Ethics enables organizations to effectively resolve dilemmas while acting socially responsible. This study aims to examine how current communication practices involving AI technologies align with domain-specific Generative AI (GenAI) guidelines to foster the quality of organizational-public relationships (OPR).
The discussion on ethical principles governing AI in computer science, along with the conceptualization of OPR linked to professional codes of ethics, informed by the feminist philosophy known as the ethics of care, contributes to the development of a proposed AI ethics framework to foster OPR. Drawing upon this framework, the use of content analysis to unpack industry discourse reveals the extent to which the industry’s understanding of GenAI ethics aligns with this proposed framework. The implications of implementing this framework to foster OPR are discussed.
Communication professionals view social responsibility and authenticity crucial for ensuring ethical AI practice, with truthfulness, respect and equity following closely. However, adherence to ethical AI use in communication depends on the implementation of explainability, accuracy, fairness and machine autonomy in computer science. Embracing the ethics of care to integrate these ethical principles into the current AI ethics framework in communication becomes crucial for easing this tension.
The proposed AI ethics framework bridges AI ethics between communication and computer science by capturing social responsibility, authenticity, truthfulness, respect and equity. This framework helps shape professional codes of ethics to address challenges in the rapidly evolving AI-driven communication landscape and advocates for the engineering of responsible AI tools to foster quality OPR. The outcomes advance cross-disciplinary research and cross-sectoral practices of AI ethics by leveraging the ethics of care, thereby connecting AI ethics across computer science and other fields.
1. Introduction
The impact of artificial intelligence (AI) on society is profound. A growing body of research explores its potential threats, challenges and opportunities from various perspectives, including education (Ballantine et al., 2024; Petricini et al., 2025; Prosen and Ličen, 2025), business and accounting (Kerr et al., 2025), healthcare (Alain et al., 2025; Asad and Faran, 2025) and communication (Gans, 2024; Van Quaquebeke and Gerpott, 2024). Current AI applications based on deep learning architectures and complex neural networks enable everyday users to generate content (e.g., images, videos and texts) and computer code (He et al., 2025; Kerr et al., 2025). Nevertheless, since the introduction of ChatGPT in November 2022, the conversation around the opportunities and risks of AI technologies for communication professionals has come to the forefront. Recent advances in Generative AI (GenAI) (e.g., ChatGPT), using large language models (LLMs), have been recognized for enhancing communication practices. However, trust in GenAI responses has emerged as an ethical concern, especially when organizations rely on them for decision-making. Recent surveys found only 10% of respondents trust the information produced by ChatGPT (Malwarebytes, 2023). This accounts for the inability of industry leaders to fully capitalize on AI-related opportunities (58%) and manage the AI-based risks (59%), with only 18% being trained to use GenAI tools responsibly (Sandpiper, 2023). Simultaneously, AI anxiety, described as “the fear and trepidation being expressed about out of control AI” (Johnson and Verdicchio, 2017, pp.4), is becoming a barrier for communication practitioners to leverage AI’s strengths in strategic communication.
Ethics is a fundamental principle of communication practices. It enables organizations to effectively resolve dilemmas while acting in a socially responsible manner (Bowen, 2016). Professional associations, particularly in public relations, often release codes of ethics to raise awareness of ethical issues and ensure professional conduct among practitioners (Yang et al., 2016). However, AI ethics discourses, including its ethical guidelines, have been criticized for being performative, simplistic and superficial (Bowen, 2024). Technological companies’ elaboration are often viewed as “ethics washing”, a tactic to polish corporate image (de Laat, 2021), challenging the implementation of professional codes. Additionally, research probing the link between technologies and values shows that the knowledge informing technological design conveys various layers of human values (Barry et al., 2017; Friedman et al., 2013). Following this logic, the timely incorporation of human values - conveyed by the ethical principles of GenAI - into both technological design and professional codes of ethics helps guide the ethical use of GenAI tools, shaping the reputation of an organization and the professions.
A growing number of studies has examined the attitudes of current and prospective communication practitioners toward integrating AI into communication management (see Christensen and Andersson, 2025; Jiang et al., 2025; López Jiménez and Ouariachi, 2021; Zerfass et al., 2020). They reveal a limited understanding of AI among communication professionals and call for managerial staff to recognize AI implementation as a leadership issue through educating themselves and their teams about the AI technologies (Zerfass et al., 2020). However, research on AI ethics within communication and public relations is scarce (see Bowen, 2024; Dong and van den Berg, 2025; Park and Yoon, 2024). Despite investigating perceptions of public relations professionals regarding AI adoption and ethical concerns, these studies often fall short in exploring the implications of ethical AI usage for core business of communication and public relations like relationship building, with exceptions such as Men et al. (2022) and Park and Yoon (2024). As noted, the ability to build trusted relationships with stakeholders, along with creativity and critical thinking, are central skills for public relations practitioners (Galloway and Swiatek, 2018). This points to the significance of developing AI ethical guidance to inform AI engineering and its adoption in the communication industry. This reasoning corresponds with the core principles of applying the ethics of care to technological ethics, which prioritize offering “caring responses to the particular needs and interests of those affected by a moral decision” (Weinberger, 2024, p. 1).
Drawing on the ethics of care to foster quality organizational-public relationships (OPR), this article examines the extent to which current public relations practices involving AI tools align with domain-specific GenAI guidelines in communication and public relations. To achieve this research aim, the following research questions (RQs) are proposed:
What is a proposed AI ethics framework designed for the communication field, with a focus on public relations?
To what extent does the industry’s understanding of GenAI ethics align with this proposed AI ethics framework?
What are the implications of operationalizing this proposed AI ethics framework for fostering quality organization-public relationships?
The research outcomes contribute to the theoretical and practical construct of responsible GenAI through the lens of care ethics. Theoretically, the development of an AI ethics framework for communication and public relations connects AI ethics principles from computer science with the professional codes of conduct in public relations. It advances understanding of how GenAI technologies can facilitate OPR while mitigating associated risks. Practically, this framework enriches existing professional codes of conduct, which technologists can use as a reference for designing, engineering and evaluating responsible AI systems in a given domain. It provides integrated mechanisms, including metrics, to AI users, developers and regulators for ethically designing, deploying and evaluating the effectiveness of GenAI tools. This prompts ongoing trans-disciplinary and cross-sector dialogues and partnerships among interdependent stakeholders (e.g., practitioners, technologists, scholars and policymakers) to explore responsible GenAI.
This article begins by reviewing the current use of AI technologies in communication practices, followed by a discussion on AI ethics and the ethical principles governing AI in computer science. It then conceptualizes OPR via linking to the role of ethics and codes of ethics in public relations, using the ethics of care as an analytical framework. This is followed by reasoning the influence of ethical messaging on building trust and forging OPR in the rapidly evolving AI-empowered communication landscape. The discussion contributes to the development of a domain-specific AI ethics framework tailored for the communication and public relations field. Aligning with the Code of Ethics released by the Public Relations Society of America (PRSA), a globally recognized association leading the establishment of guidelines for regulating ethical behavior in public relations (Fitzpatrick, 2002; n.d.), it illustrates how to operationalize this framework to nurture quality OPR, addressing RQ1. The methodological design adopts content analysis to analyze industry discourse based upon this framework, with the findings presented in the “Results” section, responding to RQ2. The final section refines the proposed AI ethics framework based upon the empirical findings and explores its implications for fostering quality OPR, addressing RQ3. This article concludes by discussing its limitations and avenues for future research.
2. Literature review
2.1 Defining GenAI and the use of GenAI in the communication industry
Artificial intelligence (AI) is acknowledged as a technical concept rooted in computer science (Celik, 2023). This term originated in the mid-1950s, coined by John McCarthy, a US-American computer scientists and AI pioneer. This term, with an emphasis on “intelligence”, underpins AI as computational agents possessing human-like competencies in perception, reasoning, learning and communication, performing tasks contextually by learning from experience and making appropriate decisions (Zerfass et al., 2020).
GenAI uses a very large corpus of data to generate new set of data at the request of users. It is understood as algorithms that examine extensive data sets to identify patterns, trends and make predictions (Martin, 2019). AI-related technologies encompass robots and digital systems that use learning methods, like machine learning systems, deep learning systems, rule-based systems, imaging recognition systems, natural language processing and voice recognition (Sun, 2021). They can be categorized into three forms according to their development stages and capacities: narrow AI (e.g., chatbots, service robots and voice assistants performing tasks better than humans), artificial general intelligence (often considered a theoretical future technology outperforming humans at every cognitive level, with the ability to generalize knowledge and improve itself through self-reflection using accessible data) and artificial superintelligence (that overtakes human brain capacity in all aspects) (Neuhofer et al., 2021).
AI-powered tools are beneficial for enhancing core functions of communications. They assist with target optimization (e.g., Cision Insights, Notified), copywriting and editing (e.g., ChatGPT, Midjourney), quantifying value (Acree, 2023) and gathering social insights (e.g., Sprout’s Social) (Gomez, 2023). The first three functions enable communication professionals to generate draft content, pitch stories to suitable outlets and reporters, refine existing content to match a brand’s tone and convey organizational value by identifying data patterns (Acree, 2023). The fourth function uses multilingual sentiment analysis and machine learning to deepen understanding of customer attitudes (Gomez, 2023).
However, concerns about the lack of competencies in individuals and organizations to mitigate AI-based risks are increasingly being raised. Insufficient training is evident, with only 17% of communication professionals trained in managing deep fakes and just 10% in identifying such content (Sandpiper, 2023). AI alarmists fear that future developments could cause a loss of control and disasters for society (Johnson and Verdicchio, 2017). This AI anxiety is criticized for sociotechnical blindness, overlooking the role of humans and social institutions within which AI operates (Johnson and Verdicchio, 2017). According to Celik (2023), AI literacy should encompass two dimensions: interaction with AI-based technologies (e.g., knowing, comprehending, using, evaluating and developing) and awareness of ethical issues. To maximize AI strengths in communication and alleviate downsides, research should offer clear, practical guidance for using AI tools.
2.2 GenAI ethics in computer science
Since 2016, AI ethics have been popularly discussed by companies, civil society organizations and governments (de Laat, 2021). AI ethics comprises data ethics and business ethics, reflecting AI’s reliance on large data volumes for training to meet business purposes (Zhou et al., 2020). Adhering to AI ethics nurtures responsible AI, characterized as “fair and non-biased, transparent and explainable, secure and safe, privacy-proof, accountable, and to the benefit of mankind” (de Laat, 2021, p. 1135). Key principles of AI ethics include transparency, fairness, security, accountability, accuracy, privacy and machine autonomy (Jobin et al., 2019; Kieslich et al., 2022), with global consensus highlighting privacy, transparency and safety (Kerr et al., 2020). GenAI is expected to uphold these principles (Wang et al., 2023) for the purpose of developing human-centric AI that benefits society (Kieslich et al., 2022). Table 1 explains what each AI ethics principle means.
AI Ethics principles in computer science (Kieslich et al., 2022, p. 6; Jobin et al., 2019)
| AI ethics principles | Descriptions |
|---|---|
| Explainability | Explainability enhances transparency and acts as a key indicator for measuring it. By clarifying processes, explainability helps minimize harm and improve both AI tools and trust. For instance, being transparent involves explaining how data is sourced and used, as well as how automated decisions are made |
| Fairness | Fairness is closely linked to equity and justice, focusing on preventing, monitoring, and mitigating unwanted bias and discrimination. This involves acquiring and processing data, especially training data, that is accurate, complete and diverse |
| Security | Security, in relation to safety, serves as a key indicator for evaluating “non-maleficence.” the use of AI should not cause harm, such as discrimination, privacy violations, or physical, psychological, or emotional harm and should also consider socio-economic aspects such as social well-being and infrastructure |
| Accountability | Accountability measures responsibility, requiring individuals to act with integrity and comply with legal obligations |
| Accuracy | Accuracy means accepting almost no errors and can be used to evaluate societal well-being |
| Privacy | Privacy is not only a value to uphold but also a fundamental right to be protected, often involving the safeguarding of data and ensuring data protection |
| Machine autonomy | A computer system is capable of functioning properly within a defined scope without human supervision |
AI development aims to build autonomous agents with general intelligence and expert systems to perform tasks in specialized domains (Kerr et al., 2020). The goal of AI development suggests that the interpretation of GenAI ethics depends on the domains and contexts of application. Additionally, human values, intrinsic (e.g., happiness, health) and extrinsic (e.g., privacy, security, trust) (Barry et al., 2017), are embedded through technology design elements like content, context and user experience (Barry et al., 2017; Friedman et al., 2013). However, blind trust, over-informing, false forecasting, over-regulation and self-reference in monitoring (Zhou et al., 2023), plus competing metrics to assess ethics principles (Kieslich et al., 2022), hinder GenAI ethics implementation. For instance, reducing algorithmic discrimination against social minorities requires collecting sensitive information, causing a trade-off between fairness and privacy (Kieslich et al., 2022). Hence, calls have been made to unveil societal expectations on GenAI ethics principles across domains among AI developers, researchers and industry users (Jobin et al., 2019).
2.3 The ethics of care and GenAI ethics in communication through Organization-Public relationships
Ethics are at the core of building trust-based relationships between an organization and the public (Bowen, 2016). Ethical behavior is measured through four frameworks: virtue, deontology, utilitarian and the ethics of care. Virtue ethics prioritize character development over right actions, nurturing qualities like gratitude, self-respect and courage in the quest for truth (Baker, 2008; Bowen, 2016, 2018). Deontological ethics rely on principles, rights, duties, freedoms and logic to form moral standards (Bowen, 2016). The utilitarian perspective defines what is ethical by assessing potential consequences and conditions, comparing them to possible positive outcomes (Bowen, 2016). These three schools of ethics concentrate on “what moral situations have in common” (Weinberger, 2024, p. 5). In contrast, the ethics of care pays attention to generating caring responses for “particular needs, wants, and values in the concrete situation” (Weinberger, 2024, p. 5). It focuses on creating a caring environment, using a feminist approach to support humanity, genuineness, reciprocity and the protection of vulnerability in human interaction to address public needs (Dong and Morehouse, 2022).
The ethics of care provides a valuable moral framework for examining AI ethics, especially by addressing the shortcomings of machine learning such as generalization. Machine learning, a core AI subfield (Celik, 2023), involves generalizing from data by identifying patterns of correlations and statistical significance to interpret new inputs and make predictions (Weinberger, 2024). This training process streamlines complex situations by memorizing particulars about trained items, but fails to find patterns to new inputs it has not been trained on (Weinberger, 2024), potentially causing harm to individuals and society – a primary source of concern regarding AI bias. Integrating care ethics into AI ethics entails a commitment to weaving together an array of particulars in ethical decision making. Instead of blindly following ethical guidelines and principles, this shift focuses on particulars (Weinberger, 2024), such as interdependent relationships, context and circumstances, vulnerability and voice when unpacking algorithmic decision-making (Villegas-Galaviz and Martin, 2024).
Care ethics offers organizations a mechanism to minimize AI-related risks to their publics. A key goal of care ethics is to prevent harm to individuals and society by supporting vulnerabilities, relationships, context, emotions and empathy in ethical decision-making (Hamington and Sander-Staudt, 2011). GenAI often struggle to distinguish facts from fakes within human-like conversations and differentiate authorship between humans and AI in written work (Zhou et al., 2023). The PRSA specifies AI-based risks, including misinformation, leaking proprietary information, trade secrets and intellectual property, infringing on copyrights and biases ingrained in AI (Staley et al., 2023). These risks resonate with six social concerns related to LLM-based GenAI technologies: discrimination, hate speech and exclusion; information hazards; misinformation harms; malicious uses; human-computer interaction harms; and environmental and socioeconomic harms (Weidinger et al., 2022). Table 2 provides definitions of these risks.
Social concerns about AI technologies
| Social concerns about AI | Descriptions (Weidinger et al., 2022) | Examples of coding excerpts from PRWeek |
|---|---|---|
| Discrimination, hate speech and exclusion | Language models (LMs) can reflect harmful language from their training data, which may perpetuate social stereotypes or unjust treatment of marginalized groups, provoking hate, violence, offence | “A search for 'gay couple’ on Midjourney returned hundreds of square-jawed, white, male pairs” |
| Information hazards | LM predictions using true but private data can cause privacy violations by leaking sensitive information, potentially cause emotional distress and infringing on a person’s rights | “While some mistakes generate funny images, in others, the results can be damaging for individuals, sections of society and brands” |
| Misinformation harms | LMs that unintentionally produce false, misleading, or poor-quality information can cause harm by misinforming, deceiving, or resulting in material damage to a person | “The only issue was that the cancer survivor was made up by AI - her emails, her quotes, and even her image” |
| Malicious uses | This risk is associated with humans deliberately using the LMs to cause harm, such as through scams, fraud and targeted disinformation campaigns | “… measurement firm onclusive had fallen victim to a malicious and targeted cyber-attack …” |
| Human-computer interaction harms | LM-based conversational agents, such as advanced care robots, can make human-computer interactions more akin to human-human interactions, potentially exploiting and violating users’ privacy and reinforcing discriminatory stereotypes | “Generative AI will also invariably predict the most likely response to questions, which risks losing the individual viewpoints and nuances captured through speaking with real people” |
| Environmental and socioeconomic harms | Training and using language models require substantial energy and when combined with the uneven impacts of automation – such as job losses – can cause environmental and socioeconomic harms | “The PR industry is antsy about AI … from nervousness about automation replacing human workers” |
An organization’s strong commitment to care ethics evokes quality OPR. Developing and sustaining quality OPR is a primary goal in the research, practice and education of communication and public relations, because of their impact on organizational effectiveness (Ma et al., 2023). Trust, commitment and satisfaction are the three key enablers of quality OPR (Aghazadeh and Aldoory, 2023; Cheng and Lee, 2023). These drivers provoke interactions between organizations and its publics, reinforcing the commitment to engagement with their publics and communities (Aghazadeh and Aldoory, 2023). Boosting public trust relies on an organization’s ethical conduct, centering around privacy, confidentiality agreements, informed consent, credibility and information veracity (Sebastião et al., 2017; White and Boatwright, 2020).
The implementation of AI is a key aspect of the broader digital transformation in organizational processes. Technology adoption empowers end users by enabling them to generate and transfer knowledge, build connections and exert influence on others (Foucault, cited in Lund et al., 2018; Li and Feng, 2024; Schivinski et al., 2025). Specifically, AI applications enhance storytelling (e.g., media pitches, messaging and executive communication), while supporting listening, publishing, customer care, advocacy and reputation management, especially during crises (Gomez, 2023). Communication professionals have increasingly integrated GenAI systems into their workflows since 2023, requiring them to explain their organization’s use of AI technology to external publics (Christensen and Andersson, 2025) to foster public trust in their content production. In this case, AI end users comprise communication professionals, the organizations they work for (Bowen, 2024).
Quality OPR involving the commitment to care ethics can be mediated by power distribution between an organization and its publics. Unequal distribution of power – defined as the ability to influence through control of resources - among stakeholders can cause conflicts between organizations and their public, causing mistrust toward the organizations (Vallet et al., 2020). From the end-user perspective, individual empowerment is confined by an individual’s perceived sense of power and willingness to adopt technology, which depends on whether the environment supports the use of these tools (Li and Feng, 2024; Tye et al., 2018). To minimize the impact of asymmetrical power relationships, organizations must engage stakeholders authentically, using two-way, ethical communication mechanisms to ensure information flow between the organization and its publics (Li and Feng, 2024). This suggests information about the processes of designing, developing, implementing and evaluating AI technology should be transparently shared among actors, including AI developers, researchers, scientists and end users.
The public tends to develop a trust-based relationship with an organization, when it is perceived as ethically and socially responsible (Oh et al., 2023). Conversely, deceptive messages undermine trust by affecting the perceived ethicality of its actions (Baker and Martinson, 2001). As communication professionals increasingly use GenAI tools for content generation, risks often involve source reliability and data security (Gomez, 2023). Accordingly, ethical messaging founded on truthfulness (of the message), authenticity (of the persuader), respect (for the audience), equity and social responsibility (Baker and Martinson, 2001) builds source reliability and data security. Among these, truthfulness, less mentioned in computer science (ScienceDaily, 2023), should be reinforced as a key principle for GenAI ethics in communication and public relations.
Based upon the requirement for ethical messaging, this article posits that the AI ethics framework in communication, particularly in public relations, should encompass the principles of truthfulness, authenticity, respect, equity and social responsibility. Prioritizing these principles help achieve the ethical principles of AI - transparency, accountability and explainability – which are central to computer science. According to Russo et al (2024), shifting the focus from trusting the output of AI systems to trusting the entire process (e.g., design, development and use) that generates the outcome, and evaluating this process with inputs from both experts and non-experts, ensure fair and just AI, thereby advancing the principles of transparency (a form of accountability) and explainability. Hence, deploying this proposed framework in communication practices helps mitigate AI-based risks and should become integral to an organization’s ethical behavior, thereby fostering quality OPR.
2.4 Professional codes of ethics in communication and the PRSA
Successful translation of the proposed AI ethics framework into real-world communication practices relies on its alignment with professional codes of ethics. Rather than departing from the logic of care ethics, the contextualization of AI ethics principles within professional history and established norms of good behavior contributes to AI design and governance (Mittelstadt, 2019). Professional codes of ethics are often enacted by professional associations and applied to their members to ensure a high level of professionalism (Kim and Ki, 2014). Codes of ethics in public relations serve four functions: “providing guidelines for practitioners”, “illustrating what clients and supervisors should expect from practitioners”, “providing a basis for charges that wrongdoing has occurred” and “providing a defence against charges of wrongdoing” (Kruckeberg, cited in Wright, 1993, p. 16). Internationally recognized codes of ethics for the public relations profession highlight six core values: professionalism, advocacy, moral standards, clients’ rights, expertise and relationship building, with the focus on the first three values (Yang et al., 2016).
Codes of ethics should also act as a crucial guide for computing professionals to practice the ethics of care in AI innovation. Technologists’ efforts to embed domain-specific codes of ethics and relevant values into AI development are essential for building responsible AI, resolving the challenge of converting abstract ethical principles into operational systems (Diakopoulos et al., 2024). Seen in this light, proper interpretation and implementation of domain-specific codes of ethics enable computing professionals to enhance AI technologies, facilitating ethical AI usage among communication professionals.
The PRSA has been leading the development of codes of ethics for the public relations profession (Fitzpatrick, 2002). Since introducing the first Code of Professional Standards in 1950, it has revised the code of ethics several times to reflect industry practices and rising expectations for ethical conduct (Fitzpatrick, 2002). The current PRSA Code of Ethics exhibits the state of ethics in public relations (Bowen, 2007), featuring six core values: advocacy, honesty, expertise, independence, Lent et al. 2024.
PRSA Code of ethics: Six values and the provisions (PRSA, 2025a, 2025b)
| PRSA code of ethics | Value | Interpretation | PRSA code of ethics provisions |
|---|---|---|---|
| Advocacy | Informed public debate | We serve the public interest by acting as responsible advocates for those we represent. We provide a voice in the marketplace of ideas, facts, and viewpoints to aid informed public debate | N/A |
| Honesty | Accuracy and truth | We adhere to the highest standards of accuracy and truth in advancing the interests of those we represent and in communicating with the public | Free flow of information – to verify the AI-generated information and the sources to ensure accuracy |
| Expertise | Providing the highest quality work | We acquire and responsibly use specialized knowledge and experience. We advance the profession through continued professional development, research, and education. We build mutual understanding, credibility, and relationships among a wide array of institutions and audiences | Enhancing the profession – public relations professionals should work with interdisciplinary teams across the organization to ensure ethical design, development, and usage of AI tools |
| Independence | Being objective and accountable for actions | We provide objective counsel to those we represent. We are accountable for our actions | Disclosure of information – public relations practitioners should be encouraged to undertake and support AI fact-checking efforts to prevent the spread of mis- and disinformation |
| Loyalty | Being faithful and working for the public interest | We are faithful to those we represent, while honoring our obligation to serve the public interest | Safeguarding confidence – to manage confidential information cautiously so as to protect the privacy rights of clients, organizations and individuals |
| Fairness | Encouraging free expression | We deal fairly with clients, employers, competitors, peers, vendors, the media, and the general public. We respect all opinions and support the right of free expression | Competition – to ensure a diverse Pool of applicants, thereby minimizing bias and enhancing diversity, equity and inclusion in hiring processes |
The values stated in the PRSA Code of Ethics support the aforementioned principles of AI ethics in communication and public relations. These principles encompass truthfulness, authenticity, respect, equity and social responsibility, all contributing to the requirement for ethical messaging. Hence, this article proposes an AI ethics framework tailored to the communication and public relations field, highlighting these principles. Table 4 illustrates the components of each AI ethics principle and how the proposed framework is supported by the PRSA Code of Ethics and its values.
A proposed AI ethics framework in communication and public relations
| AI ethics principles | PRSA code of ethics | Descriptions of the AI ethics principles | Examples of coding excerpts from PRWeek |
|---|---|---|---|
| Truthfulness | Honesty | Truthfulness, or veracity, means that GenAI-generated content is accurate and truthful | “The platform [CisionOne]… helps users more quickly and accurately identify relevant news and trends …” |
| Authenticity | Expertise | Authenticity, related to credibility, refers to the trustworthiness or integrity of the information source or communicator | “'Chat GPT is inaccurate … google has accuracy problems, too … all of this highlights something marketers and communications professionals have always known: accuracy and trust matters above all else” |
| Respect | Fairness | Respect involves treating recipients and individuals involved in GenAI-generated messages with dignity by protecting their privacy and confidentiality, and obtaining informed consent | “Privacy, security, and AI: Although a hot topic … consumers are far more cognizant and concerned about where information is stored and what content they share” |
| Equity | Fairness | Equity refers to the fair treatment of recipients and individuals involved in GenAI-generated messages, such as ensuring the content provides sufficient information, tailored to their capacity and experience | “… AI portrait generator lensa oversexualizing female users as an example of how the technology can bring human biases front and center …” |
| Social responsibility | Advocacy | Social responsibility means ensuring that communicators of GenAI-generated messages contribute to social good, particularly by addressing AI-related security and safety issues | “We need to work to ensure that all sectors of society, including differences in gender, race, economic circumstances and more, are included and acknowledged when designing data sets and algorithmic systems” |
| Independence | |||
| Loyalty |
3. Method
This study adopted content analysis to analyze the data. Content analysis involves systematic examination of communicative texts to uncover patterns, including coding and interpreting the context and meaning of the texts (Selvi, 2020; Li et al., 2024). The data was sourced from PRWeek (Link to the webcite of prweek), a leading trade magazine and news website for the public relations and broader communications industry. Published by Haymarket Media Group, PRWeek is renowned for its extensive reach and sizable subscriber base, targeting communication practitioners (Edwards and Pieczka, 2013). Public relations and communication scholars often select this trade media to research professional projects in the field (see the examples of Berg and Miller, 2023; Edwards and Pieczka, 2013).
PRWeek publishes articles in four genres: “Feature”, “Opinion” and “News Analysis” and “News”. Searches were conducted within the subscribed version of PRWeek, using the genres “Feature”, “Opinion” and “News Analysis”. Unlike “News”, these genres allow authors to express personal opinions (Calame, 2006), meeting the aim of this study to investigate the industry’s understanding of AI ethics. The search terms “AI and Ethics”, “Artificial Intelligence and Ethics” and “AI ethics” were combined via the Boolean operator “OR”. The research timeframe spanned the last five years, from January 2019 to April 2024, to capture changes during this period when AI ethics have been actively discussed (Zhou et al., 2023). This produced a total of 597 articles. Using Python programming, 250 articles unrelated to AI ethics, focusing instead on general ethics, were removed. After the data screening process, 347 articles were extracted for inclusion in the final corpus for content analysis, with each article treated as a unit of analysis.
To address RQ2, we followed these steps to uncover: 1) an overview of the trend, including sentiment analysis (e.g., positive, neutral and negative, based on the article’s overall attitudes toward AI and its ethical usage), publication volume across the years and genres; 2) how AI ethics principles were reflected in public relations industry discourse; and 3) social concerns related to AI-based risks. For Step 2, relevant texts were structured into the coding categories – the five AI ethics principles – according to their descriptions in the proposed AI ethics framework in communication and public relations, as shown in Table 4. Likewise, for Step 3, relevant texts were organized into the six social concerns about AI risks, as outlined in Table 2. Examples of coding excerpts from PRWeek are presented in Table 2 and 4.
Two independent coders were well trained for the coding process. They closely read each article in the corpus and individually classified the texts into relevant coding categories according to Table 2 and 4. Prior to coding, the two coders independently manually coded 35 randomly selected articles, representing over 10% of the total sample (n = 347), to establish intercoder reliability. This was calculated using Holsti’s method, achieving a score of 0.91. A reliability score of 0.80 or above is considered reliable, indicating strong agreement between the two coders’ judgements (Li et al., 2024). The coders then proceeded to code the remaining articles. Any disagreements in coding results were discussed between the coders. If unresolved, further discussion with the researchers took place until agreements were reached.
4. Results
RQ2 examined to what extent the industry’s understanding of GenAI ethics aligned with the proposed AI ethics framework in this study. The analysis from Step 1 showed that over 66% of the articles (n = 230, 66.28%) demonstrated a neutral sentiment toward the use of AI and its associated ethical considerations. Nevertheless, the articles with a positive stance (n = 97, 27.95%) outnumbered those with a negative tone (n = 20, 5.76%). This finding was in line with the aforementioned, positive impacts of AI-powered tools on communication practices. Thereby, communication professionals are encouraged to experiment with rather than abandon AI tools, and organizations should invest in AI-enabled tools as part of their communications toolbox (Acree, 2023; Gomez, 2023).
Positive articles portrayed how AI-powered virtual influencers, like Rae and VaVa, enhanced authenticity by creating brand personalities and effectively engaging with their followers to build brand communities (Ragavan, 2021). Moreover, Singal AI, an AI-powered media intelligence tool, assisted communication teams in tracking stories internationally across multiple platforms, seeding the success of the “Santa School” campaign (PRWeek, 2020). In contrast, negative articles exemplified AI’s drawbacks. For example, AI-generated images of high-profile individuals, which are often indistinguishable from real ones, can damage brand reputations if not treated legally and with consent (Swaika, 2023). Additionally, media pitching AI tools often fail to grasp the nuances of media operations, due to the absence of human and emotional intelligence needed for effective pitches (Croydon, 2023). This finding indicated that the public relations industry has taken a cautiously optimistic stance toward AI-related issues over the last five years. The consistent sentiment trend observed annually from 2019 to 2024 further evidenced this point. Figure 1 outlines the overall sentiment distribution, and Figure 2 the sentiment distribution by years.
The image features a bar graph illustrating the distribution of records classified into three categories: Neutral, Positive, and Negative. The y-axis represents the number of records, with numerical values ranging from zero at the bottom to over two hundred at the top. The Neutral category has the tallest bar at two hundred thirty records, amounting to sixty-six point twenty-eight percent of the total. The Positive category displays a shorter bar with ninety-seven records, representing twenty-seven point ninety-five percent. The Negative category has the shortest bar, showing twenty records, which constitute five point seventy-six percent.Overall Sentiment Distribution
Source: Authors’ own work
The image features a bar graph illustrating the distribution of records classified into three categories: Neutral, Positive, and Negative. The y-axis represents the number of records, with numerical values ranging from zero at the bottom to over two hundred at the top. The Neutral category has the tallest bar at two hundred thirty records, amounting to sixty-six point twenty-eight percent of the total. The Positive category displays a shorter bar with ninety-seven records, representing twenty-seven point ninety-five percent. The Negative category has the shortest bar, showing twenty records, which constitute five point seventy-six percent.Overall Sentiment Distribution
Source: Authors’ own work
The stacked bar chart presents the proportion of sentiment categories across six years from 2019 to 2024. The vertical axis represents percentage proportion, while the horizontal axis lists the years. Each bar is divided into three sentiment levels: negative, neutral, and positive. The results indicate that neutral sentiment remains the most frequent throughout the years, with positive sentiment following closely. Negative sentiment shows minor fluctuations but remains comparatively low. The overall pattern reveals a stable yet slightly increasing trend in neutral and positive responses over time.Sentiment Distribution by Years
Source: Authors’ own work
The stacked bar chart presents the proportion of sentiment categories across six years from 2019 to 2024. The vertical axis represents percentage proportion, while the horizontal axis lists the years. Each bar is divided into three sentiment levels: negative, neutral, and positive. The results indicate that neutral sentiment remains the most frequent throughout the years, with positive sentiment following closely. Negative sentiment shows minor fluctuations but remains comparatively low. The overall pattern reveals a stable yet slightly increasing trend in neutral and positive responses over time.Sentiment Distribution by Years
Source: Authors’ own work
The articles in the “Feature” and “Opinion” genres exhibited sentiment patterns that mirrored the overall trend in Figure 1. More than 65% of these articles displayed neutral sentiment, with a minority showing negative sentiment, followed by positive sentiment. However, there were more “Feature” and “Opinion” articles than “News Analysis” pieces, reflecting the editorial policies of PRWeek. “News analysis” as an analytical form of journalism centers on examining news rather than commenting on or expressing personal views about news events (Calame, 2006). Figure 3 presents these details.
The three bar charts display sentiment distribution in feature, news analysis, and opinion articles. Each chart includes bars representing neutral, positive, and negative sentiments. The vertical axis indicates the count, while the horizontal axis specifies sentiment type. In the feature category, neutral responses have the highest count, followed by positive and a small number of negative ones. News analysis shows a similar trend, though with fewer total entries. Opinion pieces also show predominant neutrality, followed by moderate positive and minimal negative sentiment.Sentiment Distribution by Genres
Source: Authors’ own work
The three bar charts display sentiment distribution in feature, news analysis, and opinion articles. Each chart includes bars representing neutral, positive, and negative sentiments. The vertical axis indicates the count, while the horizontal axis specifies sentiment type. In the feature category, neutral responses have the highest count, followed by positive and a small number of negative ones. News analysis shows a similar trend, though with fewer total entries. Opinion pieces also show predominant neutrality, followed by moderate positive and minimal negative sentiment.Sentiment Distribution by Genres
Source: Authors’ own work
Followed by the above, Step 2 examined AI ethics principles from the perspectives of communication and public relations, as well as social concerns about AI risks. The finding showed social responsibility (n = 51.73%) was the most discussed principle, followed by authenticity (n = 41.04%). The other principles - truthfulness, respect and equity - were discussed almost equally, each around 30%. Specifically, the articles in PRWeek demonstrated a consensus between AI technology organizations (e.g., Cision and Anthropic) and communication firms (e.g., Pagefield), viewing facilitating responsible GenAI as part of the organizations’ social responsibility. They advocated for government regulations and AI codes of ethics to guide the ethical development and use of AI products by employees, highlighting that actions should be “helpful(ly), honest(ly), and harmless(ly)” for their clients (Bach, 2023; Harrington, 2023; PRWeek, 2024).
Antony Cousins, executive director of AI strategy at Cision, stated that “[…] it’s the responsibility of the communications industry, and us as individual communicators, to set the course for how this technology will be used by future generations,” and “Cision is taking a leading role […] by acting responsibly, creating a proper framework, working with government agencies and even our competitors” (Bach, 2023). Referring to the proposed AI ethics framework, this result suggested that the PRSA Code of Ethics values – advocacy, loyalty, independence and expertise - are crucial for ensuring ethical behavior in the use of AI tools in communication practices, owing to their contribution to the AI ethics principles of social responsibility and authenticity. Figure 4 summarizes the statistics, where “Y” (Yes) indicates the article discussed a specified AI ethics principle, and “N” (No) vice-versa.
The stacked bar chart presents the percentage distribution of affirmative and negative responses for the values truthfulness, authenticity, respect, equity, and social responsibility. The vertical axis represents proportions in percentage, while the horizontal axis lists the five values. Truthfulness has 68.5 percent negative and 31.5 percent affirmative responses. Authenticity shows 59.96 percent negative and 41.04 percent affirmative. Respect and equity both have 71 percent negative and 29 percent affirmative. Social responsibility records 48.27 percent negative and 51.73 percent affirmative, marking it as the most supported value.Distribution of AI Ethics Principles in Communication and Public Relations
Source: Authors’ own work
The stacked bar chart presents the percentage distribution of affirmative and negative responses for the values truthfulness, authenticity, respect, equity, and social responsibility. The vertical axis represents proportions in percentage, while the horizontal axis lists the five values. Truthfulness has 68.5 percent negative and 31.5 percent affirmative responses. Authenticity shows 59.96 percent negative and 41.04 percent affirmative. Respect and equity both have 71 percent negative and 29 percent affirmative. Social responsibility records 48.27 percent negative and 51.73 percent affirmative, marking it as the most supported value.Distribution of AI Ethics Principles in Communication and Public Relations
Source: Authors’ own work
In response to Step 3, the analysis revealed that rather than primarily focusing on issues associated with AI, these PRWeek stories highlighted how public relations firms proactively sought solutions to resolve social concerns about AI risks. Specifically, “information hazards” (n = 25.07%) was identified as the prominent area of social concern caused by AI, closely followed by “misinformation harms” (n = 24.78%) and “environmental and socioeconomic harms” (n = 23.63%). Reportedly, public relations firms were aware of the harm AI can cause in spreading misinformation and disinformation, particularly in manipulating voters during presidential elections (Daniels, 2024). Alfredo Richard, EVP of corporate communication at NBCUniversal Telemundo Enterprises commented that “AI certainly raises the possibility of more disinformation […] The misuse of AI to potentially manipulate or create fake videos and audio is a real concern” (Daniels, 2024). To address the challenge, Fenton Communications has prohibited its creators from using GenAI to produce content, instead teaching organizations to use them as trusted messengers across multi-ethnic communities (Doerrer, 2024a). Shakirah Hill Taylor, Chief Digital officer at Fenton, draw attention to the need “to translate context and language in a way that is culturally relevant to a myriad of communities”, especially noting “the Filipino community, specifically young Latino men”, “Black and Latina women” and “Black men” (Doerrer, 2024a).
Meanwhile, “human-computer interaction harms” (n = 21.61%), “malicious users” (n = 19.31%) and “discrimination, hate speech, and exclusion” (n = 15.61%) were of slightly lesser concern. For instance, one of the reports highlighted that the adoption of generative AI by public relations firms has evolved from experimenting and feeling nervous about threats, such as automation causing job cuts, racial disparities, misinformation and untested legalities, toward implementing, including significant investment into AI technologies over the next three years to augment creativity (Doerrer, 2024b). This article urged public relations practitioners to confront the fact that AI has exacerbated existing racial bias against Black people, who are a small minority in the communication and marketing industries, while over 80% were identified as white in 2023. Public relations firms were encouraged to explore how to leverage AI tools to bring more Black entrepreneurs into these industries. Simultaneously, “a lot of holding companies and larger agencies” have already established “partnerships with technology providers and brands to bring forward the creative capabilities of this technology”, according to Nicole Greene, VP and analyst at consulting firm Gartner (Doerrer, 2024b). See the data in Figure 5.
The stacked bar chart depicts the proportion of affirmative and negative responses for six harm categories: discrimination, information hazard, misinformation, malicious use, human-computer interaction harm, and environmental or socioeconomic harm. The vertical axis represents the percentage proportion. Discrimination records 84.44 percent negative and 15.56 percent affirmative. Information hazard and misinformation show around 75 percent negative and 25 percent affirmative. Malicious use reports 80 percent negative and 20 percent affirmative. Human-computer interaction harm records 78.1 percent negative, while environmental or socioeconomic harm shows 76.37 percent negative and 23.63 percent affirmative.Distribution of Social Concerns about AI Risks
Source: Authors’ own work
The stacked bar chart depicts the proportion of affirmative and negative responses for six harm categories: discrimination, information hazard, misinformation, malicious use, human-computer interaction harm, and environmental or socioeconomic harm. The vertical axis represents the percentage proportion. Discrimination records 84.44 percent negative and 15.56 percent affirmative. Information hazard and misinformation show around 75 percent negative and 25 percent affirmative. Malicious use reports 80 percent negative and 20 percent affirmative. Human-computer interaction harm records 78.1 percent negative, while environmental or socioeconomic harm shows 76.37 percent negative and 23.63 percent affirmative.Distribution of Social Concerns about AI Risks
Source: Authors’ own work
5. Discussion
In keeping with the tradition of care ethics, contextualizing abstract GenAI principles within a given domain through linkage with professional codes of ethics contributes to the quality of organization-public relationships (OPR). This mechanism supports the development of practical guidelines for the ethical use of GenAI tools, building public trust while overcoming AI anxiety. As stated by Fitzpatrick (2002), timely updates to codes of ethics are central to keeping pace with evolving practices and disciplines. Driven by the rapid growth of AI technologies, this study examined how public relations practices using AI technologies adhered to domain-specific AI ethics principles in the communication field. The analysis of industry discourse represented by PRWeek revealed a neutral yet cautiously optimistic attitude toward using AI tools to enhance communication practices. It identified social responsibility as the most mentioned principle of AI ethics in public relations and communication, followed by authenticity, with truthfulness, respect and equity trailing behind. These findings reflect the recognized contribution of AI technologies to the communication industry, including the improvement of productivity, higher-value critical thinking and efficiency while decreasing cost (Staley et al., 2023). They demonstrate the future trend of communication professionals collaborating with AI developers in the search for responsible AI solutions, while urging the prioritization of the principles of authenticity and especially, social responsibility.
Further into this, the discussion on ethical messaging and its impact on quality OPR has contributed to the development of an AI ethics framework for the communication field, highlighting its application within public relations. This domain-focused framework captures key principles of AI ethics: truthfulness, authenticity, respect, equity and social responsibility. In light of Mittelstadt (2019), from the perspective of care ethics, codes of conduct, accreditation and licensing schemes endorsed by professional societies and boards help tackle the abstraction of AI ethics principlism. Accordingly, this study operationalizes each principle in the AI ethics proposed framework by referencing the PRSA Code of Ethics. As outlined in Table 4, social responsibility is tied to advocacy, independence and loyalty, while authenticity connects to expertise, emphasizing the generation of high-quality work to build credibility. Exercising the principles of truthfulness and respect contribute to honesty and fairness, reinforcing truth and embracing free expression. Moreover, Achieving the equity principle enhances the profession and facilitates fair competitions, promoting the ethical standards of expertise and fairness.
The interpretations above embody the logic of care ethics. Rooted in the recognition of interdependent relationships, care ethics instructs AI developers to understand the web of connections between individuals, their interests, responsibilities and the impact algorithms (Villegas-Galaviz and Martin, 2024). Thus, the expectation for computing professionals is not only to make technologies functional but also to address societal, ethical, legal and other human-centered concerns within specific domains, embedding domain-specific professional codes of conducts in data training, algorithms, performance metrics and organizational processes (Diakopoulos et al., 2024). For communication professionals, a team approach to use AI technologies ethically and innovatively helps provide clients with well-informed and workable solution (Galloway and Swiatek, 2018), lifting the credibility and integrity of the source and communicator. Conforming to the particular of context and circumstances, AI users and developers should use a relational approach to care, creating contextualized responsiveness that addresses factors like space, time, concrete circumstances and historical-cultural contexts in algorithms (Villegas-Galaviz and Martin, 2024). Vulnerability and voice involve guaranteeing the needs of vulnerable groups and diverse viewpoints are included in algorithms and training data (Villegas-Galaviz and Martin, 2024).
As noted, trust, along with commitment and satisfaction, creates quality OPR. Building upon “integrity, dependability and competence”, trust in public relations is understood as “the confidence and willingness of both organizations and publics to open up to one another” (Ma et al., 2023, pp.2). Professionalism and moral standards, as universal codes of ethics in public relations, are evaluated by “how well the public relations practitioners provides high quality services to the employer or client” and “individual integrity of the professional”, respectively (Yang et al., 2016, pp. 156–158). Hence, exercising this proposed AI ethics framework through the lens of care ethics enables communication professionals to refine GenAI-related practices and capitalize on GenAI’s potential to advance OPR.
However, the outcome of practicing the principles of social responsibility and authenticity is determined by how the based form of AI – Big Data – is sourced, trained, updated and monitored. It relies on how explainability, accuracy, fairness and machine autonomy – the ethical principles of AI in computer science – are conducted. Explainability, as a key contributor to transparency, requires clear explanations of data collection, usage and automated decision-making; accuracy demands little to no errors; fairness entails minimizing bias and discrimination through accurate, complete and diverse data; and machine autonomy requires a computer system to independently perform tasks without human supervision. Nevertheless, LLMs have been found to struggle with fully encoding factual knowledge (Lent et al., 2024). Subsequently, the ways of executing these principles in computer science during data acquisition and training lays the foundation for communication practitioners to uphold the AI ethics principles.
This point above teases out the underlying factors causing the tensions when applying AI ethics to public relations practice, affecting the quality of OPR. As noted, there is a duality in technology ethics: “the ethical purposes and context of that technology’s creation” by computer engineers and scientists, and “the ethical use of that technology by the end user” like communication professionals and the organizations they work for (Bowen, 2024, p.1). The ethical use of AI by communication professionals is influenced by a supportive environment which includes the ethical conduct of AI engineers. As demonstrated through the findings, public relations firms proactively explore practical solutions in collaboration with AI developers to address social concerns about AI-based risks. However, due to advancements in AI technology featuring high levels of automation, AI developers cannot predict system behaviors, and these automated decision-making systems are exacerbating unequal power relationships (Eubanks, 2018). To this end, the care ethics perspective, which emphasizes nurturing a caring environment in support of context and circumstances, interdependent relationships among multiple stakeholders, vulnerability and voice, is more appropriate for managing AI-based ethical challenges.
6. Conclusion
This article concludes that the current ambiguity in operationalizing AI ethics in public relations largely arises for the disconnection between AI ethics in communication and in computer science. Bridging this divergence helps optimize the benefits of using GenAI tools ethically in strategic communication, contributing to quality OPR. However, the quality of OPR is subject to the interconnected power dynamics among AI end users, AI developers and their respective organizations. Algorithms – coded instructions that enable a computer to undertake specific tasks – are socially constructed and negotiated rather than existing in isolation (Guerra, 2025). They are often designed and used by individuals with the authority to control resources and visibility to others (Bishop et al., 2020). Subsequently, the development and implementation of AI ethics guidelines, grounded in fair and unbiased algorithms (Russo et al., 2024), are shaped by diverse actors, intertwined with not only technical but also political, economic, organizational and personal contexts.
Hence, incorporating AI ethics principles from computer science into those in public relations through the proposed framework – a consideration of context and circumstances based upon the ethics of care - becomes a critical solution. These research outcomes feature the development of domain-specific AI ethics frameworks through linkage with professional codes of conduct to enhance OPR. It deepens the value of care ethics in shaping the ethical development and deployment of AI technologies via incorporating context-specific particulars into ethical decision-making.
7. Limitations and future work
There are several limitations to this study. First, relying on a single data source, PRWeek, limits the ability to generalize the findings across the entire communication industry. AI tools, while useful for generating comprehensive content, often lack the capability to effectively contextualize information (Robertson et al., 2024). Future research examining a wide range of industry magazines as well as probing digital employee voice would provide a better understanding of industry expectations on ethical AI adoption and innovation. Second, due to the rapid advancement of AI technologies transforming industries, regulations and codes regarding AI ethics must be monitored and updated timely to account for factors like culture, religion, time and circumstances within their socio-technical contexts (Kim and Ki, 2014). Simultaneously, concerns about code enforcement – its inability to penalize non-compliance and reward adherence – persist (Yang et al., 2016). Hence, research in this area should be continuous, including continuously uncovering more concrete guidance on operationalizing this framework in organizational and technical settings, with the interpretation and operation of AI ethics evolving alongside technological advancements, guided by care ethics to enhance the professions and OPR. Third, this study relies on a single research method – content analysis – to analyze the data. Although widely deployed in the communication field, it is important to adopt a mixed-method approach, combining quantitative and qualitative research methods to gain further insights (Li et al. 2021). Thus, future research should also conduct longitudinal studies and in-depth interviews with industry professionals to gather their experience and input, explore other professional codes and validate them cross-culturally to provide researchers and practitioners with guidance on ethical AI practices across various domains and in computer science.
The authors would like to acknowledge the contributions of Christopher Dodds, co-founder. Managing Director, at ICON Agency. The authors also extend gratitude to the following students: Jaekuk Lee at North Carolina State University, Maimaiti Munisha at Zhejiang University, and Sitong Wang at Capital Normal University/University of Northern Colorado, for their assistance in coding the data, also The authors thank Amy Yang from University of Technology Sydney for assisting the data collection.
