User experience and early adoption of an artificial intelligence application: a mixed-methods sentiment analysis of DeepSeek reviews on Google Play
- 20 October 2026
- journal article
- Published by Emerald Publishing in Information Discovery and Delivery
- p. 1-10
Abstract
Purpose: The rapid emergence of artificial intelligence (AI) chatbot applications has transformed digital information access, yet the theoretical understanding of how users evaluate new AI tools during early adoption remains limited. Grounded in the Technology Acceptance Model and Expectation Confirmation Theory, the study aims to address a conceptual gap regarding mixed sentiment and explore how demographic factors, such as gender, shape user experience during early adoption. Design/methodology/approach: A mixed-methods approach was used to analyze 1,000 Google Play Store reviews. Reviews were categorized by sentiment: positive, negative, mixed (both positive and negative statements) or neutral. Gender predictions were also made from usernames to explore demographic differences in sentiment. Findings: Of the 1,000 reviews, 487 were mixed reviews (48.7%), 373 were negative reviews (37.3%), 139 were positive reviews (13.9%) and one was neutral (0.1%), indicating that mixed and negative sentiment predominated. Users were mostly satisfied with DeepSeek’s usability and output quality, but criticized its performance and stability issues. Female users gave more negative reviews than male users (p < 0.05). Research limitations/implications: The analysis was conducted on a stratified sample of 1,000 “most relevant” Google Play reviews from a short early adoption period, which may not accurately represent later user experiences or experiences on other platforms. Only reviews from phone users were included, and the neutral sentiment category contained too few cases for comparison. Gender inference was based on usernames and reflects predicted rather than self-identified gender. Therefore, the results should be interpreted as indicative of early user perceptions rather than long-term attitudes. Practical implications: The findings indicate that infrastructure reliability is the primary determinant of early user satisfaction. Transparent communication about service limitations is recommended prior to large-scale deployment. For policymakers, the equity implications of performance-unstable free AI services warrant attention, particularly for users with limited access to alternatives. Originality/value: The mixed-methods analysis of early DeepSeek user reviews, which distinguishes between mixed sentiment and positive and negative feedback, and incorporates a gender-based comparison, provides timely insight into user expectations and challenges during early adoption, supporting the development of more user-centered and stable AI systems.Keywords
This publication has 33 references indexed in Scilit:
- Automatic gender detection: a methodological procedure and recommendations to computationally infer the gender from names with ChatGPT and gender APIsScientometrics, 2024
- Assessing the impact of artificial intelligence on customer performance: a quantitative study using partial least squares methodologyData Science and Management, 2024
- Sentiment analysis on google play store app users’ reviews based on deep learning approachMultimedia Tools and Applications, 2024
- ChatGPT: A comprehensive review on background, applications, key challenges, bias, ethics, limitations and future scopeInternet of Things and Cyber-Physical Systems, 2023
- Time to Publication in High-Impact General Medical Journals Differs Between Female and Male ResearchersJournal of General Internal Medicine, 2022
- Sentiment Analysis Covid-19 Spread Tracing on Google Play Store Application2022
- Gender-sensitive sentiment analysis for estimating the emotional climate in online teacher educationLearning Environments Research, 2022
- Standardised PaO2/FiO2 ratio in COVID-19: Added value or risky assumptions?European Journal of Internal Medicine, 2021
- Sentiment Analysis of COVID-19 tweets by Deep Learning Classifiers—A study to show how popularity is affecting accuracy in social mediaApplied Soft Computing, 2020
- Gender and attitudes toward technology use: A meta-analysisComputers & Education, 2017