Since Rowley’s (2008) first scholarly publication about digital content marketing, the matter of how to establish expertise and advance brand awareness with brand content on different digital platforms has been well studied and documented (Bubphapant and Brandão, 2023). While there are various perspectives on content marketing in the literature, Beard et al.'s (2021) study of its history reveals consistent objectives and strategies, such as educational content distribution and brand management. Furthermore, the emphasis on value creation as part of marketing communication has been a consistent theme in content marketing practices over time, aligning with marketing theory’s service-dominant logic principles (Beard et al., 2021).

The diverse perspectives on content marketing are also evident in the numerous definitions. The Content Marketing Institute’s (n.d.) definition of content marketing as “a strategic marketing approach focused on creating and distributing valuable, relevant, and consistent content to attract and retain a clearly defined audience — and, ultimately, to drive profitable customer action,” is still widely recognised. Proposing an activity-based view, Koob (2021) defines content marketing as “a set of specific activities, comprising content marketing strategizing, content production, content distribution, content promotion, performance measurement and content marketing organization.”

To interact with, entertain, and build long-term loyalty with the intended audience, content marketers create brand content such as blog posts, articles, videos, social media posts, whitepapers, infographics, webinars, emails, newsletters, eBooks, case studies, and podcasts (Pulizzi and Piper, 2023). Content marketing thus allows for more interactivity in brand conversations (Ahmad, 2025). To rank well in search engines, content marketers depend on human–created content that requires a significant amount of time and resources to produce (Reisenbichler et al., 2022). Creating resonating and helpful brand content also requires repetition and consistency, while content marketers must face a target audience with “content fatigue” (Shehu, 2023). Marketers can thus use GenAI tools to create and distribute content in less time and effort (Capgemini Research Institute, 2023).

This study draws from Arrivé’s (2021) work, which states that digital brand content is a hybrid strategy that includes relational, transactional, and product aspects. Brand content aims to attract consumers and establish connections (relational), whereas product-related content drives sales and meets specific marketing objectives (transactional) (Arrivé, 2021).

GenAI is rapidly evolving with OpenAI, an artificial intelligence research laboratory, adding new advanced features to GPT-4-Turbo (an advanced version of ChatGPT) as of the writing of this paper. Already, brands are able to streamline all aspects of the content creation process, from demographic research, and brainstorming ideas to generating content (Brüns and Meißner, 2024).

ChatGPT and other GenAI applications are focused on content production, which makes them especially useful for content marketing (Wahid et al., 2023). For example, content marketers can create a prompt to generate a headline that matches the content, a blog post on a specific topic, or a topic outline, to name a few. The AI will then in response generate the requested outputs in seconds (De Cremer et al., 2023; Soni, 2023).

Using AI for content marketing has been reported to have many benefits. Early empirical data emphasises GenAI’s profound role in content creation in terms of personalisation and the generation of large-scale persuasive content (Brüns and Meißner, 2024). A case in point is that content can be created and optimised for search engines more quickly and affordably, improving productivity and lowering content strategy costs. Also, topic clusters generate more content ideas and enable more personalisation (Rodrigue, 2023). GenAI thus provides significant advances in content generation efficiency, scalability, and personalisation, all of which are critical to the success of digital marketing campaigns (Soni, 2023).

Table 1, below, indicates selected areas of content marketing activities where GenAI tools can be used, with examples.

However, views on AI’s effectiveness for generating brand content are still mixed. Puntoni et al. (2021) found that AI-generated content was less engaging and less often shared than human-generated content. On the other hand, an earlier study by Thontirawong and Chinchanachokchai (2021) showed that content marketing campaigns using AI had higher click-through and conversion rates. Similarly, a study by Hartmann et al. (2023) revealed that AI can produce comparable or better findings than human-generated content. However, Reisenbichler et al. (2022) argue that, while machine-generated content is intended to rank well in search engines, the role of the human editor is still necessary.

Nonetheless, a report published in 2023 by the Capgemini Research Institute identified an increasing trend in the use of GenAI for content marketing purposes, which is expected to continue in the future. Their report is based on the findings of a survey of 1800 chief marketing officers from 14 countries and 25 in-depth interviews with industry experts across different sectors. Their findings reveal that GenAI enables independent creation, innovation, and adaptation and is rapidly transforming traditional marketing strategies. For example, marketers adopt generative AI to develop campaigns, improve customer experiences, and perform data analysis (Capgemini Research Institute, 2023). Similarly, content marketers are increasingly using GenAI technologies to improve their strategies and operations. From content development, curation, and search engine optimisation (SEO) through distribution and performance analysis, AI can be applied at many phases of the content marketing process (Capgemini Research Institute, 2023; Wahid et al., 2023).

Standards and guidelines are important in addressing ethical concerns about AI. Because of the rapid advancement of AI technologies, the discussion of AI ethics has gained prominence, as has the number of ethical principles and standards published globally over the years (Cox, 2022). Jobin et al. (2019) identified 84 ethics guidelines issued by international agencies, governments, technology corporations, and others to guide ethical AI practices.

Academic debate about AI ethics is also widespread. In 2022, the journal AI published a special issue on “Standards and Ethics in AI” that examined various AI ethics standards and legislation being developed globally (Rivas and Bejarano, 2022).

Of note is a 2022 systematic review of the ethics of AI that revealed 22 global ethical principles and 15 challenges. Amidst the many global ethical principles, transparency, fairness, privacy, and accountability are highlighted as the most commonly needed AI ethics principles. On the other hand, the most significant challenges are a lack of knowledge about ethics and vague principles (Khan et al., 2022).

Hagendorff (2020) evaluated 22 AI ethics guidelines and realised that current AI ethics regulations fail due to a lack of reinforcement systems, making ethics appear as an afterthought rather than as an integral part of using AI. Also, software developers frequently prioritize economic incentives over ethical values.

In recent years, the ethical principles of GenAI have faced criticism, especially in business and academic settings. There is also scholarly concern regarding bias and discriminating trends in the outputs of GenAI systems. To illustrate, Huanga et al. (2025), in a systematic mapping review of 39 publications, observed that GenAI models routinely reproduce established social stereotypes. This tendency can be attributed to biases within the datasets used during model training. Although various corrective strategies have been proposed, their application remains inconsistent, with shortcomings in sectors such as healthcare and public governance.

Concerns regarding transparency, authorship, and intellectual property rights are also gaining attention. The OPUS Project (2024) highlights the lack of transparency of many GenAI models, which are often referred to as “black boxes” due to their limited interpretability. The lack of transparency, along with the absence of clear disclosure guidelines, makes it difficult to verify results and determine who is accountable. Transparancy and intellectual property can become important in brand content creation, where originality and traceability of content are essential.

Furthermore, the European Commission (2024) cautions against the use of third-party data without consent and refers to the legal implications of generating content based on copyrighted materials. The Commission’s recommendations call for disclosure of any GenAI tools used in research and emphasize the importance of correct attribution of such work. Singh et al. (2025) add to these concerns by addressing the broader legal and ethical uncertainties surrounding authorship and intellectual property. They point out that existing legal systems do not recognize AI as capable of holding rights or responsibilities, thus complicating the allocation of credit and accountability in AI-assisted work.

Another review of 47 AI ethics guidelines found that they were inapplicable in both business and political economy settings. According to Attard-Frost et al. (2023), AI ethics guidelines prioritise algorithmic decision-making over aspects such as fairness, accountability, sustainability, and transparency in the context of business decision-making for AI systems.

These differing views show that while existing AI ethics frameworks provide a helpful starting point, they are often inadequate when it comes to dealing with the complex, real-world ethical challenges of using GenAI in areas like content marketing. A more tailored ethical framework is therefore required that considers both normative principles and the practical implementation challenges associated with using GenAI for brand content creation in content marketing.

While the adoption of GenAI has increased rapidly, marketers around the globe are both cautious and aware that there are potential ethical risks when using GenAI to create content (Capgemini Research Institute, 2023). Already, large language modules, such as Chat GPT, have revealed that they can show training prejudice when generating content (Ray, 2023). Also, many incidences of inaccurate information (known as hallucinations) have been reported (Gocklin, 2023). Fake or fabricated information can harm the business’s reputation and affect trust and customer experiences (Louw, 2023). Furthermore, the ethical implications of GenAI in content marketing include privacy concerns, data protection, transparency and fairness, algorithmic bias, and the possibility of manipulation or disinformation (Coltri, 2024; Mao, 2023).

Table 2 summarizes the most common ethical concerns in the literature, including those of practitioners, regarding GenAI for brand content creation. The literature beyond those sources listed did not add new insights.

A typical QCA study comprises 10 to 50 cases, although several studies have used more cases (Hanckel et al., 2021). Cases can be selected from either primary or secondary sources, which for this study comprised secondary sources (Mello, 2021).

Selection bias was addressed by establishing transparent and systematic selection criteria (Ragin, 2000). For adequate case selection, the documents had to be homogeneous with sufficient heterogeneity (Wagemann and Schneider, 2015). To accomplish this, documents were selected that could possibly include or exclude the conditions identified for this study as follows (see Table 4):

To ensure a broader perspective, the guidelines were collected from diverse sectors, including (see also Table 4):

The 2021 UN Resource Guide on AI Strategies Around the World served as a starting point for this study’s case selection (United Nations, 2021). Search engine queries yielded recent AI frameworks and Acts relevant to AI that were published after 2021, reflecting the latest perspectives. Although the documents examined do not represent all AI guidelines published worldwide, they were adequate to represent the eight conditions investigated. It was thus necessary to include enough documents to adequately represent the eight conditions being studied, as depicted in Table 3 (Ragin, 2000).

To become acquainted with the cases, the NVivo Plus software program was used. NVivo is a powerful qualitative data analysis software that enables researchers to explore, organise, and analyse qualitative data (Tang, 2023). For the purpose of this study, detailed annotations helped in understanding the degree of membership in ethical considerations related to the most common conditions associated with GenAI adoption for brand content creation in content marketing (Mello, 2021 and Table 4).

The insights gained from the literature in Step 1 (see Table 2) guided the selection of the most common conditions associated with the ethics of GenAI adoption for brand content creation in content marketing (Rihoux and Ragin, 2009; Thomann et al., 2022). Conditions in QCA refer to key factors or variables whose presence or absence may influence the outcome and must not exceed eight (Mello, 2021). The eight ethical conditions (transparency, privacy, fairness, accuracy, accountability, compliance, discrimination, and intellectual property) were selected based on their academic relevance and practical pertinence for brand content creation in content marketing. To illustrate, these conditions include ensuring transparency in AI-generated content, protecting user privacy, promoting fairness by preventing bias, maintaining accuracy, holding content marketers accountable, ensuring compliance with legal standards, preventing discrimination, and protecting intellectual property rights.

These conditions are also frequently cited in AI ethics literature (see Table 2), demonstrating wide agreement on their importance in guiding ethical AI use which can be extended to brand content in content marketing. Also, these conditions align with deontological ethics, emphasizing duties like protecting human dignity and ensuring fairness that are crucial to ethical GenAI practices. These conditions thus address crucial ethical considerations that encourage the responsible adoption and decision-making of GenAI for brand content in content marketing.

Because of their multidimensional nature, the conditions identified in this study were examined through the lens of personal integrity literature and deontological theory, as shown in Table 3. It is important to note that these conditions are interdependent and should not be viewed in isolation.

During step 4, membership scores were assigned via a process known as calibration. For this study, indirect calibration to assign membership scores for a data matrix was compiled (see Tables 5 and 4).

Indirect calibration relied on the researcher’s broad groupings of cases according to their degree of membership based on case knowledge (Ragin, 2008). Guided by the work of Rihoux and Ragin (2009), and due to the complexity and interconnection of ethics, a four-value approach to assigning membership scores was adopted.

The membership scores for this study were assigned as follows:

1: full membership (a case fully includes the condition).

0.67: more in than out (a case mostly includes the condition but not entirely).

0.33: More out than in (a case mostly lacks the condition but contains some dimensions).

0: full non-membership (a case does not include the condition at all).

However, it is acknowledged that the indirect calibration process inherently involves some degree of subjectivity, particularly when it comes to assigning membership scores. Thus, to minimise subjectivity, the calibration process relied on clearly defined thresholds for each score (the four-value approach) and transparent documentation (annotations). These thresholds draw upon QCA literature (Rihoux and Ragin, 2009) while the scores allocated were derived from a combination of case knowledge and empirical observations (Ragin, 2008; Rihoux and Ragin, 2009).

Table 5 depicts the thresholds for calibration of the conditions. For transparency, Table 5 was developed using a coding scheme derived from NVivo annotations, which identified the membership of each ethical condition across the 33 documents. To illustrate, a full membership score of 1 for IP was only assigned if the document clearly mentioned enforceable IP protections or referenced legal instruments (for example, World Intellectual Property Organization or national copyright laws), or provided strategies for infringement prevention. However, documents that lacked such recommendations or treated IP in vague terms were given lower scores.

In this study, the independent variables were transparency (TP), privacy (Priv), fairness (Fair), accuracy (Accur), accountability (Account), compliance (Compl), discrimination (Discr), and intellectual property (IP), while the dependent variable, representing the outcome of interest, was precedence (Prec).

The calibrated data matrix in the table below shows the values of each independent variable and their corresponding outcomes for 33 cases. Using a fuzzy set theory approach, the outcome score is based on the weakest link in the dataset (Kacprzyk, 2023) and represented as a binary crisp number (Dusa, 2020).

The calibrated data matrix was then imported into the fsQCA software (version 4.1) for further analysis. The software is intended to make it easier to apply QCA through the use of fuzzy set theory. The descriptive statistics for the dataset are as follows:

The findings show consistent high levels of transparency, privacy, fairness, accuracy, accountability, compliance, and discrimination across the cases examined, as well as significant variations in intellectual property (Table 6).

During this step the data matrix was regenerated as a truth table, displaying condition configurations and their effects on the outcome. Truth tables list all logical combinations of the conditions under consideration. Each combination was compared to the empirical data to establish if it caused the outcome. Providing evidence for a combination was sufficient for the outcome (Mello, 2021; Wagemann and Schneider, 2015). The truth table was then minimized while retaining essential data. The minimized truth table includes only the essential configurations that cover all outcomes, simplifying the representation (Ragin and Davey, 2022).

Using the Quine-McCluskey algorithm, the Table 7 depicts the minimized truth table with a frequency of 1 and a consistency threshold of 0.8 (Goertz, 2017).

The minimized truth table analysis revealed two main configurations: one where all conditions, including intellectual property, are present, leading to precedence in 26 cases; and another where all conditions except intellectual property are present (for 7 cases), resulting in the absence of precedence. This variation highlights the essential role of intellectual property in determining precedence.

This step examined how much each condition is needed for the outcome of interest by comparing cases and determining how it affected the outcome (Schneider and Wagemann, 2012).

QCA emphasizes consistency and coverage. Consistency measures how steadily a set of conditions causes the outcome across cases, while coverage measures how well the identified conditions account for all cases with the outcome (Mello, 2021; Schneider and Wagemann, 2012). Exploring the complexity of causal relationships within the dataset by focusing on the overarching patterns and configurations of the outcome was necessary to find the necessary conditions with which a subset/superset analysis helped. Subset/superset analysis tested the sufficiency of a condition or any combination of conditions, meaning that their presence was enough to produce the outcome even if they were not present in every case (Ragin, 2008; Schneider and Wagemann, 2012).

Table 8 highlights the conditions or combinations of conditions that are crucial for the outcome, precedence. It is evident that all the conditions tested individually show high consistency and raw coverage, implying they are all crucial and applicable to a broad range of cases in determining the outcome, precedence. Interestingly, the condition of intellectual property stands out with a high consistency score of 0.971342, but when its contribution to the outcome is not considered, the consistency score drops to 0.584025 (TP* Priv* Fair* Accur *Account*Compl* Discr ~ IP).

When all of these conditions are taken into account (TP* Priv* Fair* Accur* Account* Compl* Discr, *IP), the consistency score is the highest, indicating that these conditions have a significant effect on precedence.

Step 7 involved an analysis of sufficiency (how much each condition is required for the outcome to occur). The conditions that must be present (or absent) for the outcome to occur consistently across cases were highlighted. The complete range of sufficiency solutions is presented in terms of the necessary, complex, the parsimonious and intermediate solutions (Mello, 2021; Schneider and Wagemann, 2012).

The complex solution identifies a set of conditions that, when present together, consistently lead to the occurrence of the outcome, which for this study represent the eight conditions measured (Schneider and Wagemann, 2012).

During this analysis, no counterfactuals were considered.

The complex solution considers the necessity of the condition intellectual property which is a counterfactual for the outcome (Table 9).

The findings of the parsimonious solution highlight the essential role of intellectual property for the outcome, emphasising its relevance despite being a “difficult counterfactual” (Schneider and Wagemann, 2012; Table 10).

The intermediate solution, which strikes a balance between the complex and parsimonious solutions, shows that all eight conditions must be met to achieve the outcome (Ragin, 2000). Table 11 now summarises all solution types for this study.

While considering the views of Schneider and Wagemann (2012:278), the intermediate solution is presented for further discussion.

According to the findings in Table 11, while all eight conditions are required for ensuring responsible GenAI adoption, intellectual property stands out as a prerequisite for precedence. Without considering intellectual property, the outcome is significantly lower, demonstrating its importance in protecting creative ownership and avoiding legal consequences, despite the fact that general AI guidelines frequently fail to include it.

In contrast, other conditions (such as transparency and privacy) were consistently applied across cases, indicating that they are important but less likely to change than intellectual property. Thus, to maintain brand integrity and avoid reputational damage, content marketers must respect intellectual property rights in addition to other ethical considerations.

Because GenAI ethical guidelines for content marketers are currently lacking, this study focuses on the unique ethical challenges that arise from integrating GenAI into content marketing, distinguishing it from broader discussions on AI ethics.

The method used in this study also adds to the literature by demonstrating how different conditions interact to support responsible GenAI adoption, resulting in a multi-condition analysis that can be used to guide future research in AI ethics and content marketing.

QCA confirmed the importance of ethical decision-making in GenAI for brand content creation in content marketing, implying that both the nature of ethical actions and complex decision-making are crucial. This demonstrates the importance of ethical guidelines that emphazise moral responsibilities in a fast-changing industry.

This study bridges the gap between general AI guidelines and GenAI’s specific ethical challenges in content marketing. The findings highlight inconsistencies between general AI ethical recommendations and real-world ethical decision-making in content marketing, particularly intellectual property risks that can affect brand reputation.

The findings add to the ongoing AI ethics debate by addressing the ethical concerns unique to the specialised area of using GenAI for brand content creation in content marketing.

The study will also stimulate additional academic debate and inform future research on the ethical use of GenAI for brand content creation in content marketing.

The findings of this study have significance for marketing professionals involved in GenAI for brand content generation in that they can help brands implement responsible use and ethical decision-making in their content marketing strategies.

Given the eight ethical conditions measured, the findings may assist businesses to use GenAI responsibly for brand content creation and brand reputation. Intellectual property was shown to be essential for brand content creation in content marketing in order to protect brand reputation.

Businesses can use the findings of this study to develop internal guidelines for responsible AI brand content generation.

Businesses must monitor compliance with legal and ethical standards to ensure that AI-generated brand content complies with regulatory requirements.

Guidelines for GenAI ethics for creating brand content in content marketing are now proposed (Table 12).