Abstract
This article reviews investigations of online survey chatbots across disciplines and offers insights into their potential to encourage innovative and practical approaches to marketing research surveys. Online surveys deployed in service practice can effectively capture customer feedback. Using chatbots to administer such surveys can have desirable outcomes, although specific advantages and disadvantages are not clear at this point. Based on a meta-analysis of prior research on the use of chatbots in online surveys, this article presents a typology of relevant themes and avenues for further research. The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) method initially yielded 1,827 unique papers from Scopus, of which 30 qualify for in-depth analysis. The analysis identifies five key survey administration capabilities of chatbots: data collection, conversational interfaces, interactivity, humanization, and effective response generation. It also highlights the challenges and research gaps associated with each capability.
Introduction
Market transformations that are driven by artificial intelligence (AI) include the expanded use of chatbots, which are computer programs that replicate human communication in text- or voice-based dialogue systems (Gaikwad et al., 2018). The global chatbot market continues to grow, with an expected value of US$27.29 billion by 2030 (Tyumeneva, 2026). Chatbots can handle and promptly respond to simple queries in call centers, while human employees handle more complex issues, thus promising substantial reductions in labor costs. It is reported that Chatbots can respond quickly and answer an estimated 80% of frequently asked questions (Downie & Hayes, 2025).
Chatbots are also promising for online surveys (Jin et al., 2024) and may replace other, less efficient survey methods, such as in-person or telephone surveys. Chatbot surveys are continuously available (Araujo, 2020). In this sense, chatbots offer enhanced convenience (Kim et al., 2019) and efficiency (Belhaj et al., 2021). Current implementation of chatbots can engage in natural, conversational interactions with survey participants (Celino & Re Calegari, 2020), as exemplified by surveys embedded in Facebook Messenger, which encourage participation by appearing in a familiar setting for many consumers (Smutny & Schreiberova, 2020).
Extant research has examined the use of chatbots across various domains, including education (Belhaj et al., 2021), computer science (Jin et al., 2024), and health care (Te Pas et al., 2020). We conduct a literature review, using the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) approach (Page et al., 2021). We review the application of chatbot surveys across these domains, focusing on applications relevant to market research. Marketing researchers can draw from established knowledge in other domains. By identifying 1,827 unique papers and further analyzing 30 highly relevant papers, we established five main themes and noted several related but less prominent themes.
Academically, the findings summarize the literature on the use of chatbots in market research surveys and outline an agenda for future research. While previous studies have examined chatbots from different perspectives, the literature review in this paper provides a more comprehensive overview of relevant themes than existing work. For instance, Dwivedi et al. (2021) did not investigate how chatbots can be implemented in specific contexts, such as online surveys, whereas Höhne et al. (2024) primarily examined the adoption of chatbots and their effects on outcomes (i.e., completion rates). Building on these studies, we contribute to the literature by providing a systematic synthesis of chatbot-administered surveys, offering structured insights into their design and implementation in online survey research. For marketers, this paper highlights the potential and limitations of chatbots as data collection tools and provides guidance on their implementation. The discussion section elaborates on the study’s implications for academic literature and marketing research.
The paper is structured as follows: first, we present the theoretical background; next, we describe how we applied PRISMA methodology to select papers, detailing how articles were identified, screened, deemed eligible, and included. Lastly, we apply thematic content analysis to evaluate the 30 key articles, identifying five main themes and five additional emerging themes. The analysis of primary themes underscores the benefits and limitations of chatbot capability dimensions and provides directions for both academic research and practice, as outlined in the discussion.
Theoretical Background: Chatbots
Contemporary chatbots can engage in humanlike dialogue and perform natural language processing (NLP). They can be integrated into various application programming interfaces (Tsai et al., 2021) to resolve customer requests and automate repetitive tasks, such as answering frequently asked questions or sending customer emails. When applied to online surveys, chatbots can also improve efficiency (Belhaj et al., 2021) by gathering vast amounts of data and streamlining the survey process, such as by explaining the purpose of the questionnaire and guiding participants as needed (Söderström et al., 2021). Furthermore, whereas conventional online surveys tend to use a formal language style, the casual language style often programmed into chatbots can resonate with participants’ expectations of social interactions, though it may also engender embarrassment (Kim et al., 2019).
Noting the many studies on the use of online chatbots, we also identify a gap: there is no clear overview of the current state of knowledge. Literature can provide insights into current knowledge of chatbot use in market research surveys, highlight promising avenues for future research, and offer guidelines for practitioners applying chatbots in market research. Based on this review, we address three focal research questions that drive our investigation:
What does current literature say about chatbots’ online survey capabilities?
What are the key benefits and limitations of chatbot implementation in online surveys, as identified by existing research?
What directions for research can be derived from the extant academic literature to guide ongoing chatbot implementation in online marketing surveys?
We undertake a systematic literature review to assess what is known, particularly the key advantages and limitations of chatbots, and to propose directions for further research.
PRISMA-Guided Selection of Relevant Papers
We applied the PRISMA framework, a structured and reproducible search strategy (Page et al., 2021), comprising identification, screening, eligibility, and inclusion. We searched across disciplines beyond marketing, including, for example computer science and information systems, for studies on chatbots in online surveys. By rigorously adhering to predefined criteria, we captured all relevant studies. Figure 1 depicts the step-by-step process for screening and excluding articles. Search results for PRISMA
Keyword Combinations
In Step 2 (Screening), we applied filters to exclude topics that are not relevant for the use of chatbots in surveys, such as chatbots’ social presence in communication (Tsai et al., 2021). Journal articles were limited to 2013-2025, as we identified Hasler et al.’s (2013) as the earliest relevant study. Conference papers were limited to 2019-2025 to anticipate emerging research. Titles, abstracts, and reference lists were checked for reliability and credibility. Three screening exclusion criteria (SEC) were applied. Therefore, 603 articles remained for eligibility assessment.
In Step 3 (Eligibility), we further narrowed the selection based on journal credibility and empirical evidence, following Patino and Ferreira (2018) and Meline (2006). Two eligibility exclusion criteria (EEC) were applied, and 30 remained after this step.
Step 4 (Inclusion) ensured robustness by evaluating the 30 articles for relevance, focus on chatbots in online surveys, and methodological rigor. No articles were removed, ensuring our findings are based on high-quality, current, and thematically relevant sources. The Online Appendix summarizes these articles, including publication source (journal or conference), applied theories, research focus, survey approach, methodology, chatbot capabilities, and key findings.
The descriptive statistics reported in Figure 2 illustrate the publication year of the 30 articles, revealing increased research interest since 2019. Figure 3 shows an increased interest in the application of chatbots in surveys, in terms of the number of times each of the 30 selected articles is cited each year. Table 2 lists the number of cited articles by journal or conference, and Computers in Human Behavior publishes the largest number of papers on the application of chatbots in surveys. Publication year Total citations per year Cited Articles by Journals/Conferences
Thematic Content Analysis
We conducted a thematic analysis of the 30 articles in Table 2 to identify themes and trends across studies. Using the method outlined by Cruzes and Dyba (2011), we began with an exploratory text analysis to identify common themes. This iterative process (Neuendorf, 2018) allowed us to refine themes iteratively and assess each study’s contribution, providing insights into the scope, depth, and breadth of the research. The first author analyzed all papers, while the other two authors derived themes from half of the papers each. Such qualitative evidence synthesis enhances the reliability of the findings.
Five Key Themes and Related Studies
Theme 1: Data Collection Capability
A growing body of literature examines the feasibility of using chatbots embedded in surveys to collect data, replacing traditional open- or closed-ended questions with conversational chatbot-based interview or questionnaire prompts (e.g., Kim et al., 2019; Xiao et al., 2020a; Zarouali et al., 2024). They show that chatbot-assisted data collection can enhance participant disclosure (Beam, 2023; Jin et al., 2024; Lucas et al., 2014; Te Pas et al., 2020; Zhou et al., 2019) and improve response quality compared with conventional online survey formats (Beam, 2023; Jin et al., 2024; Kim et al., 2019; Sidaoui et al., 2020). Positive outcomes are more frequently found when chatbots are used to administer structured or semi-structured questions. Under such conditions, chatbots are hypothesised to reduce satisficing behaviour and response biases by encouraging greater engagement and attentiveness (Beam, 2023; Hassina & Hiba, 2025).
These benefits are not consistently replicated across studies. Notably, using chatbots to collect large-scale survey data can be challenging, due to the inherent inaccuracies of NLP technologies. Therefore, chatbots may misinterpret respondents’ inputs or generate inappropriate follow-up responses, leading to dissatisfaction (Jacobsen et al., 2025; Kim et al., 2019; Li et al., 2025) and conversational disengagement (Hasler et al., 2013; Jin et al., 2024; Tang et al., 2025). These issues might ultimately diminish respondents’ willingness to answer subsequent questions, thereby compromising response quality, and challenge the claim that chatbots can reduce respondent satisficing behaviour.
This contradiction seems to be driven by differences in chatbot design (Jin et al., 2024) and interaction quality (Celino & Re Calegari, 2020; Xiao et al., 2020a). Some studies report that chatbots with clear conversational structures (Jin et al., 2024; Ward & Pond, 2015), limited or repeated probing (Li et al., 2025; Park et al., 2019; Zarouali et al., 2024), and constrained response formats (Jacobsen et al., 2025; Jin et al., 2024) function better and enhance interaction quality. This reduces cognitive load and minimizes misinterpretation. In contrast, when chatbots are required to interpret diverse open-ended responses or to dynamically adapt to respondents, interaction breakdowns become more likely (Park et al., 2019; Xiao et al., 2019). Respondents’ unfamiliarity with conversational survey formats (Lucas et al., 2014; Tang et al., 2025; Zarouali et al., 2024) can increase response times and this may result in shorter or less detailed answers (Zarouali et al., 2024). Similarly, if a chatbot continues to probe the same topic while ignoring cues to shift direction, it can generate frustration (Park et al., 2019), disrupt conversational flow, and ultimately lead to disengagement (Zarouali et al., 2024).
These findings suggest that the effectiveness of chatbots for data collection in online surveys depends heavily on their design and deployment. Chatbots seem more effective for structured data collection tasks, whereas their application to complex, open-ended survey contexts remain prone to interactional and technological limitations.
Based on this theme, we propose three research avenues. First, future research could explore the design of structured conversations in chatbot surveys to provide clear guidance and a sense of dialogue (Araujo, 2020; Jin et al., 2024; Ward & Pond, 2015). Researchers should draw attention to minimal or repeated probing in chatbot surveys because it can clarify questions or responses but may also cause fatigue when overused (Li et al., 2025; Park et al., 2019; Zarouali et al., 2024). Constrained responses may minimize ambiguity and misinterpretation (Jacobsen et al., 2025; Jin et al., 2024), whereas diverse open-ended responses are crucial for obtaining richer insights (Xiao et al., 2020a, 2020b; Zhou et al., 2019). Second, future studies could investigate how to predict and address interaction breakdowns when chatbots misunderstand inputs (Park et al., 2019; Xiao et al., 2019). Third, respondents’ unfamiliarity with conversational survey formats affects their expectations and subsequent behaviours (Lucas et al., 2014; Tang et al., 2025; Zarouali et al., 2024).
Theme 2: Conversational Interface Capability
By interacting via natural language in text-based or voice-based formats, chatbots are designed to simulate human dialogue and provide a more natural, engaging interaction experience for respondents (Celino & Re Calegari, 2020). A conversational survey aims to make participants feel as though they are engaging with a person rather than simply completing a traditional questionnaire (Duka & Njeguš, 2021; Kim et al., 2019). In such settings, chatbots offer a unique advantage by combining the efficiency of digital data collection with the interactive qualities of a human interviewer (Belhaj et al., 2021; Pickard et al., 2016). Chatbots facilitate quick, cost-effective collections of both numerical and textual data. These conversational interfaces create a natural and familiar medium for user expression, improve usability, and support flexible interactions that do not require a rigid path (Xiao et al., 2020a; 2020b). Using anthropomorphic and personalizing features can help capture user attention and build trust (Rhim et al., 2022; Tang et al., 2025).
If the chatbot effectively guides participants through a conversational survey, it provides clear and easy-to-understand information. Nevertheless, chatbots sometimes fail to address specific user demands or clarify survey features (Rhim et al., 2022). At times, a specific question of a respondent may necessitate intervention from a third-party actor, such as a human survey administrator (Belhaj et al., 2021). Therefore, although conversational interfaces are promoted as flexible and engaging, their actual operation often depends on pre-scripted conversational flows that cannot accommodate the unpredictability of real conversational behaviour (Jacobsen et al., 2025).
The benefits of conversational interfaces are often evaluated in controlled interactions where participants follow anticipated conversational paths. In contrast, studies that report limitations typically examine situations in which participants engage in conversation rather than in procedural tasks (Belhaj et al., 2021). In this case, chatbots cannot adapt to respondents, which undermines the expectations they create because of their different design from traditional surveys. Features that initially increase engagement might later undermine trust-building when expectations are not met (Rhim et al., 2022). Furthermore, requiring continuous internet connectivity leads to an additional challenge. Conversational interfaces aim to create an interactive experience for respondents, but a poor network connection can disrupt the flow of conversation between respondents and chatbots. Thus, the effectiveness of conversational interfaces is shaped by design and technological limitations.
Recent advances continue to leverage NLP to create generic conversational interfaces that support more humanlike interactions. Future research could investigate how respondents engage with chatbots to distinguish their behaviors between conversational and procedural modes (Belhaj et al., 2021) and whether the design of conversational interfaces meets respondents’ expectations for responsiveness and naturalness (Rhim et al., 2022). Furthermore, future studies could solve the practical challenge of internet connectivity by exploring how human-chatbot collaboration can be maintained effectively when connectivity issues occur.
Theme 3: Interactivity Capability
Interactive information exchanges are possible in a chatbot-enabled survey, such as repeated question elicitation. Furthermore, chatbots facilitate two-way communication, they can foster dialogue while still allowing users to maintain control over the flow and depth of the conversation. This contributes to reducing survey fatigue (Hassina & Hiba, 2025) and thereby enhances response quality (Beam, 2023; Guo et al., 2021; Jacobsen et al., 2025; Kim et al., 2019). Guo et al. (2021) find that participants remain positively engaged and willing to interact for extended periods if the chatbot demonstrates strong conversational capabilities. Chatbot interactivity enhances various other interaction outcomes, including user engagement (Belhaj et al., 2021; Celino & Re Calegari, 2020), favorable attitudes (Duka & Njeguš, 2021). Such benefits encourage positive affective, cognitive, and behavioral responses.
These findings assume that respondents view the chatbot as a conversational partner and are willing to engage; however, other literature highlights the negative effects on participant engagement in chatbot surveys. As defined by Dwyer et al. (1987), a disengaged state can occur at any point in a relationship and may lead to termination. In the context of survey chatbots, participants exhibit disengagement by expressing unfavorable attitudes toward the survey or the chatbot. This may undermine the system’s capabilities (Gupta & Chen, 2022; Li et al., 2025). For example, disengaged respondents might skip open-ended questions (Jacobsen et al., 2025; Zarouali et al., 2024), provide irrelevant or minimal answers (Xiao et al., 2020a; 2020b), or spend longer completion time than traditional surveys (Kim et al., 2019; Soni et al., 2022; Zarouali et al., 2024).
On another note, when respondents perceive the chatbot primarily as a survey tool rather than a conversational partner, interactivity may be perceived as time-consuming (Rhim et al., 2022; Zarouali et al., 2024) and cognitively demanding (Jacobsen et al., 2025), leading them to minimize participation or engagement. The effectiveness of interactivity depends on how respondents view the chatbot’s role in the process, which can either facilitate engagement or trigger disengagement.
In contemporary marketing research, personalized interactions and communication have become vital. Scholars could examine respondents’ perceptions of chatbots (as survey tools or conversational partners) and how these perceptions affect engagement and quality. Previous research indicates that respondents spend more time answering questions in chatbot surveys (Rhim et al., 2022; Zarouali et al., 2024), leading to survey fatigue or disengagement. Thus, future studies could adopt a balanced approach to interactivity and cognitive load to optimize respondents’ experience.
Theme 4: Humanization Capability
Relative to traditional online surveys, chatbots offer a more personalized approach that can enhance participants’ engagement (Conrad et al., 2015; Rhim et al., 2022; Xiao et al., 2020a, 2020b), willingness to respond, and response quality (Conrad et al., 2015; Xiao et al., 2020b). The perceived humanlike qualities of chatbots have encouraged respondents to share their thoughts and insights (Zarouali et al., 2024). In social media settings, chatbots provide interactive, personalized experiences by delivering interventions and can conduct surveys on these platforms (Kim et al., 2019; Rhim et al., 2022; Tang et al., 2025).
Anthropomorphic cues motivate greater self-disclosure among survey respondents, thereby eliciting open-ended, emotionally intense responses (Conrad et al., 2015; Hasler et al., 2013; Kim et al., 2019). Such increased self-disclosure likely stems from the chatbot’s ability to foster social bonding while preserving the respondent’s anonymity. While humanization capability encourages openness in low-sensitivity survey tasks, in sensitive survey contexts, respondents may become more cautious or even unwilling to disclose (Zhou et al., 2019).
Although a positive correlation exists between a chatbot’s verbal anthropomorphism and users’ affinity for it, this association holds only up to a point. As a chatbot’s interactions become excessively humanlike, users’ positive feelings diminish (Hasler et al., 2013; Kim et al., 2019; Zhou et al., 2019) and give way to feelings of eeriness (Rhim et al., 2022), in line with the “uncanny valley” phenomenon (Pizzi et al., 2023). That is anthropomorphism initially creates a sense of companionship, but it becomes uncomfortable when users begin to attribute a mind to the chatbot. This feeling of uncanniness diminishes the positive effects of anthropomorphism on perceived social presence (Rhim et al., 2022).
Another concern is that chatbots’ sophisticated interactive features, such as voice, gestures, and facial expressions (Tang et al., 2025), can elevate consumer expectations of the survey experience. Problems arise when there is a mismatch between these heightened expectations and the chatbot’s conversational capabilities. The more humanlike chatbots look, the more respondents expect them to perform like humans. When chatbots cannot maintain natural dialogue, offer flexible responses, or adapt appropriately to respondents’ needs, the perceived social presence might worsen rather than improve.
Instead of being inherently beneficial or problematic, the influence of humanization depends on task sensitivity (Hasler et al., 2013; Zhou et al., 2019) and the chatbot’s humanlikeness (Conrad et al., 2015; Hasler et al., 2013; Rhim et al., 2022; Tang et al., 2025; Xiao et al., 2020b) in the survey interaction. Future researchers should balance chatbot humanization, user comfort, and task sensitivity when designing surveys. For example, anthropomorphic design elements (Conrad et al., 2015; Hasler et al., 2013; Rhim et al., 2022; Tang et al., 2025; Xiao et al., 2020b), such as humanlike tone, voice, and gestures, can enhance perceived social presence and user engagement, but they can trigger eeriness when they exceed users’ expectations. A promising research direction would be to determine the optimal degree of humanization to maximize engagement without causing discomfort, while accounting for the moderating role of task sensitivity.
Theme 5: Effective Response Capability
Offering a critical capability in surveys, chatbots can generate higher response rates and superior response quality compared with traditional methods (Celino & Re Calegari, 2020; Guo et al., 2021; Hasler et al., 2013; Li et al., 2025; Soni et al., 2022; Ward & Pond, 2015; Zhou et al., 2019). In social media settings, both chatbots and online surveys help mitigate social desirability bias and thus respondents are more likely to share information when interacting with chatbots (Beam, 2023; Jacobsen et al., 2025; Kim et al., 2019; Tang et al., 2025; Xiao et al., 2019). Beam (2023) also notes that many respondents click on an online survey invitation, but very often do not start with the survey. In contrast, respondents appear more willing to spend time completing a survey if receiving an invitation from a chatbot (Celino & Re Calegari, 2020), due in part to their interactivity, which increases engagement (Abbas et al., 2021; Belhaj et al., 2021; Söderström et al., 2021; Xiao et al., 2020b).
The perceived anonymity of chatbots increases respondents’ willingness to disclose. However, the low sense of accountability may lead them to offer careless or deliberately fake answers (Li et al., 2025). Xiao et al. (2020a) show that once participants realize that a survey chatbot cannot adequately understand their responses, they may intentionally provide false information, thereby undermining response quality.
Current survey chatbots are far from flawless, and respondents may feel limited by their limited conversational abilities (Ramesh & Chawla, 2022). This may imply that participants are less willing to offer qualified answers. Contemporary NLP capabilities are not sophisticated enough to handle the nuances of human conversation (Hasler et al., 2013); they often struggle with slang, processing synonyms, and extracting entities. Although conversational surveys are designed to feel natural and engaging, limited NLP capabilities can disrupt conversational flow and frustrate respondents, ultimately reducing their willingness to provide thoughtful responses. Furthermore, a conversational approach may require more time and effort than participants initially anticipate, making it unclear whether they will complete all questions or provide in-depth responses (Ward & Pond, 2015; Zhou et al., 2019).
Five Key Themes
Other Emerging Themes
Five Emerging Themes
Chatbot Conversational Styles
Different chatbot conversational styles might affect both user perceptions and response quality. For example, a study comparing reactions to a travel booking chatbot that uses modern versus Shakespearean language styles reveals that people find the former more user-friendly and the latter more entertaining (Elsholz et al., 2019). Abbas et al. (2021) demonstrate that a casual script, filled with emojis, results in higher-quality responses to multiple-choice questions than a standard survey script. A notable gap arises in this research stream, though, in that these studies typically use a terse, non-conversational baseline for comparison, rather than a formal yet conversational style.
Gamification
Incentives (financial or otherwise) can increase response and survey completion rates and have been used extensively. Survey chatbots can apply a more varied range of motivational elements to encourage survey completion. The introduction of nudging techniques, such as gamification, might be beneficial. As detailed in behavioral psychology, nudging involves gentle, non-forceful efforts to stimulate a desired action. Te Pas et al. (2020) report a strong positive impact of gamification on survey completion rates. If adding gamification makes chatbot surveys more pleasant to complete, it could also improve the quality of the survey results.
Response Delay
Response delays offer critical social cues in chatbot surveys. Unlike human discussants, who take time to read and formulate responses, retrieval-based chatbots can respond almost instantaneously. Immediate responses feel unnatural, thereby reducing user satisfaction relative to chatbots that use dynamic delays. Rhim et al. (2022) propose effective methods for adjusting response speed and mitigating response delays by calculating response time based on response length, including the number of characters, sentence length, and average syllables per word. A dynamically timed response can enhance perceptions of social presence and trustworthiness. Users are also likely to develop subjective interpretations of a delay.
Chatbot Questionnaire Contextuality
In a chatbot survey, the nuanced richness comes from the chat data. As with face-to-face interviews, the most valuable insights often do not emerge from individual participant statements alone, but from the broader context of the questionnaire, such as patterns in response timing, follow-up answers, or repeated clarifications. Jacobsen et al. (2025) observe that participants initially struggled to articulate their thoughts about complex topics, such as personal finance habits. However, as the chatbot guided them through step-by-step probing questions, such as asking about daily routines first, then exploring underlying motivations, participants gradually provided deeper and more reflective insights. This example underscores the potential of chatbots to uncover rich, context-dependent information in survey settings.
Comfortable Information Sharing
For participants, it often is challenging to discuss sensitive or personal topics, particularly when they feel judged or stressed. Retaining respect for participants’ autonomy in user-centered design underscores the significance of the boundaries they set, which can help create a comfortable or safe environment for disclosure. The survey chatbot TigerGPT (Tang et al., 2025) maintains a relaxed, supportive, and non-judgmental tone when interacting with participants, particularly when addressing sensitive questions. This chatbot can adapt its questioning when it detects participants’ hesitation or distress, refraining from further probing unless participants indicate a willingness to continue. By fostering a comfortable survey environment, this balanced method respects participants, encourages them to share information more openly, and helps preserve their sense of psychological safety.
Discussion
AI transforms the marketing research landscape into the collection and analysis of information, the execution of strategies (e.g., ad placement), or the evaluation of their impact. We seek to establish AI applications at an initial stage, namely, collecting consumer information through chatbot surveys. A thematic analysis of chatbot capabilities for online surveys identified five key themes, i.e., data collection, conversational interface, interactivity, humanization, and effective response, and five less commonly occurring themes, i.e., conversational styles, gamification, response delay, chatbot questionnaire contextuality, and comfortable information sharing, see Tables 4 and 5. We explored the benefits and limitations of chatbots in terms of each theme. The reported findings provide valuable insights for researchers and marketers seeking to improve survey design, data or feedback collection, respondent engagement, and response quality by integrating chatbots into their online surveys.
As a theoretical contribution, this review synthesizes prior work across disciplines, resulting in an organization of chatbot survey characteristics into five key themes and five emerging themes. These themes offer systematic lenses through which future studies can examine key chatbot survey outcomes. Attributes that can be evaluated are, for example, respondents’ disclosure willingness (Beam, 2023; Jacobsen et al., 2025; Jin et al., 2024), survey completion rates (Celino & Re Calegari, 2020), and respondent engagement (Belhaj et al., 2021).
The five key themes and the five emerging themes that we report and discuss extend theoretical understanding of human-AI interaction in survey contexts. Earlier research only focused on specific aspects relating to the use of chatbots for surveys. For instance, Höhne et al. (2024) focused on Gemini Pro; this review examines the broader use of general-purpose survey chatbots in data collection, while we synchronized the evidence about humanization of chatbots for conducting surveys. Furthermore, Dwivedi et al. (2021) addressed AI-related challenges and opportunities at a macro level; the literature review reported in our paper focuses on chatbot-enabled online surveys and empirically reinforces the role of “humanness” in shaping survey effectiveness, see also Rhim et al. (2022) and Xiao et al. (2020b). Our review highlights the need to establish the optimal level of chatbot humanization that enhances engagement without eliciting discomfort (Kim et al., 2019; Zhou et al., 2019) or perceived eeriness (Rhim et al., 2022). More generally, our review provides an overview of the broader range of themes addressed in previous papers. This oversight results in well-grounded managerial implications and opens novel avenues for research on how conversational design, interactional cues, and chatbot behavioral features shape respondents’ experiences.
Advice for Market Researchers and a Research Agenda
Second, we offer recommendations for deploying chatbots, including selecting and designing chatbot attributes based on the survey task and the target audience’s characteristics. Chatbots may capture user attention and may reduce survey attrition rates (Xiao et al., 2020a; 2020b). Nevertheless, when using chatbots for market research, it is key to limit unnecessary conversational features to minimize cognitive load (Soni et al., 2022; Zhou et al., 2019) and respondent fatigue (Hassina & Hiba, 2025). Managers should evaluate tone, interactivity, response length, and levels of personalization to ensure chatbots support the intended research objectives.
Third, we emphasize several design and implementation considerations. For example, adopting humanization cues (i.e., tone, persona, and social presence) can foster trust and rapport but requires careful calibration to avoid perceived discomfort or creepiness (Rhim et al., 2022). Creepiness occurs when the chatbot recalls and uses too many personal details, or when the conversation becomes so friendly that it is only suited to friends (Pizzi et al., 2023).
Fourth, we highlight several practical challenges that marketers need to anticipate and manage, summarized in Table 4 as mitigating factors. These include respondent fatigue caused by overly long or repetitive conversations (Park et al., 2019; Zarouali et al., 2024), a lack of human intervention or interaction breakdowns (Park et al., 2019), and misunderstandings of open-ended responses. Addressing such possible issues requires conducting several pretests, refining chatbot logic, and integrating fallback solutions, such as hybrid human-chatbot survey support.
This study has several limitations that suggest avenues for future research. Research into optimizing chatbot designs across diverse user demographics, technological settings, and survey purposes is critical. Studies might explore integrating advanced AI techniques, such as sentiment analysis (Eyu et al., 2025), NLP enhancements, and personalization, to enhance the effectiveness of chatbot surveys. As Hasal et al. (2021) recommend, we note the need for research to identify options for improving chatbot security and privacy protection across two key domains: secure data transfer and secure data storage. Ongoing research must account for the rapid evolution of generative AI and its potential to increase consumer acceptance of chatbots across contexts, including surveys. By addressing these methodological and ethical considerations, researchers could unlock the full potential of chatbots as reliable, scalable, and valuable tools for data collection and analysis in both academic and industry settings.
Supplemental Material
Supplemental Material - Chatbot-Administered Surveys in Marketing: A Literature Review
Supplemental Material for Chatbot-Administered Surveys in Marketing: A Literature Review by Yingxue Zhao, Xiaoyi (Sylvia) Gao, Leo Paas in International Journal of Market Research.
Footnotes
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Supplemental Material
Supplemental material for this article is available online.
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
