UTAUT2: Unified Tech Acceptance Model

• UTAUT2 is a comprehensive model that defines seven key determinants of technology use behavior, including performance expectancy, effort expectancy, social influence, facilitating conditions, hedonic motivation, price value, and habit.
• The model extends its framework with constructs like security, trust, and perceptual realism to enhance predictive power in domains such as generative AI, mobile banking, and automated vehicles.
• Recent empirical studies using SEM techniques and contextual adaptations reveal that factors like habit and performance expectancy consistently drive user acceptance across diverse settings.

The Unified Theory of Acceptance and Use of Technology 2 (UTAUT2) is a comprehensive, empirically validated model designed to explain and predict user acceptance and use of technology. Introduced by Venkatesh, Thong, and Xu (2012), UTAUT2 extended the original UTAUT by incorporating additional constructs and refining paths for consumer adoption contexts. Recent empirical and conceptual work continues to adapt and extend UTAUT2 for diverse domains, including generative AI, mobile banking, automated vehicles, platform fairness, and cross-cultural technology adoption.

1. Core Constructs and Mathematical Foundations

UTAUT2 includes seven principal determinants of behavioral intention (BI) and technology use behavior (UB):

• Performance Expectancy (PE): The degree to which using a technology provides utility or benefits in achieving goals.
• Effort Expectancy (EE): The perceived ease of use associated with the system.
• Social Influence (SI): The extent to which important referents (peers, supervisors, family) encourage use.
• Facilitating Conditions (FC): Users’ perception of organizational and technical support.
• Hedonic Motivation (HM): The perceived enjoyment or pleasure derived from use.
• Price Value (PV): The trade-off between perceived benefits and monetary cost.
• Habit (HB): The extent to which use becomes automatic through prior experience.

The canonical UTAUT2 structural equations are typically expressed as:

$$\mathrm{BI} = \beta_1\, \mathrm{PE} + \beta_2\, \mathrm{EE} + \beta_3\, \mathrm{SI} + \beta_4\, \mathrm{HM} + \beta_5\, \mathrm{PV} + \beta_6\, \mathrm{HB} + \epsilon$$

$$\mathrm{UB} = \gamma_1\, \mathrm{BI} + \gamma_2\, \mathrm{FC} + \gamma_3\, \mathrm{HB} + \epsilon$$

Moderating factors such as age, gender, and experience were part of the original specification but are variably included in empirical extensions.

Measurement of each construct typically involves multi-item Likert scale instruments as operationalized in large-sample SEM or PLS-SEM applications (Apaua et al., 2022, Lambiase et al., 8 Sep 2024, Voria et al., 18 Dec 2024, Lambiase et al., 3 Apr 2025).

2. Empirical Applications and Key Findings

Extensive recent work illustrates the adaptability of UTAUT2 across contexts:

• LLM and Generative AI Adoption: Studies in enterprise and higher education contexts operationalize PE as “perceived usefulness,” EE as “perceived ease of use,” and so forth. Empirical findings underscore PE, HM, and HB as consistently strong predictors of adoption intent, with SI often exhibiting weaker or context-dependent effects. Frequency or habit of prior use frequently emerges as a dominant explanatory variable, and practical interventions often recommend organizational support for direct experience to increase acceptance (Agossah et al., 2023, Xie et al., 17 Jun 2025).
• Mobile Banking: UTAUT2 successfully models adoption, but with substantial enhancement of predictive validity upon inclusion of domain-specific constructs such as perceived security (PS), institutional trust (IT), and technological trust (TT), which directly drive BI and actual UB. In this context, traditional PE and SI may become non-significant, while demographic moderation becomes salient (Apaua et al., 2022).
• Automated Vehicles: UTAUT2 constructs—especially SI and PE—are robust predictors of BI. Cross-national differences are observable: U.S. users emphasize SI, PE; European users weight HM. Extended path modeling reveals indirect effects (e.g., SI and HM driving PE; FC affecting EE and HM) (Saravanos et al., 18 Feb 2024).
• Text-to-Video (T2V) AI: Extended models (e.g., SORA use) integrate perceptual realism and novelty value, which independently and jointly drive PE and willingness to use, outstripping traditional constructs in explanatory power—PR (visual fidelity, realism) is the largest single driver of behavioral intention (Mvondo et al., 7 May 2024).

A representative table of construct–context mappings is below:

Domain	Main Significant Predictors	Core Measurement Highlights
LLM/Generative AI (enterprise)	PE, HM, HB	5- or 7-point Likert; frequency of use critical
Mobile banking	PS, IT, TT, PV, HB	7-point Likert, domain-specific trust/security
Shared micromobility	EE (barrier), SI, HM, reliability	EFA/LCCA segmentation; ~50 items
Automated vehicles	SI (US), PE, EE (indirect)	PLS-SEM; extended inter-construct paths
T2V (SORA)	PR, NV, PE, SI, HM	Combined SEM and fsQCA; 6-driver “recipes”

3. Model Extensions and Contextualization

Recent research adapts UTAUT2 to domain idiosyncrasies by introducing new constructs or merging with other theoretical frameworks. Notable extensions include:

• Warm-Glow: Augmentation with intrinsic and extrinsic “warm-glow” captures both personal and social-approval motivations beyond classical UTAUT2 hedonic and social constructs. Intrinsic warm-glow, in particular, exerts a strong direct and total effect on behavioral intention (Saravanos et al., 2022).
• Security and Trust: Constructs such as perceived security, institutional trust, and technological trust provide significant incremental explanatory value, especially in privacy- and risk-sensitive verticals (e.g., banking) (Apaua et al., 2022).
• Perceived Realism and Novelty: In AI-generated content, perceived realism and novelty value are critical antecedents of both PE and BI. Their inclusion in adoption models significantly raises explained variance in intention and actual use (Mvondo et al., 7 May 2024).
• Ethical, Political, and Cultural Layering: Qualitative and mixed-methods studies expand UTAUT2 with context-sensitive factors—regulatory barriers, domestic–global model access, cultural fit, and affective ritual behaviors—yielding richer models for cross-jurisdictional technology adoption (Xie et al., 17 Jun 2025).

4. Methodological Paradigms and Measurement

Standard UTAUT2 implementations employ covariance-based SEM or PLS-SEM, with rigorous evaluation of construct validity (Cronbach’s α, composite reliability, AVE, HTMT), indicator loadings (>0.7), and fit indices (CFI, TLI, RMSEA for CB-SEM). Recent studies report R² values for BI up to 0.881 and for UB near 0.547 with full models, with domain adaptations (e.g., additional predictors or mediation effects) increasing explanatory power (Apaua et al., 2022, Saravanos et al., 18 Feb 2024, Lambiase et al., 8 Sep 2024).

In domain-specific applications, task-level analyses and clustering approaches (e.g., latent class cluster analysis, LCCA) have surfaced to identify heterogeneity among user segments and task-specific determinants (Geržinič et al., 15 Apr 2025, Lambiase et al., 3 Apr 2025).

High-dimensional exploratory factor analysis, multi-condition fsQCA, and the use of moderation analyses for demographic and cultural variables further refine the operational accuracy of UTAUT2 constructs in empirical settings (Chukwuere, 28 Mar 2024, Ulloa et al., 10 Jul 2025).

5. Key Determinants, Moderators, and Practical Implications

Across applications, the most consistent drivers of behavioral intention are performance expectancy and habit. Effort expectancy and hedonic motivation play significant roles in contextually demanding or pleasure-oriented technologies. Social influence demonstrates variable relevance—crucial in collectivist or highly networked settings, minimal in privacy-dominated or individual-use cases. Facilitating conditions predict actual use more than intention, especially when behavior requires organizational or infrastructural support.

Moderation by demographics (age, gender, education, experience) is frequently tested, with some studies finding substantial heterogeneity—particularly with security and trust constructs in consumer finance contexts—but nonsignificant moderating effects for cultural variables in high-performing, utility-dominated domains (LLM adoption in software engineering) (Lambiase et al., 8 Sep 2024, Ulloa et al., 10 Jul 2025).

Practical recommendations routinely include:

• Fostering habitual use via integration and training.
• Emphasizing demonstrable performance/utility gains.
• Addressing domain-specific barriers (security, regulatory, realism).
• Segmenting interventions by user group, need, or barrier.

6. Limitations, Critiques, and Future Directions

Empirical and conceptual studies elucidate both the strengths and boundaries of UTAUT2:

• Omission or adaptation of constructs is common (e.g., Price Value in open-source, zero-cost adoption scenarios).
• Constructs such as PE and SI may lose predictive power in certain domains; conversely, new constructs (trust, realism, cultural fit) may be dominant.
• Reliance on self-report, cross-sectional surveys without behavioral log validation introduces bias; field-experimental and longitudinal work is recommended.
• Model performance can be highly sensitive to context, necessitating careful operationalization and reporting of item-level measurements and invariance tests.
• A plausible implication is that as new technologies and societal challenges emerge, UTAUT2 will continue to require adaptation—both at the construct and measurement levels—to remain explanatory and actionable across socio-technical frontiers (Agossah et al., 2023, Xie et al., 17 Jun 2025, Lambiase et al., 8 Sep 2024).