A new study published in Frontiers in Human Dynamics by Katharina Frehmann and Marc Ziegele of Heinrich Heine University Düsseldorf tackles this question from an unusual angle: the users’ own perspective. Rather than presupposing that trust in voice assistants is technical, interpersonal, or media-based, the researchers asked what trust actually means to the people who live with these devices every day. The answer, based on 22 in-home interviews conducted in summer 2024, is neither simple nor uniform. Participants drew on interpersonal concepts when defining trust in general, describing it as a willingness to relinquish control and the ability to rely on someone without needing to intervene or verify. But when asked whether this definition applied to a voice assistant, many hesitated. Some rejected the term outright for a device, while others restricted trust to particular tasks and functions, such as playing the songs they wanted or triggering an alarm clock reliably.

This pattern, which the researchers call task-specific trust, sits at the heart of their findings. Participants commonly described selective reliance grounded in bounded trust judgments concerning distinct functions rather than holistic trust in the assistant as a whole. One participant explained that trust was less fitting than the term reliable, meaning the assistant always works. Another stated that they trusted their voice assistant to play the songs they wanted to a large extent, while simultaneously expressing doubts about whether the device was listening in on conversations. These coexisting judgments show that trust in one area and a lack of trust in another could exist side by side without canceling each other out. For some participants, the actual trustee was not the assistant at all but the parent company behind it, with privacy concerns and institutional trust in organizations like Amazon shaping their evaluations separately from the device’s performance.

The study employed a multi-method qualitative design that combined semi-structured interviews with observational tasks involving participants’ own assistants and ChatGPT’s voice mode. Interviews lasted between 70 and 110 minutes and took place in participants’ homes to enhance external validity. The sample included 22 users of different ages, genders, and products, with 16 identifying Alexa as their primary assistant, 3 using Siri, and 3 using Google Assistant. Participants ranged from 18 to over 70 years old, and most were familiar with ChatGPT in text-based form before the study. During the observational phase, participants first performed a typical task of their choice with their own assistant and then completed six predefined tasks modeled after common use cases, covering technical, interpersonal, and media-related dimensions. They then repeated the same tasks with ChatGPT’s voice mode on an iPhone 15 running the iOS app with GPT-4 selected, keeping the model and default voice constant across sessions.

When it came to trustworthiness criteria, technical and media-related qualities were especially prominent in participants’ accounts. Reliability and functionality provided common evaluative references for technical trustworthiness, with some participants explicitly linking knowing that a function worked to trusting that it would work. One participant noted that if an important reminder function failed, they would no longer trust the device. Media-related criteria included accuracy, source transparency, relevance, and selectivity of information. Participants appreciated when assistants named the source of their answers, and one described trust precisely as accepting information without double-checking it, having stopped verifying sources after initially configuring them during setup. However, verification varied with the stakes: the same participant who trusted weather information without rechecking it consulted additional sources for medication side effects. Relevance also mattered, as one participant gradually stopped using their assistant for information searches after repeatedly receiving answers about entirely different topics.

Interpersonal qualities, by contrast, did not necessarily translate into interpersonal trust. Although participants sometimes reacted socially or emotionally to their assistants, using polite phrases or acknowledging that the technology was good at building up something emotional, they frequently reflected on these reactions and continued to describe the devices in functional terms. One participant, identifying as a technician, explained that they lost a bit of trust because the device was pretending to be someone, calling the assistant’s human-like presentation potentially deceptive. Another valued Siri’s integrity in being willing to acknowledge its limitations, contrasting this with ChatGPT, which did not do so. These accounts suggest that users have developed more differentiated interaction scripts than early research on the computers-as-social-actors paradigm might predict: they respond spontaneously to human-like voices while simultaneously reflecting on the technical nature and limitations of the interaction partner, and they restrict the situations in which they are willing to rely on the assistant.

The relationship between trust, experience, and use emerged as a recursive calibration loop rather than a one-directional path. Participants described an initial period of experimentation in which they tried different functions and learned how to phrase requests usefully. Based on these experiences, they adjusted their expectations and communicative behavior, continuing to use functions perceived as reliable while avoiding those that appeared error-prone. One participant stopped using a function after repeated failures, explaining that the new expectation was that it would probably not work. Others anticipated greater trust and broader use if reliability improved. Reliable functions that fitted recurring situations became ordinary practices or supported new routines, such as maintaining a shared shopping list, an evening ritual with lights and music, or using the assistant for simultaneous kitchen timers while abandoning separate devices. This selective habituation did not require general trust in the assistant, and negative experiences with one function did not generally lead participants to abandon the device entirely.

The comparison between familiar assistants and ChatGPT’s voice mode revealed perhaps the most counterintuitive finding of the study. ChatGPT was frequently evaluated as more capable, informative, and socially convincing, with participants appreciating the breadth and detail of its responses and its ability to address complex questions. Some associated its detailed answers and human-like voice with the impression that someone had looked up information for them and was responding personally. Yet these favorable evaluations did not uniformly translate into greater trust. Many participants trusted their own assistants more because they were familiar, predictable within an established range of uses, and easier to handle. One participant stated intuitively that they trusted Alexa more than ChatGPT because they had become accustomed to it, even though ChatGPT’s technical performance was almost a little better. Perceived interpretability also played a role: participants described traditional assistants as directing them to recognizable sources or following a comprehensible sequence between command and outcome, while ChatGPT’s generated responses could not readily be traced back to identifiable origins. Greater capability could therefore increase perceived usefulness while simultaneously increasing the amount of trust required when participants felt less able to control and inspect the result.

Where participants did trust ChatGPT more, that trust was also task-specific and experience-based, restricted to complex information tasks for which they considered it especially suitable. The researchers propose a preliminary conceptual framework of task-specific hybrid trust that distinguishes among technical, interpersonal, and media-related trustworthiness evaluations, bounded trust judgments concerning distinct or overlapping roles, and associated use behaviors, emotions, and attitudes. These judgments are embedded in a temporal calibration process involving experience, expectations, use, and trust, while institutional trust in the parent company and perceived privacy practices form contextual frames discussed separately from the assistant’s role-specific performance. The framework is empirically informed rather than a validated causal model, and the authors emphasize that future research should test and refine these relationships through quantitative, comparative, and longitudinal designs.

The study’s limitations include its focus on German users, whose historically distinct approach to privacy and self-disclosure may have influenced limited trust in assistants and stronger reactance toward anthropomorphic elements. The exploratory qualitative nature of the research also means the findings are not representative of all users, and the combination of retrospective accounts with a single session of elicited tasks cannot establish causal directions. Nevertheless, the implications are significant for research on human-machine communication and for the design of conversational AI. Using a general evaluation of trust to explain usage intentions may be too broad, since trust and a lack of trust can coexist when they relate to different functions or roles. An aggregate trust score may conceal considerable variation across functions and situations, and researchers should specify the functions and roles to which their trust measures refer. As LLM-based assistants grow in capability and reach, and as their use expands into more consequential contexts such as education, understanding how users calibrate trust at the level of particular tasks rather than the assistant as a whole may prove essential for predicting adoption, safeguarding against misinformation, and refining the theoretical frameworks that explain our relationships with increasingly human-like machines.

Subject of Research: Trust in voice-based assistants: a multidimensional conceptualization based on in-home user interviews

Article Title: Trust in voice-based assistants: a multidimensional conceptualization based on in-home user interviews

Article References: Frehmann, K., & Ziegele, M. (2026). Trust in voice-based assistants: a multidimensional conceptualization based on in-home user interviews. Frontiers in Human Dynamics, 8, Article 1941118. https://doi.org/10.3389/fhumd.2026.1941118

Image Credits: AI Generated

DOI: 10.3389/fhumd.2026.1941118

Keywords: voice assistants, conversational AI, trust, ChatGPT voice mode, human-machine communication, media trust, Alexa, Siri, voice-based, assistants, multidimensional, conceptualization

The study’s central insight is methodological as much as theoretical. Rather than asking people to rate one AI article against one human article in a laboratory setting, Barnidge and Nah conducted a two-wave online survey of 942 U.S. adult internet users and then applied a statistical technique known as bifactor modeling. This approach allows researchers to separate the variance that people’s credibility judgments share across all types of news sources from the variance that is unique to each specific source. In practical terms, the model can distinguish between a person’s general orientation toward news credibility and their particular skepticism or trust toward AI-generated news, professional journalism, or citizen-produced news. The bifactor model fit the data well, with a comparative fit index of 0.94, a Tucker-Lewis index of 0.94, a root mean square error of approximation of 0.058, and a standardized root mean square residual of 0.045, all of which indicate an acceptable fit by conventional psychometric standards.

Why does this decomposition matter? Because, as the authors show, the competing explanations for the AI-journalist credibility gap operate at different levels of the judgment process. When the researchers examined what predicted people’s generalized credibility orientations, the strongest associations were communicative factors: how much exposure people had to AI-related news and how much they discussed such news with others. In other words, people who encounter and talk about stories concerning artificial intelligence develop broad dispositions about whether news in general can be believed. These everyday communicative habits, the study suggests, shape a kind of baseline skepticism or trust that colors judgments of all news sources, whether the byline belongs to a person or an algorithm.

When it came to the specific difference between AI-generated and professional news, however, a different predictor dominated. Trust in journalists was the strongest predictor of the source-specific gap between machine-written and professionally produced news, with a standardized coefficient of 0.25, statistically significant at the p < 0.001 level. This is a striking result. It means that the more people trust journalists as a professional group, the more they differentiate in favor of professional journalism when comparing it against AI-generated alternatives. The gap, in this sense, is not primarily driven by attitudes toward artificial intelligence at all. It is driven by the institutional authority that audiences attach to journalism as a profession, an authority that machine authorship simply does not carry.

To probe this further, the researchers decomposed the professional-AI credibility gap into its component parts, comparing it against the gap between professional and citizen-produced news. The result was revealing: slightly more than half of the professional-AI gap corresponded to the professional-citizen component. This suggests that audiences perceive AI-generated news in a way that is partly analogous to citizen journalism, as news produced outside the boundaries of the professional field. The AI-journalist credibility gap, therefore, reflects two distinct things at once: the simple fact of human versus non-human authorship, and the institutional authority, the norms, routines, and legitimating apparatus, that professional journalism has accumulated over more than a century. AI-generated news suffers not merely because a machine wrote it, but because it lacks the institutional stamp that professional newsrooms confer.

These findings arrive at a moment when the use of AI in news production is both widespread and largely invisible to audiences. Recent research cited in the study, including a 2025 analysis of American newspapers, found that AI use in newsrooms is common, unevenly distributed, and rarely disclosed to readers. Meanwhile, a growing body of experimental work has documented how machine attribution affects perception. Studies dating back to Clerwall’s 2014 experiment on robot journalism, and extended by meta-analyses of machine authorship effects, have shown that readers often rate automated content as more objective but less credible, less trustworthy, and less able to convey expertise or nuance. Other work has invoked the machine heuristic, the cognitive shortcut by which people evaluate machine-produced content differently from human-produced content, and has even found that machine-written political news can intensify hostile media perceptions among partisans.

What the new study adds to this literature is a structural explanation. Previous experiments established that the gap exists; Barnidge and Nah explain where the gap comes from in the architecture of audience judgment. By separating generalized from source-specific variance, the study shows that the credibility deficit of AI news is not simply a matter of novelty effects or technophobia that might fade as AI becomes familiar. Communicative factors, including AI-related exposure and discussion, do shape generalized credibility orientations, but the specific preference for professional over machine-written news is anchored in trust in journalists themselves. That anchoring is likely to be durable, because it rests on institutional legitimacy rather than on fleeting attitudes toward a new technology.

The theoretical implications extend beyond journalism. The authors argue that their findings refine credibility theory by demonstrating that audience evaluations of emerging media are tied not only to attitudes toward new technologies but also to the institutional authority audiences attach to established media institutions. This reframing connects the study to a broader scholarly conversation about metajournalistic discourse and boundary work, the processes by which journalism defines itself and legitimizes its authority in the digital era. It also echoes earlier research on citizen journalism, which found that social trust and perceptions of journalistic roles shape how audiences evaluate non-professional news sources. AI-generated news, in this framework, occupies a position of peripherality in the journalistic field, much like citizen journalists and other new actors that have challenged the boundaries of professional practice.

For the news industry, the practical stakes are considerable. As publishers increasingly deploy large language models to draft routine stories, summarize events, and personalize content, they face a credibility problem that disclosure alone may not solve. If audiences’ preference for professional news stems from institutional trust rather than from ignorance of AI’s capabilities, then simply telling readers that a story was AI-assisted will not close the gap. Conversely, the finding that slightly more than half of the gap corresponds to the professional-citizen component hints that hybrid forms of human-AI collaboration, in which professional journalists supervise and validate machine output, might recapture some of the institutional authority that pure automation lacks. Recent work on visualizing human-AI collaboration disclosures in news production suggests that newsrooms are already experimenting with how to communicate such arrangements to audiences.

The study is not without limitations inherent to its design. It relies on self-reported survey data from U.S. adult internet users, and generalized credibility judgments measured in a survey may not perfectly capture how people react to specific articles in the wild. The bifactor approach, while powerful, requires careful interpretation, and the authors note that regression among factor scores involves well-known statistical subtleties. Still, the two-wave design and the strong model fit lend considerable weight to the conclusions. And the core message is one that both technologists and journalists should heed: the battle for trust in AI-generated news will not be won or lost on the merits of the technology itself. It will be won or lost in the contested terrain of institutional authority, where audiences decide whether the institutions behind the news, human or machine, deserve their belief. As AI-generated content becomes indistinguishable from professional reporting in style and accuracy, that institutional dimension may prove to be the last, and most decisive, frontier of the credibility debate.

Subject of Research: Audience credibility judgments of AI-generated versus professional and citizen-produced news

Article Title: Explaining the AI–journalist credibility gap: generalized and source-specific judgments of news credibility

Article References: Barnidge, M., & Nah, S. (2026). Explaining the AI–journalist credibility gap: generalized and source-specific judgments of news credibility. AI &amp; SOCIETY. https://doi.org/10.1007/s00146-026-03374-2

Image Credits: AI Generated

DOI: 10.1007/s00146-026-03374-2

Keywords: AI-generated news, news credibility, journalism, bifactor modeling, trust in journalists, automated journalism, citizen journalism, machine heuristic, institutional authority, media trust, survey research, AI & Society