Consumer acceptance of the use of artificial intelligence in online shopping: evidence from Hungary

Introduction

• Retailers’ global spending on AI services was expected to reach $12 billion by 2023, alongside adoption by more than 325,000 retailers.The passage presents these figures as an expectation reported by Maynard (2019).Source: Introduction · p. 2
• AI can support personalised, customer-centric shopping and more efficient sales, but technology-driven transformation is based on trust.The authors cite Amazon as an example of localised shopping journeys, including location-specific pricing and destination-specific messages.Source: Introduction · p. 2
• Few researchers had examined trust and consumer acceptance of AI in online retail, motivating a TAM-based framework and analysis of trust, usefulness, and ease of use.The study also investigates how these factors relate to attitudes and behavioural intention.Source: Introduction · p. 2
• The paper reviews AI and trust, presents TAM, then describes its methodology and sample before reporting model fit, hypothesis tests, and comparisons with prior research.Source: Introduction · p. 2

1. Literature review

• The review situates retail AI amid pandemic-related acceleration of online purchasing and increased demand for AI in retail.Source: §1 · p. 2

1.1. The use of AI in online shopping

• AI can personalise product selection, automate assistance, and reduce search costs, but effective systems rely on big data and require substantial investment.Examples include chatbots, recommendations, and automated analysis of consumer behaviour.Source: §1.1 · p. 3

1.2. The role of trust in online shopping

• Online trust relates to the general e-commerce environment and to a consumer’s experience in a particular store; higher e-commerce trust is associated with more frequent online shopping.The review also notes that shoppers may not notice when a website uses AI.Source: §1.2 · pp. 3–4

1.3. Technology Acceptance Model (TAM)

• TAM links perceived ease of use and usefulness with attitudes and behavioural intention, but prior criticism questions whether consumers can choose whether to use AI in a webshop.The authors argue consumers retain the choice to shop in an AI-powered webshop or not.Source: §1.3 · pp. 4–5

2.1. Methodology

• The authors adapted measurement instruments from Park (2009) and Hu and O’Brien (2016), modifying each variable to measure AI’s perceived role in online shopping.Source: §2.1 · p. 5
• The 26-variable questionnaire also collected six demographic variables, with the measurement instruments presented in Table no. 1.Source: §2.1 · pp. 5–6
• Respondents were told that AI-powered webshops offer personalised recommendations, chatbots, image search, virtual changing rooms, voice recognition, and other shopping tools.The description also covered recommendations based on prior searches and purchases, and reminders about items left in a cart.Source: §2.1 · p. 5
• The questionnaire operationalised perceived usefulness and ease of use as beliefs about shopping effectiveness and effort, respectively.It also included measures of consumers’ experience and beliefs about whether AI works in their best interest.Source: §2.1 · p. 6

Trust (TRUST)

• The measures cover trust, usefulness, ease of use, experience, social norms, attitudes, behavioural intention, shopping enjoyment, and perceived product quality.The questionnaire includes items about finding better deals, saving time, ease of learning, and willingness to use AI-powered shops more frequently.Source: Trust (TRUST) · pp. 6–7
• The survey required prior experience with AI-powered webshops, used convenience sampling in Hungary in July and August 2020, and tested the model with SEM.The 439-response database was complete and had no missing data.Source: Trust (TRUST) · p. 7

2.2. Research questions and hypotheses

• The research asks whether TAM can investigate AI-shopping acceptance and which factors influence intention to visit AI-powered webshops and apps.The initial model represented hypothesised relationships among constructs with directed arrows.Source: §2.2 · pp. 7–8

2.3. The sample

• The 439 respondents averaged 32.2 years of age, and the sample size fell within the study’s stated ideal range of 380 to 760 for its initial model.The authors note that the sample’s representativeness of Hungary’s e-shopper population could not be assessed because population-distribution data were unavailable.Source: §2.3 · p. 8

3. Results and discussion

• Respecification followed rejection of the initial model: PCA removed variables with factor loadings below 0.7, yielding a nested model with eight hypotheses.The retained measures covered trust, perceived usefulness, ease of use, attitude, and behavioural intention.Source: §3 · pp. 8–9

3.1. Validity

• Factor loadings exceeded .84, and AVE scores were .79 for ATT, .76 for PU, and .81 for PEU; reported construct correlations were all below .85, supporting convergent and discriminant validity.Source: §3.1 · pp. 10–11

3.2. Reliability

• Cronbach’s alphas were at least .87, AVE scores exceeded .76, and composite reliability scores exceeded 0.9, supporting measurement-model reliability.Source: §3.2 · p. 11

3.3. Model fit

• The respecified structural model showed good fit, including CFI=0.99, RMSEA=0.02, and HOELTER (CN)=546, supporting TAM’s suitability for this analysis.The initial model was rejected, and theoretically justifiable alternatives were tested on the same data.Source: §3 · pp. 8–9, §3.3 · p. 12

3.4. Hypothesis testing and estimates

• The perceived-ease-of-use-to-attitude path was not significant, the only rejected hypothesis; the authors suggest AI webshop shopping may not require much mental effort.The other hypotheses were accepted.Source: §3.4 · p. 13
• Perceived ease of use had strong indirect effects on attitude (0.48) and intention (0.43), and trust indirectly affected intention (0.42).Ease of use also positively affected perceived usefulness and trust directly.Source: §3.4 · pp. 13–14
• Perceived usefulness had the highest total effect on behavioural intention; trust most strongly affected usefulness, and attitude strongly and positively affected intention.Trust also directly influenced attitude.Source: §3.4 · pp. 13–14

• Trust, perceived usefulness, perceived ease of use, and attitudes were confirmed as key factors influencing behavioural intention to use AI-powered webshops and apps.Source: Conclusions · p. 14
• The direct relationship between perceived ease of use and attitudes was insignificant, although ease of use indirectly affects attitudes and behavioural intention.Source: Conclusions · p. 14
• Trust is central: consumers who do not trust an AI-powered webshop tend to consider it less useful and form a negative attitude toward it.Source: Conclusions · p. 14
• Favourable attitudes toward AI-powered webshops lead to more frequent online traffic, while the authors expect AI use in online shopping to expand further.They connect that expectation to the positive effect of the recent COVID-19 crisis on e-commerce.Source: Conclusions · p. 15
• The authors recommend repeating the study in a multicultural context as a future research direction.Source: Conclusions · p. 15

• The reference list includes sources on structural equation modelling and model-fit assessment.Source: 170 Amfiteatru Economic · pp. 17–18
• The references include work on technology acceptance and online-shopping acceptance, including TAM and consumer e-shopping.Source: 170 Amfiteatru Economic · pp. 17–19
• Cited topics also include AI in retail and marketing, e-commerce trust, and online-shopping behaviour.Source: 170 Amfiteatru Economic · pp. 17–19