adopt AI-powered art tools. Unlike traditional models, AIDUA simultaneously accounts for supportive drivers (PU, HM, SI, TRU) and opposing constraints (PR), thereby more comprehensively explaining the psychological contradictions and cognitive conflicts in the process of creative technology adoption. This paper addresses two core research questions. First, which core factors affect designers’ acceptance of AI-powered art tools? Second, how do supportive and opposing factors interact to trigger psychological conflict and shape adoption decisions, thereby verifying the applicability of AIDUA in creative technology contexts and expanding its theoretical scope?

By systematically answering these questions, this paper not only expands the application of the AIDUA model to the creative industries, but also provides a theoretical basis and practical suggestions for AIPT designers, developers and educators, thereby facilitating sustainable integration and value realization of AI tools in artistic creation. Furthermore, this study verifies the dual-path mechanism of AIDUA in creative contexts, thereby compensating for traditional models’ inability to explain the parallel emotions of creation and risk, while providing actionable strategies for developers, institutions and policymakers to remove psychological barriers for designers. The remainder of this paper is structured as follows: first, it reviews relevant theoretical research; then, it proposes a research model and hypotheses; next, it verifies the effectiveness of the model through empirical research; and finally, it discusses the theoretical contributions and practical significance of the research findings.

Theoretical background and hypotheses development

AIDUA

In recent years, research on the predictive factors and foundational mechanisms influencing customer technology acceptance has expanded¹². The TAM has served as the primary analytical framework for evaluating consumer propensity toward adopting technological innovations⁹. For instance, HM has been recognized as a critical determinant of designers’ intention to engage with AI-powered devices¹³. However, there are limitations in applying these traditional models when studying customer acceptance of AI devices¹⁴, as some core driving factors, such as PU, are better suited to explaining the learning process of new technologies¹⁵. Building on TAM, Venkatesh et al. introduced the Unified UTAUT, which primarily examines the factors influencing designers’ behavioral intentions toward the adoption of non-intelligent technologies. UTAUT primarily focuses on the adoption of non-intelligent technologies, including systems such as self-service kiosks and electronic banking¹⁶.

While numerous studies have utilized the TAM and UTAUT models to examine the acceptance of various AI devices, these models fail to encompass the multifaceted effects of AI devices. To address this limitation, Gursoy et al.¹⁷ introduced a theoretical framework for examining AI device acceptance. This model, grounded in the Cognitive Appraisal Theory, identifies six key factors as primary predictors of designer acceptance: utilitarian performance expectancy, effort expectancy, anthropomorphism, HM, emotion, and SI.

Additionally, existing research has primarily focused on designers’ willingness to embrace AI devices, particularly in terms of behavioral intention. In the creative process, designers typically face two choices: AI-generated painting and traditional manual painting. Designers are more likely to adopt AIPT when they find these tools easy to use and effective in performance. Therefore, designer ATT may play a pivotal role in the decision between these two options. From this perspective, studying designers’ willingness to accept AIPT becomes critically important (see Table 1).

Table 1.

Summary of related studies in AIDUA.

Study	Model/perspective	Research focus	Key constructs/determinants	Main contribution	Limitations in AI or creative context
Niemelä et al.12	Technology acceptance research stream	Monitoring customer acceptance of emerging technologies (e.g., service robots)	General acceptance mechanisms	Demonstrates growing interest in understanding predictive factors of technology acceptance	Does not provide a specific theoretical framework for AI device decision-making
Venkatesh et al.9	UTAUT/UTAUT2	Consumer adoption of information technologies	Performance expectancy, effort expectancy, social influence, facilitating conditions, hedonic motivation, price value, habit	Establishes UTAUT/UTAUT2 as dominant frameworks for predicting technology adoption	Mainly designed for traditional IT systems; limited in addressing AI-specific experiential and emotional factors
Gursoy et al.13	AIDUA (AI device use acceptance model)	AI-powered device usage intention	Hedonic motivation	Identifies HM as an important determinant of intention toward AI devices	Focus remains primarily on behavioral intention rather than deeper psychological mechanisms
Lu et al.14	Service robot integration	Willingness to integrate AI/robotic services	Integration willingness factors	Fills the gap on consumers’ psychological attitudes toward service robots in tourism and hospitality	Suggests need for more context-specific constructs
Mortenson and Vidgen15	TAM review	Computational review of TAM literature		Mainly explains learning/efficiency aspects of CLR	Cannot fully capture emotional or anthropomorphic features of AI devices
Sohn and Kwon16	TAM, TPB, UTAUT, and VAM	Adoption of intelligent products	Enjoyment, subjective norms	Traditional TAM is insufficient for AI; enjoyment drives adoption	Excluded some AI/VR tech, high factor correlations, and decomposition analysis needs verification

Research hypotheses derived from a three-stage process

Although the technology acceptance model (TAM) has been extensively validated in explaining user adoption of conventional technologies, its predictive power has gradually diminished amid the rapid evolution of intelligent AI devices¹⁸, which feature autonomy, interactivity, and uncertainty beyond traditional systems¹⁸. This decline underscores the urgent need for a context-specific theoretical framework tailored to AI acceptance to improve both predictive accuracy and contextual adaptability¹⁹. To address this gap, the present study adopts the AIDUA framework, integrating additional factors salient to AI painting tool acceptance, to examine designers’ behavioral intentions toward AI Painting Tools.

As specified in AIDUA theory, designers’ acceptance of AIPT follows a sequential three-stage cognitive process: primary appraisal, secondary appraisal, and outcome stage²⁰. In the primary appraisal, designers form preliminary judgments of the technology’s value based on HM, SI, and TRU. In the secondary appraisal stage, they deliberate on PU and PR, which collectively shape their attitudinal responses toward the tools. The outcome stage stage, informed by prior evaluations, determines whether designers accept or reject he technology¹⁴. In this model, rejection is operationalized as designers’ reluctance to use AIPT and preference for conventional manual painting. This dual-outcome structure distinguishes the model from traditional unidirectional acceptance frameworks and contributes theoretical value to scholars who conceptualize acceptance and rejection as non‑mutually exclusive psychological states. The following sections elaborate the specific constructs and evaluative mechanisms embedded in the AIDUA model.

Primary appraisal

Grounding in the AIDUA model developed by Gursoy et al.¹³, HM, TRU, and SI serve as foundational determinants of the primary appraisal phase for AIPT utilization in artistic creation. If designers’ initial assessment signals that AIPT aligns with their creative needs and expectations, they progress to the secondary evaluation stage, where they systematically weigh the functional benefits and potential risks of integrating AIPT into their creative workflows.

SI reflects an individual’s emphasis on the societal significance embedded in technology adoption, and this influence is typically shaped by observing others’ usage patterns²¹. When individuals lack sufficient knowledge, they tend to rely on the attitudes of their social groups to form judgments²². For instance, colleagues and peers serve as key informational sources in consumer decision-making, and their opinions often influence designers’ acceptance of technology²³. Therefore, if designers’ social networks (such as art enthusiasts) perceive AIPT as useful and recommend its adoption, designers are more inclined to align with these group norms, thereby fostering a positive ATT toward AIPT adoption. The model proposed by Liaw²⁴ explains the influence of environmental factors on PU. In addition, previous studies have also confirmed that SI has a positive effect on PU²⁵. Building on the previous discussion, the following hypothesis is proposed:

Hypothesis 1

SI positively influences the PU of AIPT acceptance.

SI is rooted not only in the TAM but also in relevant models within the field of consumer behavior, such as the Theory of Planned Behavior and the theory of reasoned action²⁶. If core members of an individual’s social group in a work setting believe they should use a particular system, using that system may help enhance their social status within the group²⁷. Both PR and SI are considered key concepts²⁸. Studies have confirmed that SI can have a negative impact on PR²⁹. Furthermore, research has shown that SI may influence designers’ PR regarding AIGC-assisted design tools³⁰. The degree to which designers adhere to group norms may directly influence their risk assessment methods. For example, if the designer community generally holds negative attitudes toward using AIPT for artistic creation and perceives it as risky, its members are more likely to view AIPT usage as a high-risk choice. Conversely, higher group consensus effectively reduces PR. Based on this discussion, we propose the following hypothesis:

Hypothesis 2

SI negatively affects the PR of AIPT acceptance.

HM pertains to the pleasure or expected enjoyment individuals experience from using AI-powered devices¹⁴. According to existing research, HM is a critical factor in predicting user behavior in technological systems³¹. In this study, HM specifically refers to the pleasure or satisfaction that designers derive from the use of AI-powered devices, such as AIPT. When designers experience enjoyment or fulfill personal entertainment needs while using painting robots, this intrinsic pleasure can encourage them to accept and continue using the technology³². Moreover, HM not only enhances designer interest but also increases the willingness to use the technology by providing novel and entertaining experiences, especially when designers perceive these devices as highly useful. In innovative technology systems, intrinsic motivation is a key factor in shaping user perceptions and attitudes toward technology adoption³³. Research has found that HM positively influences users’ PU³. In addition, previous studies have shown that HM has a positive effect on PU³⁴.Therefore, we propose the following hypothesis:

Hypothesis 3

HM positively influences the PU of AIPT acceptance.

According to CAT¹¹, individuals’ evaluation of events generates corresponding emotional states. In this process, individuals must determine whether the events they face imply potential threats or offer potential benefits. Although AI art has been widely applied in the field of AI-generated content (AIGC)², its novelty introduces uncertainty³⁵. From this perspective, considering that new technologies typically lack established usage norms and predictable outcomes, designers may initially associate the hedonic and novelty attributes of AIPT with potential risks. Koenig-Lewis et al.²⁶ proposed that the perceived pleasure users derive from mobile payment services attenuates their PR. Similarly, Wu et al.³⁶ noted that the enjoyment and pleasurable experiences users derive from mobile payment services exert a significant negative influence on their PR. Building upon these perspectives and integrating findings from Abikar³⁷, this study proposes a hypothesis of a direct relationship between HM and PR. Therefore, we propose the following hypothesis:

Hypothesis 4

HM negatively influences the PR of AIPT acceptance.

PT refers to users’ approval of a system’s reliability and credibility³⁸. As dependence on new technologies grows, TRU becomes increasingly important in technology adoption³⁹. In this study, TRU refers to the designer’s approval of the credibility and reliability of AI-powered devices, such as AIPT. With the widespread application of AI technology, TRU is a key factor influencing designers’ decisions to adopt AI technologies and significantly affects their intention to accept these systems⁴⁰. Studies have indicated that a new technology’s TRU plays a crucial role in shaping users’ PU development⁴¹. Previous studies have demonstrated that trust has a significant impact on users’ perceived usefulness⁴². Building on the previous discussion, we propose the following hypothesis:

Hypothesis 5

TRU positively influences the PU of AIPT acceptance.

PR is a key factor in TRU models because TRU aids individuals in risk assessment⁴³. Research based on the TAM demonstrates that TRU exerts a negative influence on PR^44,45. Designers’ TRU in AIPT technology affects their perception of risk, ultimately influencing their decision to adopt the technology. When individuals judge such tools as trustworthy, both their level of fear and PR decrease⁴⁶. TRU plays a crucial role in mitigating uncertainty and reducing PR. Therefore, RU can reduce a user’s PR²⁷. Based on this, we propose the following hypothesis:

Hypothesis 6

TRU negatively impacts the PR of AIPT acceptance.

Secondary appraisal

In this evaluation stage, designers assess the usefulness and risks associated with AIPT, which shapes their attitudes toward the technology. Different from the preliminary judgments in the Initial Evaluation phase, this stage involves a systematic analysis of AIPT’s integration into creative workflows. Designers weigh its perceived benefits against potential drawbacks, and the balance between PU and PR directly forms their attitudes toward AIPT.

PU indicates the extent to which users believe using a particular technology can improve their work and daily life performance⁴⁷. It can also be understood as a subjective evaluation of a technology’s effectiveness from the perspective of potential users⁴⁸. If users perceive that a technology has the potential to improve their workflow or achieve desired outcomes, they are more likely to embrace it. Empirical studies show that users’ attitudes (ATT) toward conversational AI chatbots are significantly influenced by both perceived ease of use (PEOU) and PU⁴⁹. Research indicates a close relationship between PU and ATT toward using the system online⁵⁰. In the context of AI painting, PU reflects an individual’s assessment of the utility and value of AIPT in the creative workflow. This perception is based on designers’ understanding of the tool’s features and functions, as well as its potential impact on artistic workflow and outcomes. The level of PU directly influences designers’ attitudes toward AIPT and their willingness to continue using them. Based on this discussion, we propose the following hypothesis:

Hypothesis 7

PU positively affects ATT toward AIPT acceptance.

PR involves two key aspects: uncertainty (the probability of negative outcomes) and consequences (the significance of potential losses)⁵¹. These risks can be further categorized into psychological, physical, time, performance, and social risks, with their relative importance varying across different product categories⁵². For example, in the context of AI painting applications, performance risk and time costs may be primary concerns for designers. More importantly, these risk dimensions may operate independently of each other and can have varying emotional effects on designers. Previous studies have shown that PR has a negative impact on ATT⁵³. Numerous studies have examined the impact of PR on the adoption of innovative technologies by consumers⁵⁴. The anxiety and uncertainty triggered by PR can undermine designers’ confidence in the technology, ultimately reducing their willingness to adopt AIPT. Existing studies have demonstrated that rising perceived risk adversely affects users’ attitudes toward technology adoption^55,56. Therefore, emotional reactions during the secondary evaluation stage, such as anxiety and a lack of TRU, may significantly influence designer attitudes. Based on this discussion, we propose the following hypotheses:

Hypothesis 8

PR negatively affects ATT toward AIPT acceptance.

Outcome stage

In the outcome stage, designers’ attitudes toward AIPT serve as the primary determinant of their continued usage decisions. This judgment is influenced by PU and PR, such that individuals’ cognitive perceptions and actual usage experiences collectively form their overall evaluation of the tool.

Research has shown that the adoption of emerging technologies is closely linked to designer ATT, which plays a key role in decision-making⁵⁷. Additionally, designer ATT is a critical factor in determining whether a person will adopt a new technology¹⁴. For AI designers, attitudes influence whether they will actively engage with the technology. Previous research has shown that user attitudes have a positive impact on intention to use^58,59. Furthermore, research indicates that attitudes have a negative impact on objection to use⁶⁰. Hans⁶¹ defines ATT toward behavior as an individual’s positive or negative evaluation of a specific action. In the case of AIPT, designers’ attitudes can be either positive or negative, reflecting their overall impressions and willingness to use the tool. Furthermore, individuals with greater confidence in their abilities are more likely to adopt technology than those who are skeptical. Thus, designers’ decisions to adopt or reject AIPT largely depend on their attitudes toward the tool. Based on this discussion, we propose the following hypotheses:

Hypothesis 9

ATT positively impacts the Intention to Use (INT) toward AIPT.

Hypothesis 10

ATT negatively influences the Objection to Use (OBJ) toward AIPT.

Based on the above hypotheses, this study proposes an AIDUA research model (see Fig. 1) to explore the factors influencing designers’ acceptance of AIPT.

Fig. 1.

Adapted the AIDUA model to further develop and validate.

Methodology

This study employs a quantitative research method based on questionnaire surveys. It does not merely verify hypothesized relationships but, adopting the AIDUA model as its theoretical framework, delves into the complex psychological conflicts and contextual characteristics designers experience when adopting AIPT in artistic creation. The research details the questionnaire design, sample selection, data collection, and analysis processes, emphasizing how each step supports the exploration of unique user psychological mechanisms within the creative domain.

Questionnaire design

The questionnaire structure encompasses both basic information and psychological constructs within specific contexts, aiming to reveal designers’ inherent contradictions regarding AIPT. It comprises two core sections: Part One collects standard demographic data (e.g., age, education level) while adding participants’ frequency of AIPT use in artistic creation. This ensures subsequent analysis remains grounded in authentic creative scenarios, avoiding decontextualized measurement. The second part draws upon the AIDUA model while adapting to the unique nature of artistic creation. It incorporates eight items designed to measure key variables reflecting conflict-prone psychological tendencies, aiming to assess designers’ genuine attitudes toward AI-generated art. All items were scored on the five-point Likert scale, with responses ranging from 1 (strongly disagree) to 5 (strongly agree). The questionnaire was initially developed in Chinese and then translated into English using a back-translation procedure to ensure consistency (see Table 2). Additionally, a pilot test was conducted with 145 art enthusiasts before the official release to assess internal consistency and gather feedback. The results indicated that Cronbach’s alpha exceeded 0.8, demonstrating an acceptable level of reliability.

Table 2.

Questionnaire items.

Construct	Items	References
SI	SI1: People who influence my behavior think that I should use AIPT for painting	Venkatesh et al.62 and Yang and Forney63
	SI2: People who are important to me think that I should use AIPT for painting
	SI3: I would use AIPT anywhere
	SI4: In general, the AI development has supported the use of AIPT
HM	HM1: Painting through AIPT is fun for its own sake	Sharifi Fard et al.64 and Chang et al.65
	HM2: Painting through AIPT makes me feelgood
	HM3: Painting through AIPT is one of my favorite activities
	HM4: I am a person who is looking for more fun and enjoyment in painting through AIPT
TRU	TRU1: I would trust AIPT to perform the expected task without any errors	Ribeiro et al.66
	TRU2: I believe that AIPT performance would be trustworthy
	TRU3: My tendency to trust AIPT performance would be high
	TRU4: I would feel confident using AIPT
PU	PU1: Using AIPT would enable me to accomplish painting more quickly	Davis67 and Venkatesh and Davis68
	PU2: Using AIPT in my painting increases my art productivity
	PU3: I find the AIPT to be useful in my painting
	PU4: Using AIPT improves my painting performance
PR	PR1: I am concerned that prolonged use of AIPT might diminish my painting capabilities	Ribeiro et al.66 and Fan and Jiang69
	PR2: I am concerned that prolonged use of AIPT might reduce my independent painting capabilities
	PR3: I am concerned that prolonged use of AIPT might lead to the leakage of my painting works
	PR4: I would feel there is a risk involved with using AIPT
ATT	ATT1: Using AIPT is a good idea	Xu et al.4
	ATT2: I am positively impressed with the ability of the AIPT
	ATT3: I find AIPT to be valuable systems for creating works
	ATT4: I am very satisfied with the artwork generated by AIPT
INT	INT1: I would be willing to use AIPT for painting	Ribeiro et al.66
	INT2: I would feel happy to creative using AIPT for painting
	INT3: I would use AIPT for painting
	INT4: I would use AIPT for painting more than an average person when creating
OBJ	OBJ1: Painting by use of AIPT would lack social contact	Gursoy et al.13 and Ribeiro et al.66
	OBJ2: Painters need emotional exchange with people during creating when painting
	OBJ3: I would prefer to create by human beings when painting
	OBJ4: I prefer human contact in painting

Sample and data collection

This study specifically targeted designers who had accessed or used AIPT, ensuring that the participants had practical experience with AIPT and related concepts. To ensure the representativeness and diversity of the sample, participants were recruited through professional survey platforms, online forums, and design academies. To encourage participation, designers who completed the survey and shared the link were given a chance to enter a prize draw. Data were collected between January 23 and February 3, 2025, using the Chinese online survey platform Wenjuanxing. Of the 500 participants, 480 valid responses were gathered, resulting in a high response rate of 96%. Notably, the sample size exceeds the minimum threshold required for conducting SEM analysis, which is over 200 responses⁷, ensuring that the data analysis is robust and comprehensive.

All research activities in this investigation fully followed the ethical standards set by the Research Ethics Committee of Jiangxi Institute of Fashion Technology (Approval No. JIFT-REC-2025-0004). Official ethical clearance was issued on 16th January 2025. This approval covered all planned steps involving human participant recruitment, data collection, data analysis and subsequent academic publication. Every operational procedure also complied with the core ethical principles outlined in the Declaration of Helsinki.

All participating individuals provided signed consent documents before data gathering commenced on 20th January 2025. The research team directly obtained these formal agreements from adult volunteers; the documents covered study participation, legitimate research use of collected information, and permission to share de-identified data in published outputs. Every participant received comprehensive explanations about the study’s core objectives, their right to join or withdraw freely, and all protective measures put in place to safeguard anonymity and the confidentiality of personal data.

Data analysis

The data analysis framework transcends simple model validation, instead focusing on dissecting the inherent contradictions and particularities of user psychology within the unique context of artistic creation, thereby measuring designers’ acceptance of AIPT. This study employed two methods for data analysis. First, reliability and validity were assessed using SPSS 26 statistical software; additionally, correlation analysis was conducted. Second, the hypotheses of the research model were tested using AMOS 26 software, the model fit was evaluated, and the relationships between variables were analyzed through SEM path analysis.

Analysis of basic population information

In terms of gender, the sample includes 247 male participants (51.46%) and 233 female participants (48.54%), showing a nearly equal distribution. Regarding age groups, the sample covers a wide range, from under 18 to 51–60 years old. The distribution is as follows: 18–20: 13.33%, 21–25 years old: 27.92%, 26–30 years old: 24.58%, 31–40 years old: 18.54%, 41–50 years old: 10.83%, 51–60 years old: 4.79%. Overall, the data suggest that the sample is predominantly composed of younger individuals, aged 18–30. In terms of educational background, the distribution is as follows: Associate degree or below: 24.79%, Bachelor’s degree: 33.54%, Master’s degree: 28.54%, Doctoral degree: 13.13%. This indicates that the sample primarily consists of participants with bachelor’s and master’s degrees. Regarding AIPT usage frequency: Daily use: 27.71%, At least once a week: 37.5%, At least once a month: 20%, At least once a year or more: 14.79%. Overall, the data show a high level of engagement with AIPT (see Table 3).

Table 3.

Basic information analysis.

Name	Options	Frequency (n = 480)	Percent (%)
Gender	Male	247	51.46
	Female	233	48.54
Age (years)	18–20	64	13.33
	21–25	134	27.92
	26–30	118	24.58
	31–40	89	18.54
	41–50	52	10.83
	51–60	23	4.79
Education level	College degree or below	119	24.79
	Bachelor	161	33.54
	Master	137	28.54
	Doctor	63	13.13
Frequency of use of AIPT	At least once a day	133	27.71
	At least once a week	180	37.5
	At least once a month	96	20
	At least once a year or more	71	14.79

Questionnaire reliability analysis

Cronbach’s alpha is a commonly used indicator for assessing the reliability of a questionnaire and is widely employed in empirical data analysis. Typically, a Cronbach’s alpha value below 0.7 indicates poor internal consistency among the variables in the scale, suggesting the need for revision. In reliability assessment standards, a Cronbach’s alpha coefficient exceeding 0.70 is generally regarded as acceptable for reliability⁷⁰. Furthermore, when the Cronbach’s alpha coefficient exceeds 0.9, it reflects excellent internal consistency, indicating very high measurement reliability for the designed variables.

Cronbach’s alpha values for each dimension indicate that all dimensions in this study have values greater than 0.7, suggesting good internal consistency, as shown in the table below. Therefore, the reliability of the survey results is considered strong (see Table 4).

Table 4.

Reliability analysis.

Dimensions	Items	Sample size	Cronbach
SI	4	480	0.827
HM	4	480	0.839
TRU	4	480	0.845
PU	4	480	0.845
PR	4	480	0.849
ATT	4	480	0.835
INT	4	480	0.842
OBJ	4	480	0.843

Validity analysis (KMO and Bartlett test)

In this study, the KMO test and Bartlett’s test of sphericity were performed to determine whether the variables met the requirements for factor analysis, allowing for the extraction of common factors from the original indicators. According to Kaiser’s criterion, the data are not suitable for factor analysis if the KMO is less than 0.5. Typically, when the KMO value falls between 0.6 and 0.7, the data are considered only marginally suitable for factor analysis. KMO values greater than 0.7 indicate that the conditions for conducting factor analysis have been satisfied.

As shown in the table above, the KMO value is 0.907, which exceeds the 0.7 threshold, indicating that the data meet the requirement for factor analysis. Furthermore, Bartlett’s test of sphericity was significant (p = 0.000), further validating the suitability of the data for factor analysis (see Table 5).

Table 5.

KMO and Bartlett test.

KMO	0.907
Bartlett’s test of sphericity
Approximate Chi-Square	7778.444
df	496
p	0

Discriminant validity (HTMT and Fornell–Larcker criterion)

This study employs two distinct approaches to evaluate discriminant validity. First, one key criterion for assessing discriminant validity is to examine the square root of the average variance extracted (AVE). For adequate discriminant validity, the square root of the AVE for each construct should exceed the correlation coefficients between constructs⁷¹. Based on the measurement results in this study, the square root of the AVE for each factor is higher than the corresponding inter-variable correlation coefficients (see Table 6). This finding confirms that the measurement instrument used in this study exhibits satisfactory discriminant validity, with the core concepts measured by different dimensions being independent and clearly delineated.

Table 6.

Discriminant validity (Fornell–Larcker criterion).

	ATT	HM	INT	OBJ	PR	PU	SI	TRU
ATT	0.818
HM	0.425	0.821
INT	0.455	0.365	0.825
OBJ	− 0.493	− 0.428	− 0.411	0.824
PR	− 0.475	− 0.453	− 0.374	0.401	0.83
PU	0.461	0.395	0.378	− 0.329	− 0.382	0.826
SI	0.412	0.339	0.283	− 0.395	− 0.369	0.368	0.81
TRU	0.476	0.414	0.378	− 0.406	− 0.376	0.406	0.378	0.826

Second, this study adopts the heterotrait–monotrait ratio (HTMT) to further assess discriminant validity. To evaluate the correlations among different factors and the internal consistency of each factor, HTMT values should be lower than 0.85⁷². The results indicate that all factors in this study meet the criterion of HTMT values below 0.85, which further supports the presence of strong discriminant validity among the constructs (see Table 7).

Table 7.

Discriminant validity (HTMT values).

	ATT	HM	INT	OBJ	PR	PU	SI	TRU
ATT	–
HM	0.500	–
INT	0.535	0.427	–
OBJ	0.583	0.503	0.482	–
PR	0.555	0.527	0.432	0.473	–
PU	0.543	0.460	0.441	0.386	0.443	–
SI	0.479	0.397	0.325	0.459	0.424	0.428	–
TRU	0.559	0.484	0.437	0.475	0.433	0.473	0.440	–

Variable correlation analysis

There is a significant positive relationship between SI and PU, with a correlation coefficient of 0.352. Similarly, HM is significantly related to PU, with a correlation coefficient of 0.378, indicating a positive correlation. TRU also demonstrates a significant positive relationship with PU, with a correlation coefficient of 0.392.

SI, on the other hand, is significantly negatively correlated with PR, with a correlation coefficient of − 0.347. Similarly, HM shows a notable negative association with PR, with a correlation coefficient of − 0.438. TRU also demonstrates a significant negative correlation with PR, with a correlation coefficient of − 0.360.

ATT is significantly and positively associated with PU (r = 0.449) and significantly negatively correlated with PR (r = − 0.459). Additionally, INT has a positive association with ATT, with a correlation coefficient of 0.442, while OBJ is negatively associated with Attitude, with a correlation coefficient of − 0.482 (see Table 8).

Table 8.

Correlation analysis.

	SI	HM	TRU	PU	PR	ATT	INT	OBJ
SI	1
HM	0.325**	1
TRU	0.362**	0.400**	1
PU	0.352**	0.378**	0.392**	1
PR	− 0.347**	− 0.438**	− 0.360**	− 0.368**	1
ATT	0.390**	0.409**	0.462**	0.449**	− 0.459**	1
INT	0.265**	0.351**	0.362**	0.366**	− 0.359**	0.442**	1
OBJ	− 0.374**	− 0.414**	− 0.393**	− 0.318**	0.395**	− 0.482**	− 0.399**	1

*p < 0.05 **p < 0.01.

The results of the correlation analysis effectively identified the latent relationships among key concepts within the AIDUA model in an artistic context, laying the groundwork for subsequent research to validate causal relationships through SEM.

Model fit analysis

The questionnaire data were imported into AMOS software, and the table presents the model fit indices obtained through the maximum likelihood method. As shown, most of the fit parameters meet the required standards, indicating that the model fits well. Thus, the structural equation model aligns well with the survey data (see Table 9).

Table 9.

Model fit.

Indicator	Adaptation standards	Test results	Accept
CMIN/df	3	2.578	Accept
GFI	> 0.8	0.861	Accept
AGFI	> 0.8	0.839	Accept
RMSEA	0.08	0.057	Accept
NFI	> 0.8	0.853	Accept
IFI	> 0.8	0.905	Accept
CFI	> 0.8	0.904	Accept
TLI	> 0.8	0.895	Accept

Path coefficients within the structural equation model

The results of the path analysis reveal several significant relationships. The path “SI → PU” (β = 0.245, p < 0.001) shows a notable positive effect. Similarly, the path “HM → PU” (β = 0.249, p < 0.001) indicates a substantial positive influence, while the path “TRU → PU” (β = 0.293, p < 0.001) also demonstrates a significant positive impact. In an artistic context, this implies that SI, HM, and PT collectively reinforce designers’ perception of the tool’s value for artistic practice.

In contrast, the path “HM → PR” (β = − 0.388, p < 0.001) shows a significant negative effect, with a similar negative relationship observed in “TRU → PR” (β = − 0.232, p < 0.001). The path “SI → PR” (β = − 0.234, p < 0.001) also reflects a significant negative influence. The findings indicate that positive SI, HM, and PT can alleviate designers’ concerns in artistic contexts, such as fears of losing creative autonomy or compromising artistic authenticity.

Further analysis reveals that “PU → ATT” (β = 0.367, p < 0.001) has a significant positive impact on attitude. On the other hand, “PR → ATT” (β = − 0.451, p < 0.001) demonstrates a considerable negative effect on designer attitude. This indicates that designers’ attitudes toward AIPT become more positive as perceived value increases and more negative as risk concerns deepen.

Finally, the path “ATT → INT” (β = 0.549, p < 0.001) indicates a strong positive impact on intention, while “ATT → OBJ” (β = − 0.564, p < 0.001) exhibits a significant negative impact on the adoption intention. The results indicate that designers’ attitudes toward AIPT directly influence their behavioral tendencies. The more positive the attitude, the stronger the INT AIPT. Conversely, the more negative the attitude, the more pronounced the tendency to oppose AIPT usage. Path analysis reveals how supporting factors within an artistic context interact to counteract opposing risk perceptions, elucidating the psychological mechanisms underlying designers’ acceptance of AIPT and extending the applicability of the AIDUA model to creative domains. These results are summarized in Table 10 and illustrated in Fig. 2.

Table 10.

SEM path coefficients.

Path	β	B	SE	CR	P	Significant
SI → PU	0.245	0.350	0.073	4.783	0.000	Supported
HM → PU	0.249	0.332	0.067	4.930	0.000	Supported
TRU → PU	0.293	0.435	0.077	5.687	0.000	Supported
HM → PR	− 0.388	− 0.388	0.054	− 7.219	0.000	Supported
TRU → PR	− 0.232	− 0.258	0.056	− 4.580	0.000	Supported
SI → PR	− 0.234	− 0.251	0.055	− 4.584	0.000	Supported
PU → ATT	0.367	0.318	0.043	7.320	0.000	Supported
PR → ATT	− 0.451	− − 0.521	0.062	− 8.356	0.000	Supported
ATT → INT	0.549	0.602	0.062	9.765	0.000	Supported
ATT → OBJ	− 0.564	− 0.665	0.064	− 10.429	0.000	Supported

Fig. 2.

Structural equation model results.

Results

This study explores the key factors influencing designers’ support for or opposition to AIPT adoption. The results indicate that SI positively affects PU (β = 0.245, p < 0.001), consistent with findings from previous studies on AI-related technologies and systems⁷³. The research suggests that SI enhances designers’ perception of the practicality of AI technology, thereby increasing acceptance. However, SI negatively impacts PR (β = − 0.234, p < 0.001), highlighting its role in shaping risk perceptions related to AIPT adoption. On a macro level, the stronger designers’ perception of AIPT risks, the lower their willingness to continue using or adopt the technology⁴.

Secondly, HM positively influences PU (β = 0.249, p < 0.001), which is consistent with findings from studies on AI-related technologies^74,75. Specifically, the primary factor driving designers’ adoption of AIPT for artistic creation is the hedonic experience. AIPT offers a platform where designers can create art without requiring traditional painting skills, allowing for a more intuitive and accessible interaction with the system. To further enhance HM, AIPT developers could introduce a variety of painting styles and themes, catering to diverse aesthetic preferences and emotional needs. Additionally, the results indicate that HM negatively affects PR (β = − 0.388, p < 0.001).

Next, TRU positively impacts PU (β = 0.293, p < 0.001). This result confirms the idea that TRU is critical in shaping designer engagement with AI technologies⁷⁶. In the context of AIPT, higher TRU levels increase the likelihood of promoting designers’ behavioral intentions towards AIGC⁷⁷. Additionally, the study shows that TRU negatively impacts PR (β = − 0.232, p < 0.001). Increased TRU reduces designers’ perceptions of uncertainty, making them more willing to accept the potential risks associated with new technology. According to McKnight, building initial TRU requires addressing concerns about risk, which ultimately enhances designers’ willingness to adopt the technology⁷⁸.

Furthermore, PU positively affects the designer’s ATT (β = 0.367, p < 0.001), a conclusion supported by multiple studies⁷⁹. Previous research has shown that PU plays a crucial role in influencing designers attitudes toward the adoption of new technologies and systems. The intuitive interface design of AIPT enables designers to quickly understand its functionalities, enhancing efficiency and fostering a positive ATT toward the technology. Further studies suggest that developing designer-friendly and easy-to-operate technologies can significantly boost designer acceptance, making usability a key trend in future development⁸⁰. Conversely, the findings indicate that PR negatively affects designer ATT (β = − 0.451, p < 0.001). When designers perceive AIPT as highly risky, their positive ATT toward the technology decreases significantly. Perceived risks may include data privacy breaches, loss of creative autonomy, or copyright disputes. For example, artists who fear that AIPT might replicate their style or commercialize their work may develop mistrust toward the technology, leading to a negative attitude. Therefore, minimizing designers’ perceived risks regarding AIPT is essential for increasing acceptance.

Finally, designer ATT positively influences INT (β = 0.549, p < 0.001). An individual’s cognitive ATT toward an object may manifest as a positive or negative evaluation, which in turn affects their attitudinal responses⁸¹. Specifically, a positive ATT increases designers’ willingness to adopt AIPT, while a negative ATT reduces this intention. For example, Sánchez et al. found that students’ attitudes toward AI-based learning programs are a key factor in determining whether they actively engage with the technology⁴. Additionally, designer ATT negatively affects opposition to use (OBJ) (β = − 0.564, p < 0.001). In other words, the more designers recognize the value of the technology, the less likely they are to reject or resist its application. Research has shown that designer ATT is strongly linked to behavior, with a positive ATT often reducing resistance to new technologies.

Discussion

This study addresses the two core research questions proposed earlier. First, PU, HM, SI, and TRU serve as key positive drivers that promote designers’ acceptance of AI-powered art tools, while PR acts as the primary barrier that impedes their willingness to adopt and use such tools continuously. Second, these positive factors not only strengthen perceived usefulness but also reduce perceived risk simultaneously, thereby alleviating the cognitive and emotional conflict experienced by designers. These findings provide empirical support for the dual-path interaction mechanism embedded in the AIDUA model within high-creativity contexts. The results further confirm that the dual-path structure of the AIDUA model effectively captures the complex tension between creative benefits and potential risks, which cannot be fully explained by traditional models such as TAM and UTAUT. When interacting with generative AI tools, designers do not form one‑dimensional rational evaluations. Instead, they hold mixed attitudes that combine expectations of improved creative efficiency and concerns about losing creative autonomy, originality, and control. Therefore, converting skeptical or resistant users into active adopters requires a balanced strategy that simultaneously enhances benefit perceptions and reduces risk perceptions, rather than focusing solely on functional advantages.

This study provides several important theoretical implications. First, it extends the application of the AIDUA framework to the high-creativity and high-uncertainty context of AI-powered art creation tools, verifying the robustness and generalizability of its dual-path mechanism. This expansion broadens the model’s interpretive scope from general intelligent products to creative technology domains and enriches the theoretical system of technology acceptance in artistic fields. Second, this study clarifies the parallel cognitive appraisal process underlying designers’ adoption behavior. Unlike the traditional view that primary and secondary appraisals proceed sequentially, the present findings demonstrate that positive antecedents (SI, HM, TRU) simultaneously strengthen beneficial perceptions and suppress risk concerns. This parallel mechanism supports the real-time hedging function of the dual-path model and advances the theoretical understanding of cognitive appraisal in creative technology acceptance. Third, this study quantifies the interactive effects between facilitating and inhibiting factors, complementing prior research that often examines positive and negative antecedents separately. By revealing how supportive factors counteract perceived risk, this research helps build a more comprehensive and balanced framework for understanding user acceptance of generative AI tools.

From a practical implications, the findings offer actionable and empirically grounded guidance for tool developers, design educators, and policymakers. For developers, improving functional usefulness, optimizing hedonic experience, enhancing social influence, and establishing trust should be integrated into product design. Reducing algorithmic opacity, strengthening copyright protection, and improving creative controllability can effectively mitigate perceived risk and promote sustained usage. For design educators, the dual-path mechanism provides a structured foundation for curriculum development, helping learners understand both the value and potential risks of AI creation tools and fostering rational, healthy, and responsible usage attitudes. For policymakers and industry stakeholders, establishing ethical norms, clear standards, and supportive policies can balance technological innovation and risk governance, facilitating the sustainable and healthy integration of AI in artistic creation.

Conclusion

By applying the AIDUA framework to AIPT, this study provides new insights into the complex decision-making processes underlying creative technology adoption. The results demonstrate that both positive (PU, HM, SI, TRU) and negative (PR) appraisals jointly determine designers’ willingness to accept AIPT. This dual-path perspective enriches acceptance theory by extending AIDUA beyond general consumer technologies into the creative arts. Practically, the findings suggest that developers and educators must balance innovation with trust-building and community engagement to enhance adoption. Future research could diversify participant backgrounds, distinguishing between professional artists, amateur enthusiasts, and designers, to reveal nuanced differences in acceptance patterns. Overall, this study contributes to theory by refining the AIDUA model and to practice by offering concrete guidance for fostering sustainable integration of AI tools in artistic creation.

Research limitations and future studies

This study has several limitations that suggest directions for future research. First, the sample was not differentiated by professional background (e.g., professional artists, amateur enthusiasts, or design students). Since acceptance factors may vary across these groups, future studies should employ stratified sampling or comparative analysis to capture more nuanced differences. Second, the study relied on self-reported survey data, which may be influenced by social desirability or response bias. Future research could complement surveys with behavioural data, such as actual usage logs of AIPT platforms, to validate and enrich the findings⁸². Third, while the AIDUA model provides a comprehensive framework, it was tested within a single cultural context. Designers’ perceptions of creativity, trust, and risk may differ across cultural or regional settings. Comparative cross-cultural studies could therefore broaden the generalizability of the results. Finally, this study focused on adoption intention and actual adoption at an early stage of AIPT development. As AI tools evolve rapidly, longitudinal studies are needed to explore how perceptions of usefulness, risk, and trust change over time and whether early adoption behaviours translate into sustained use. By addressing these limitations, future research can refine the theoretical understanding of AIPT acceptance and provide deeper insights into the evolving relationship between human creativity and artificial intelligence.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Must include all authors, identified by initials, for example: Y.L. conceived the experiment(s), Y.L. and H.L.S. conducted the experiment(s), Y.L. and L.J.G. analysed the results. All authors reviewed the manuscript.

Data availability

Relevant data for the study are either included in the article or provided as supplementary materials. Additionally, datasets used and/or analyzed during this research are accessible from the corresponding author upon reasonable request.

Declarations

Competing interests

The authors declare no competing interests.