The rapid expansion of artificial intelligence (AI) in retail has positioned chatbots as a central customer-brand interface in e‑commerce (Shankar, 2018; McKinsey & Company, 2024). While chatbots are expected to streamline service, reduce operational costs, and enhance availability (Dey & Bhaumik, 2022), their real impact on customer experience and brand perception remains insufficiently understood. Existing research largely focuses on system-level acceptance or perceived usefulness, with limited empirical evidence capturing naturally occurring customer reactions expressed in online reviews (Mariani & Borghi, 2021). Addressing this gap, this research investigates how consumers evaluate chatbots in real service encounters and how these evaluations shape perceptions of brand competence, warmth, empathy, and overall satisfaction, dimensions central to brand meaning (Fiske et al., 2007; Keller, 2013). This work analyses user-generated reviews to examine emotional, functional, and experiential responses to chatbots. Studies rely on a dataset of approximately 610,000 Trustpilot reviews from seven global retail brands (Adidas, ASOS, Nike, Sephora, Shein, TeePublic, and Zara), covering the period 2020–2025. Using natural language processing (NLP) methods, including sentiment analysis with the RoBERTa model (Barbieri et al., 2021), the study identifies patterns in customer evaluations and quantifies how chatbot-related experiences differ from other types of brand feedback. Results show that only 3.7% of all reviews explicitly mention chatbots, yet these reviews are disproportionately negative (Adam et al., 2021): across brands, 66% to 85% of chatbot-related comments express dissatisfaction. Sentiment has deteriorated sharply over time, with negative evaluations rising from 25% in early 2020 to 85% by 2024–2025. Negative reviews are longer and more detailed, frequently revolving around unresolved refunds, order issues, or conversational failures. Simpler rule‑based chatbots receive comparatively more positive evaluations, whereas AI‑driven and hybrid models generate deeper frustration, especially when personalisation is poorly implemented or when the chatbot cannot resolve high-stakes issues (Juquelier et al., 2025). The study also examines how chatbot interactions influence broader brand perceptions by applying thematic dictionaries for competence, empathy, warmth, and satisfaction. Linear regression models using star ratings as the dependent variable show that these four dimensions significantly predict satisfaction but differ in magnitude: competence and satisfaction are the strongest contributors, while warmth and empathy emerge less frequently in chatbot-related comments. Importantly, comments mentioning chatbots display lower levels of all four dimensions compared with other reviews, revealing a negative spillover effect on brand image. When chatbots fail, users attribute not only poor performance to the system but also reduced competence and warmth to the brand itself. Findings underscore that chatbots currently compromise customer experience in retail rather than enhance it. Technical sophistication alone does not guarantee effectiveness; Users consistently value clarity, problem resolution, and efficiency over advanced conversational features. The research contributes to AI‑service and digital retailing literature by offering large‑scale evidence of how chatbot interactions shape emotional and functional evaluations, and by demonstrating the strong influence of chatbot performance on brand perception.

--

Branding,Chatbots,Customer Experience,Brand Perception,AI in Retail