Diversity, equity and inclusion in EU public communication. Between policy mainstreaming and communicative clashes

1 Introduction

European policies and measures on diversity, equity and inclusion (DEI) emphasize public communication as a strategic tool for promoting equality and well-being through balanced media representation and open, respectful discourse (D’Ambrosi and Parito, 2026). The founding treaties of the European Union, including the Treaty on European Union (TEU) and the Treaty on the Functioning of the European Union (TFEU), as well as the Charter of Fundamental Rights of the European Union – and Article 21 in particular – form a cornerstone for the promotion and protection of diversity. This framing is itself rooted in the Union's own motto United in diversity, which has historically cast diversity primarily in cultural, national and linguistic terms, encompassing many different traditions across the European Union.

From this perspective, it is therefore important to adopt an approach aimed at effectively promoting European citizenship, including through substantial changes to communication strategies and established practices (Isin and Nielsen, 2008; Parito and Pasciuto, 2025). It should be noted, in fact, that the exercise of citizenship is not equally accessible to all members of society: more vulnerable groups, such as migrant women and refugees, often face greater obstacles in making their voices heard and having their demands recognized (Blaagaard and Ponzanesi, 2024). In response to these challenges, the commitment that European Institutions have called upon to implement together cannot be read merely as a value-based choice but must translate into a concrete evolution of policies to promote effective equality among individuals and the protection of rights (European Commission, 2025a). Active inclusion means enabling every citizen, particularly the most disadvantaged, to participate fully in society and to concretely exercise their rights.

Against this backdrop, the European Institutions have pursued several strategic approaches to strengthen communication within their organizations, adopting a holistic and intersectional perspective sensitive to the ways in which multiple factors of discrimination may interact (Council of the EU, 2018; EIGE, 2024; European Parliament, 2019). A key issue has concerned fundamental rights and freedoms across several domains (women, the LGBTQI+ community, people with disabilities, migrants), for which European Institutions have sought to adopt an inclusive stance, valuing differences across their various institutional communication channels, albeit with uneven results that have often been more declaratory than structurally transformative. While this process has accelerated, on the one hand, certain social changes by placing DEI policies at the centre of attention, it has on the other hand, exposed European Institutions to specific communication risks, frequently manifest in the gap between stated political ambition and its translation into day-to-day professional practice, a problem further complicated by the cultural and political specificities of the Member States addressed (Ducci et al., 2025). This risk is compounded by the growing use of digital technologies and artificial intelligence in the public communication sector, which can replicate or amplify gender stereotypes in the representation of different audiences. At the same time, these tools open new possibilities for more inclusive communication practices, from neutralising gendered language across multilingual contexts to improving the accessibility of institutional content for diverse publics.

This paper examines the principles of diversity, equity and inclusion in institutional communication, considering both the organization of communication work — structures, responsibilities, clearance — and the content it produces. The aim is to understand how DEI policies are implemented at European level and what challenges arise in the representation of different audiences, highlighting the tensions and ambivalences that can characterize European communication policies and practices. Building on these premises, the paper advances two related arguments. First, that the robustness of the EU's DEI communication architecture coexists with a striking dependence on individual discretion, a condition we term entrusted inclusion: inclusive practice is upheld day to day by the commitment of particular professionals instead of being secured by roles, procedures and evaluation metrics, and is therefore vulnerable to changes in personnel and leadership. Second, the gap between declared intent and daily practice generates communicative clashes, moments at which communication designed to include collides with its own production, representational or reception dynamics, producing effects — tokenism, diversity washing, algorithmic bias — that run counter to the original aim. These two concepts organize the analysis that follows: after the introduction, Section 2 reviews the literature on DEI in public communication and public relations, situating the concept within debates on gender and intersectionality. Section 3 outlines the study's aims, research questions and methodological design. Section 4 presents the findings in two parts: a policy analysis tracing the evolution of EU DEI communication frameworks since 2018, and a thematic analysis of the interviews. Section 5 discusses these findings in relation to the research questions, and Section 6 concludes by considering their implications for DEI communication research and practice.

2 Conceptualising DEI in public communication studies

Diversity is widely recognized in the literature as a multidimensional concept encompassing, among others, race, ethnicity, gender and disability (Cox, 1993; Loden and Rosener, 1991). Diversity is closely intertwined with equity and inclusion, and recent conceptualizations have increasingly treated them as a unified construct, summarized under the acronym DEI (Minow, 2021; Ravazzani et al., 2025). This conceptualization responds to the major social and cultural transformations which, in recent years, have brought inequalities into sharper relief, but have also revealed new and persistent vulnerabilities that deserve deeper scrutiny, including the deepening of forms of exclusion, the widening of socio-economic gaps, the intensification of conflicts and crises, and the rise in environmental risks. This requires reading diversity policies differently, not only in terms of individual characteristics, but also in terms of equitable treatment, ensuring that every individual feels valued and respected.

Public communication studies have addressed inclusion along different lines of research. Several studies approached diversity as a privileged analytical lens through which inclusion has been examined. Studies in this area focus on gender roles, professional representation and stereotyped language alongside continued prejudices limiting women's access to leadership positions (D’Ambrosi et al., 2024; Topić et al., 2020; Vardeman-Winter and Place, 2017). Others focus on an intersectional perspective, considering multiple factors of discrimination, in studying the role and impact of identity and difference in public relations (Ciszek, 2020; Mundy and Bardhan, 2023). Other studies have instead focused on the broader institutional mandate as an ethical mandate, framing public sector communication as an instrument for closing the gap between citizens and public organizations (Bowen and Erzikova, 2025; Canel and Luoma-aho, 2019; Oliveira and Gonçalves, 2024). Grunig (2023) conceptualizes this institutional role through the lens of social inclusion, understanding public relations as a means for individuals and groups to participate meaningfully in society.

Within this context, public communication in Europe is studied as an essential function for promoting the principle of equality, providing a basis for a vision of integration based on the recognition of differences and on various aspects of social justice (Valentini, 2019; D'Ambrosi and Parito, 2026). It is further recognized as a strategic tool for fostering an inclusive society, increasing trust in institutions and ensuring that citizens can participate in public life, while safeguarding and strengthening democratic principles (Lovari and Belluati, 2023; OECD, 2022; Rocha Menocal, 2021).

A second strand of research translates this intersectional approach into organizational and professional competences, examining what it takes for DEI to move from declared intent to structural practice. DEI policies need to become a professional competence integrated into communication processes, instead of being reduced to episodic or symbolic initiatives (Basaglia et al., 2022; Kim et al., 2025; Ravazzani et al., 2023; Sebastião and Cotton, 2025). Such structural integration is not, however, uniform across settings: the DEI approach is related to the specific socio- cultural and organizational context in which it is implemented (Ravazzani, 2016). Several factors can impact on these processes. Some studies point out the employees' linguistic and cultural backgrounds and the need to have specific language policies or guidelines suggesting the appropriate approach to adopt with different audiences (Gunnarsson, 2010; Lovrits, 2022). This competence is particularly tested by the growing role of digital technologies and AI which can simultaneously open new possibilities for inclusive communication and replicate the exclusions that DEI policies seek to address: recent research focusing on text to image generation shows that these tools can replicate entrenched gender and socioeconomic stereotypes in the representation of social roles, requiring public relations professionals to take an active role in identifying and mitigating such biases (Luttrell et al., 2025). How diversity is represented in the media depends fundamentally on language choices and professional practices embedded within journalism routines and organizational contexts (Moody-Ramirez et al., 2023; Peruzzi et al., 2024; Peruzzi and Volterrani, 2016).

In the context of public sector communication, this shift demands new competencies, ethical vigilance, and a renewed sense of responsibility toward citizens in order to counter risks such as disinformation and the erosion of institutional trust (EIGE, 2025; Lovari and De Rosa, 2025). Building such competences cannot be left to individual sensitivity: they must be embedded from the earliest stages of the professional training, in pedagogical models that combine creative, dialogical and intersectional approaches to inclusive practice (De Sá, 2026; Tsetsura, 2011).

A further line of research provides empirical evidence in support of this argument, showing how DEI practices in recruitment and assessment bring about measurable and positive change, particularly when underpinned by a genuine moral commitment (Kalev et al., 2006; Ravazzani et al., 2025), and exploring how, once embedded in organizational culture, DEI redefines professional ethics and fosters organizational cultural change, going beyond mere communicative outputs (Ducci et al., 2025; Mundy, 2016; Wolfgruber et al., 2021).

These strands of research share an underlying assumption: that communication is an instrument through which DEI policies reach their intended audiences. Yet taken together, they also point to a more constitutive function. The decision to co-design a campaign with affected communities, as distinct from producing it on their behalf, is not simply a methodological choice: it is an act through which the boundaries and meaning of inclusion are negotiated and defined. More broadly, when communication professionals advocate internally for a message that risks being diluted as it moves through institutional approval processes, communication is no longer merely downstream of policy — it becomes the site at which policy is made and contested. This generative dimension has been recognized across communication studies and policy analysis: institutional practices and communicative acts do not merely transmit pre-existing values but actively participate in constructing them (Council of Europe, 2024; Isin and Nielsen, 2008; Rocha Menocal, 2021). In the EU context, where DEI commitments rest on soft law instead of binding treaty obligations, this constitutive function carries particular weight: the gap between declared policy and implemented practice is closed or not by the daily choices of those who produce institutional communication (Belluati, 2026; D’Ambrosi and Parito, 2026).

2.1 From communicative clashes to communicative clashing

Despite this growing body of literature, a gap remains concerning how communication professionals interpret, negotiate and enact DEI policies, and concerning the effects that open a distance between declared institutional intent and actual outcomes.

Communicative clashes in the sense proposed here, are the breaking point at which institutional communication aimed at inclusion comes into conflict with its own production processes, forms of representation or reception dynamics, generating effects that are unforeseen or even contrary to the initial intention — such as tokenism, diversity washing, intersectional algorithmic bias, over-blocking in automated moderation, and forms of discursive marginalization. The concept is inductive, developed from the corpus of this research, but it resonates with excellence theory's account of communication as a two-way, symmetrical function (Grunig, 2008): a clash is precisely the moment at which that symmetrical ideal breaks down under political, procedural and platform pressures, so that the institution communicates at rather than with its publics. Read this way, clashes are not communication failures in a technical sense but structural tensions between inclusive intent and the organizational and technological conditions of its production.

If a communicative clash names the outcome, communicative clashing names the process: the ongoing collision of symmetrical, inclusive or dialogic intent with the conditions of production, control and mediation through which messages are made and circulated — a process that can turn inclusion into its opposite, whether asymmetry, tokenism or exclusion. So defined, the concept is analytical: it locates the failure of inclusive communication not in individual error or bad faith but in the routine points — clearance chains, automated systems, platform metrics — at which intent is filtered, diluted or reversed. In this sense, communicative clashing is the counterpart of entrusted inclusion: where the latter describes how inclusive practice is held up by the commitment of individuals, the former describes how, and where, that commitment is undone. To specify what is meant by the organizational conditions of production, it is useful to read communicative clashing through the communicative constitution of organizations. In McPhee and Zaug's (2000) structurational account, organizations are constituted through distinct flows of communication, and structural constraints allow locally worked-out resolutions to enter the organization's authoritative body of knowledge only in limited ways. McPhee and Zaug build this account on Giddens's (1984) theory of structuration, and it is worth drawing out the tension they inherit from it. For Giddens, structure is both the medium and the outcome of the practices it organizes — the duality of structure — so that agents, in drawing on organizational rules and resources to act, at the same time reproduce those rules and resources. Agency is thus enabled and constrained by one and the same structure. Applied to inclusive communication, this yields a pointed irony: the very acts through which professionals pursue inclusion are also acts through which existing arrangements — the clearance hierarchy, the gendered division of labour, the metrics of platform visibility — are reproduced. Inclusive intent does not stand outside structure; it is exercised through it, and can therefore reinstate, as an unintended consequence, the asymmetries it sets out to challenge.

The implication for inclusive communication is direct. An inclusive choice made locally, a drafter's decision to use non-marked language, to show real people and not merely generated images, or to consult an affected community, is not in itself the institution's position; it is one professional's. For that choice to become the organization's authoritative output, it must pass through the flows that constitute the organization itself: the negotiation of membership, the coordination of activity across units, the self-structuring of the organization's own rules and hierarchies, and the institutional positioning of the text towards other bodies, Member States and publics. It is in that passage — and above all in self-structuring, where the clearance chain resides — that inclusive intent is filtered, reweighted or lost. The failure is therefore structural: not a lapse of individual care but a property of the very flows through which a text comes to speak for the institution.

Both concepts need to be distinguished from adjacent formulations. Entrusted inclusion is not simply weak institutionalization, nor a restatement of policy–practice decoupling or organizational hypocrisy. Decoupling describes a structural relation between formal commitment and operational core in which talk and action are deliberately uncoupled to absorb external pressure. Entrusted inclusion describes something the opposite in its mechanism: commitment is not decoupled from practice but devolved to it, carried by named individuals who supply, at their own discretion, the accountability the formal architecture does not allocate. The policy is not hypocritical; it is unassigned. What follows analytically is a specific vulnerability — not the gap between talk and action, but the dependence of action on personnel — and a specific remedy, namely the allocation of inclusive responsibility to roles, procedures and evaluation criteria, not just the strengthening of declarations.

Communicative clashing likewise risks becoming a label for any divergence between intent and outcome unless it is located in a sequence. We therefore situate clashes at identifiable stages of the production chain: allocation of responsibility, drafting, clearance, representational choice, dissemination and reception. This structural location is reinforced by the material conditions of production. As Cuel and Ferrario argue, information systems are not neutral organizational assets but are bound up with pre-existing coordination processes and modes of production; the tools through which institutional communication is drafted, cleared and circulated therefore carry the coordination logics, and the exclusions, already inscribed in them (Cuel and Ferrario, 2008). Kaplún's account of organizational communication as a field structured by tensions between fragmented, instrumental communication and an integrated communication oriented to structures, processes, identities and collective action names the same fault line from the other side: inclusion is an integrative aspiration pursued through instruments that are frequently fragmentary and control-oriented (Kaplún, 2012).

Read alongside the public relations literature, this is a sharper, production-centred version of a familiar problem. Macnamara identifies a persistent theory–practice gap in organization–public communication, arguing that organizations do not merely fail to communicate but produce communication in ways that foreclose the very relationships they claim to build (Macnamara, 2016). Stoker shows how the ideals of symmetry and relationship management can themselves generate paradoxes and double-bind commitments that practitioners cannot honestly deliver (Stoker, 2014). Grunig conceives of public relations as a means of social inclusion while also noting that digital media can operate as vehicles of exclusion as readily as inclusion; in that sense, communicative clashing marks the breakdown of the symmetrical, inclusive ideal under organizational and platform pressure (Duhé and Wright, 2013; Grunig, 2023; Theunissen and Rahman, 2011). Communicative clashing thus names, in a single term, what these traditions approach separately: the structural and material process through which inclusive intent is undone at the point of production.

In the communication of public institutions, and in EU institutions in particular, inclusive commitments are enacted through clearance procedures, multilingual workflows and platform-dependent systems established for other purposes, and it is in this mediation that their intended effects are reduced or reversed.

3 Aims and methods

The study examined how European Institutions incorporate the principles of diversity, equity and inclusion into their public communication, comparing stated policies with the professional practices of those involved in communication.

The research was guided by two core research questions: RQ1- How do EU professionals (communication and policy staff) interpret and apply DEI policies in their day-to-day communication practices aimed at different audiences? RQ2- How is this approach perceived by public communicators operating within central EU institutions, and what alignments, ambivalences or communicative clashes emerge in relation to DEI policies?

Our hypothesis was built on previous studies on the risks of inclusive communication in the public sector (Ducci et al., 2025; Spalletta et al., 2024), which suggest that gender-related issues are particularly sensitive and politically contested and institutions struggle to translate their policy commitments on gender equality into effective and truly inclusive communication practices. This research aimed, in particular, to examine the complex relationship between regulatory objectives, organizational resistance and possible communicative clashes.

This study employed a qualitative approach, combining documentary analysis, policy review and interviews within a public-sector context. The first phase consists of desk research on European Union strategies and documents concerning diversity, equity and inclusion in the period 2018–2026, with particular attention to their implications for institutional communication. Strategies, action plans and guidelines were retained when they specified how institutional messages should be worded, illustrated or made accessible, or when they assigned responsibility for those choices.

The aim was to reconstruct the framework within which European Institutions define inclusion, diversity, equality, accessibility, anti-discrimination, participation and gender mainstreaming. This phase served to build the analytical framework of the research and to define the categories guiding the subsequent interviews. A corpus of 14 documents was selected against three criteria: issued or revised between 2018 and 2026; adopted by an EU institution or a European body with direct normative influence on EU communication practice; and containing explicit prescriptions for communicative output — language, terminology, imagery, formats, accessibility or representation. On the third criterion the AI Act [European Parliament and Council of the European Union (EU) 2024/1689] and the Code of Conduct on countering illegal hate speech online plus, integrated into the Digital Services Act framework in 2025 (European Commission, 2025c), fall outside the corpus. We note them here because they form the regulatory horizon against which interviewees situate their internal AI rules, and because the Code of Conduct–DSA nexus governs the platform moderation on which institutions depend.

The second phase involved 20 semi-structured interviews conducted in 2026. These interviews explored how DEI policies are implemented across communication strategies within European institutions at the central level—specifically the European Commission, the European Parliament and the Council of the European Union. Participants included public communicators, policy officials, social media managers and leadership figures directly involved in translating DEI commitments into institutional practice (see Table 1). The sample focuses on the institutional core of European public communication (Lovari and Belluati, 2023) and includes, alongside staff of the three institutions, external practitioners such as consultants and agency professionals who work on EU communication under contract. The choice to focus on these three institutions reflects their distinct political principles, legal mandates and communication functions, which together shape how DEI commitments are translated into institutional practice. Participants were identified through purposive sampling based on role and institution, and subject to availability for interview. Eligibility was established through desk research on publicly available organigrams and institutional directories, prioritizing roles in communication, press, campaign and DEI functions; recruitment proceeded by email and, in some cases, through direct contact at institutional events and through chain referral, as interviewees suggested colleagues in other divisions (Noy, 2008; Patton, 2015). Selection sought maximum differentiation across institutions, role bands and areas of specialization. Approximately fifty approaches yielded twenty interviews. Recruitment and interviewing ran in parallel, so that sample composition could be reviewed as fieldwork progressed and subsequent invitations directed towards under-represented profiles. Invitations referred broadly to online communication and diversity without specifying the study's focus on DEI practice, in order to limit priming at the point of consent.

Table 1. Profile of the interviews.

This procedure nonetheless carries a self-selection risk. Professionals who agree to take part in a study touching on DEI communication are plausibly among those most invested in it, and a sample so composed may tend towards a picture in which inclusion is sustained by committed individuals. We therefore treat entrusted inclusion as a condition observed among professionals already engaged with this work and not as a characteristic of EU communication staff generally, and the broad framing of invitations should be understood as mitigating this risk, though not eliminating it.

Overall, 14 women and 6 men were interviewed. This pattern is consistent with the composition of our sample and with interviewees' accounts of their units, and it echoes a documented gendering of career pathways in the European institutions (Alayrac et al., 2025). The asymmetry is horizontal: even where women have reached management, inclusion and communication work remains concentrated among them, in roles with more limited strategic authority.

The 20 participants are working across the principal institutions of the European Union and cooperating bodies. The Commission accounts for the largest share (n = 8), followed by the Council of the EU and the European Parliament (n = 5 each); the remaining two participants are based at an EU advisory body and an external communication provider respectively. The sample is predominantly female (14 of 20). Participants span a range of seniority, from management-level posts (heads of unit and senior managers) to policy, press, and digital officers, campaign and communication managers, and advisory roles. Substantively, their responsibilities concentrate on communication, press, and campaign functions, with further coverage of DEI portfolios. The sample therefore combines breadth of institutional coverage with a thematic focus on communication-related roles.

The distribution of interviewees does not proportionally reflect the numerical weight of the institutions considered, since the selection did not aim at statistical representativeness but at gathering diversified and information-rich perspectives relevant to the research aims. The final composition of the sample also reflects the different opportunities of access to interviewees and the availability of the actors involved. In line with the logic of reflexive thematic analysis, the sample size was not governed by the principle of data saturation, whose appropriateness for this analytical approach has been questioned by Braun and Clarke (2021); rather, the number of interviews was determined by the situated nature of the phenomenon under study and by the information richness of the accounts gathered.

To protect participant confidentiality, all directly and indirectly identifying information was removed or generalized prior to analysis and reporting. Each participant was assigned a sequential code, renumbered to break any link to recruitment or interview order. Quasi-identifiers were generalized to reduce re-identification risk: directorate-general codes and unit-level titles were removed, institutions were grouped into broad categories, and roles were collapsed into aggregate bands (management versus officer/staff). Categories represented by a single participant were merged into broader groupings to avoid unique records. The key linking pseudonymous codes to participant identities was stored separately from the research dataset under restricted access.

This article reports the first phase of a broader research project and investigates the phenomenon from within, that is, from the standpoint of the communication professionals directly involved. A more in-depth analysis of communication content is envisaged for the subsequent developments of the research, possibly complemented by the testimonies of a sample of recipients.

Interviews took place over a four-month period (March–June 2026), conducted as virtual sessions on video-conferencing platforms. The interview material was analyzed through Reflexive Thematic Analysis, following Braun and Clarke's approach (2006, 2023).

The process began with repeated reading of the interview transcripts in order to become familiar with the corpus and identify initial patterns of meaning. Analysis followed a hybrid, deductive–inductive logic. The four macro-areas derived from the documentary phase were used as sensitizing categories that oriented initial reading, in keeping with the study's overall design; coding itself, however, proceeded inductively and reflexively, in line with Braun and Clarke's approach. Coding was conducted manually, without CAQDAS software, given the limited yet information-rich corpus. The macro-areas were divided among team members, each performing the reflexive thematic analysis of their portion before the team met to discuss interpretations; interpretive rigour was pursued through reflexivity and collegial discussion to effectively control for any interpretive biases (Yardley, 2000).

The a priori areas were treated as an entry point, not as a coding frame, remaining open to patterns of meaning that cut across or exceeded them, and attending throughout to tensions, contradictions and implicit assumptions in the professionals' accounts. The themes reported below were generated from the data and the macro-areas were revised where the corpus did not support them.

The macro-areas that guided the analysis are the following: DEI as a multilevel approach explores the relationship between institutional governance structures and the translation of DEI policies into communication practices; Inclusion as a shared process examines the role of participatory and co-design dynamics in institutional communication; DEI as organizational culture addresses the internal conditions, in terms of competences, team composition and procedural frameworks, that shape how DEI is embedded in professional routines; Non-delegable responsibility focuses on the use of digital technologies and platforms in institutional communication, and on the forms of human oversight they require.

4 Results

4.1 Policy analysis

The desk research on European Union strategies and documents concerning diversity, equity and inclusion covers the period 2018–2026, with a view to examining the evolution of policy preceding the COVID-19 pandemic and taking into account the political, social and technological transformations that have affected the European public sphere. The corpus comprises documents issued by the European Parliament, the Council of the European Union, the European Commission and European Institute for Gender Equality (EIGE). The desk research reveals three phases of institutional evolution, each of which advances the regulatory framework whilst exposing its own structural limitations (see Table 2).

Table 2. Overview of the EU policy documents analysed.

Source: adapted from D’Ambrosi and Parito (2026).

The first phase (2018–2019) established gender-sensitive language as an explicit institutional commitment, introducing guidelines on gender-neutral terminology and the countering of stereotypes and sexism (EIGE, 2019; European Parliament, 2018a). This represented a significant normative step, though one that remained largely prescriptive and text-centred, with limited attention to organizational conditions or reception dynamics (EIGE, 2019; European Parliament, 2018b).

The policy on DEI does not appear in the foundational documents of 2018–2019, which operate within the established framework of “gender equality” and “non-discrimination”. It is only with the “Union of Equality” agenda launched by the von der Leyen Commission in March 2020 that diversity, equity and inclusion are formally articulated as an integrated policy framework, explicitly combining gender mainstreaming with intersectionality as a horizontal principle (European Commission, 2020a). In this second phase (2020–2025) the concept of intersectionality is introduced and recognized the accessibility of content as a right, including inclusive representation in narratives and audiovisual content (EIGE, 2024). The concept of DEI has since become one of the most contested terrains in European public debate, actively politicized by far-right movements that frame diversity initiatives as anti-meritocratic and ideologically driven.

The current cycle (2026–2030) marks a more structural shift, building explicitly on the achievements of the previous 2020–2025 strategies. The Gender Equality Strategy 2026–2030, upholds the approach of its predecessor by maintaining an intersectional perspective and combining gender mainstreaming with targeted actions, whilst enlarging its scope to cover all principles of the Roadmap for Women's Rights endorsed by all 27 Member States in 2025 (European Commission, 2025a). DEI is being integrated into organizational processes, evaluation mechanisms and professional training, whilst artificial intelligence is explicitly identified as a vehicle for gender bias requiring mandatory impact assessments (EU Strategy for Gender Equality 2026–2030; EU Strategy against Racism 2026–2030; LGBTIQ+ Equality Strategy 2026–2030). Integration therefore remains a concept fraught with tensions: it signals a transformation of the conditions under which public communication is produced, but risks reproducing the very exclusions it purports to address, unless it is accompanied by genuinely intersectional implementation requiring capacity building and systematic evaluation (Table 3).

Table 3. Policy analysis: from gender-sensitive communication to DEI mainstreaming.

As recently observed, the EU's approach to DEI risks amounting to performative allyship, paying lip-service to equality commitments without delivering the tangible policies, accountability mechanisms and measurable targets needed to enforce them (Pőcze, 2026). What mainstreaming presupposes, in other words, is not only political ambition but enforceable obligations at Member State level.

Despite an increasingly articulated policy framework, the documents analyzed appear not to address how institutional communication is actually produced, in terms of clearance procedures and the tools employed. Guidance on terminology continues to be issued alongside AI risks and concerns about information disorder, without bearing significantly on the conditions under which institutional communication is made, leaving open the tensions and clashes.

4.2 Findings from the interviews

The thematic analysis of the interviews identifies four areas that trace this complexity across the dimensions: DEI as a multilevel approach in decision -making process, inclusion in institutional communication as a shared process, DEI as organizational culture, AI risks and the non-delegable responsibility.

4.2.1 DEI in decision-making process

European Union Institutions have, over the past decades, built a substantial architecture of DEI policies through the 2020 Union of Equality agenda that formally articulated diversity, equity and inclusion as an integrated, intersectional framework. The strength of this architecture is visible in several simple, widespread signals: gender-sensitive language guidelines adopted across institutions, a network of diversity charters present in every Member State, a dedicated DEI office within human resources, and thematic months and awards that have since become institutionalized. This formal architecture, however, does not always translate into lived practice on the ground with the same solidity: the interviews reveal that, even where offices, networks and structures formally exist, DEI is often felt to function only because individual officials, units and networks choose, of their own accord, to make it work.

This tension is captured with striking clarity by one official at the Council of the European Union, who describes DEI as a “virtual policy” (id. 9): present in daily practice, driven by genuine commitment, yet lacking explicit legal grounding in the Treaties. What matters is that this steady accumulation of soft-law instruments such as guidelines, strategies, roadmaps, has never crystallized into the kind of binding obligation that Article 21 of the Charter of Fundamental Rights would, in principle, seem to demand.

Far from representing a weakness, this gap reveals something more interesting, a field sustained by conviction rather than obligation, and all the more fragile, and remarkable, for it. This dynamic resonates with recent work on the non-performativity of diversity commitments, whereby formal statements and structures can coexist with an absence of substantive change and come to depend on the discretionary effort of committed individuals (Van Douwen et al., 2025).

This reliance on individual commitment is, in the eyes of several interviewees, itself a source of difficulty, insofar as roles dedicated to DEI risk simply mirroring the organization's pre-existing gendered patterns instead of challenging them.

A first structural feature that stands out across the interviews is a pronounced horizontal segregation in the division of labour. There is, moreover, a structural paradox worth naming explicitly: the very field dedicated to overcoming exclusion and stereotype has itself become predominantly female. DEI work risks becoming a form of reverse ghettoization — a domain where women's presumed affinity with care, inclusion and communication is quietly reproduced, not questioned. What begins as an openness to diversity ends up reinforcing the very gendered divisions that DEI policies are meant to dismantle. This pattern appears both in general communication units and, more acutely, in DEI teams.

In my sector (publishing)… there are about 12 of us… and there's just one man and the rest of us are all women […] The rest of the unit deals with technical aspects related to the website, and, um, they're all men, so there's no one else — there's no woman in the technical team (id. 14).

In my unit, there are 20 women and one man […] the number of women with expertise in the subject is ten times greater than that of men […] In fact, when we find a man, we celebrate, because we'd like to achieve gender equality in our unit, but it's really very difficult (id. 8).

A second tension concerns the approval processes for messages on politically sensitive issues. The more sensitive the topic — LGBTQI+ rights, the fight against racism, disability — the further the validation process moves away from communications professionals and towards management levels that may lack the specific knowledge and cultural familiarity the message requires. As one interviewee describes, the outcome is often a hybrid product, assembled from disparate contributions and not built around a coherent vision:

The more strategically important the message, the more it is passed on to people outside the communications department; consequently, directors who may lack this sensitivity and who do not understand the audience being addressed can, from time to time, turn into a sort of Frankenstein's monster (id. 4).

According to those interviewed, the process designed to safeguard DEI policies through multi-level validation tends to become the mechanism through which inclusive intent is diluted or lost. This escalation is organizationally prescribed: politically strategic files move up the hierarchy and are handled directly by central communication directorates and, where the file is deemed strategic, by the political level. In practice, a vetting process calibrated to political sensitivity, combined with vertical escalation, confines such matters to an internal management arena in which the DEI function acts only as an optional advisor. This pattern is reported most acutely at the European Commission and the Council of the European Union, where the length of the clearance chain and the political weight attached to sensitive files place the final decision furthest from the units that drafted the message. Alongside this, a further erosion occurs as the resulting communication content moves up the hierarchical chain: what feels essential to a technical unit may register as one item among hundreds at higher levels.

If I write a document, for me it's the most important thing in the world […] but for my colleagues at DGCOM who see 300 of them a day, it's obviously worth three-hundredth (id. 3).

These distortions in the validation chain are compounded by a further factor: the professional formation of some of the officials who occupy key roles within it. A significant number began their careers as translators before moving, often organically, into communication positions. This background brings genuine strengths in linguistic precision and cross-cultural awareness where the priority is not merely to translate correctly but to communicate transformatively shaping how audiences perceive an issue or a social group, thereby contributing to broader cultural change. As a DEI coordinator argues:

The funny thing is that as many as 5 out of 9 started their careers as officials, as translators in various languages, but clearly then, over the years […] they found themselves there […] However, when it comes to communication, my unit has never really invested all that much, except in recent years, since I've been here (id. 15).

Another consideration concerns the composition of teams relative to what is produced for external audiences. Whilst gender and LGBTQ+ issues have been progressively mainstreamed, ethnic diversity remains a critical gap. Staff from racial and ethnic minorities are under-represented, especially at senior levels: a structural reality that several interviewees link directly to recruitment mechanisms. As one interviewee stated:

There is clear structural racism, given that people of colour make up a tiny percentage; even though over the last twenty years the figure has gone from zero to one – ten times more than zero – I believe this is the main issue, which reflects a more general problem (id. 20).

Read together, the feminization of the DEI field and the ethnic-diversity gap are not two separate observations but an intersectional one. The work of representing diversity is concentrated among women, yet the senior levels at which representational decisions are validated remain both male- dominated and overwhelmingly white. Inclusion work is thus performed disproportionately by those with least authority over its outcomes, while the axes along which the institutions are least diverse - ethnicity at senior grades - are precisely those least present in the spaces where communication is cleared. The asymmetries that DEI communication addresses outward are, in this sense, reproduced in the composition of the function itself. This uneven distribution of inclusion work echoes intersectional analyses of organizational power, which examine how overlapping identity positions shape who holds authority over diversity outcomes across organizational levels (Thatcher et al., 2023).

The evidence collected highlights some significant differences between the institutions involved in the survey. In Commission accounts, upward escalation is described chiefly as an administrative property of the clearance chain: length, volume and distance from the drafting unit. The difficulty is viewed through the lens of internal administration: the validation chain distances the message from the communicators (id. 4), and the volume of files dilutes its sense of urgency (id. 3).

At the Council, however, the constraint is seen as external, with the DEI viewed as an optional function tailored to the positions of the Member States (id. 9), within an intergovernmental framework. Parliament accounts introduce a party-political register, in which representational choices acquire visibility through the political affiliations of the members they concern — the compulsory anti-discrimination training for MEPs is intelligible only in that register. In particular, horizontal segregation within units (id. 14) and the process from translation to communication (id. 15)—which operates within a political-party framework—prevail. Council accounts, by contrast, locate the constraint outside the institution: sensitivity is calibrated to Member State positions, and DEI is described by one official as a challenge perceived as external to the Secretariat itself. Entrusted inclusion thus appears across all three settings, but it is produced by different mechanisms — bureaucratic dilution, political exposure and intergovernmental caution respectively.

Whether these mechanisms converge on a common condition, or whether the concept mainly describes particular Commission units, is a question our sample size cannot settle and which a design structured for institutional comparison should address.

4.2.2 Inclusion in institutional communication

The picture emerging from the interviews reflects a conception of DEI that moves away from a logic of mere integration or added value to the final message, taking shape instead as an organizational sensitivity that cuts across processes, roles and hierarchical levels. DEI tends to be progressively embedded in structures and practices through specific instruments, namely task forces, networks of DEI coordinators and staff networks, called upon to perform consultative, deliberative and verificatory functions with respect to the choice of messages, images and language.

The shift from a logic of representation to one of involvement finds clear expression in the principle Nothing about us without us, which plays a central role in the communicative practices analyzed. When a campaign or piece of content concerns people with disabilities, minorities or socially disadvantaged groups, the direct involvement of those communities makes it possible to test the appropriateness of tone, imagery and narrative before they reach the public. This is a shared orientation across the institutions studied, though expressed in different forms.

The same principle emerges at the Council of the European Union, where the formula is described as a “disability mantra”, a formula repeated in reference to the EU's disability strategy often enough to signal institutional commitment.

A related shift concerns the framing of DEI itself: away from diversity as a value to be declared, and towards inclusion and belonging as experiences to be built. As one interviewee puts it:

I think we've gone from focusing on diversity to focusing on inclusion and belonging, because we know that European societies are diverse — now it's about building inclusion and belonging into that. So the messaging is slightly changing, to make sure that everybody can recognize themselves in the communication (id. 2).

Inclusive communication is no longer simply about making diverse individuals visible; it is about working with them to craft messages capable of generating genuine recognition and belonging. The shift from mainstreaming to co-design changes not only the content of what is communicated, but the process by which it is produced: who participates, who validates, and through what mechanisms. DEI thus acquires both a narrative and a design dimension, shaping the ways in which content is produced and made accessible to audiences with different needs.

The interviews reveal, however, a meaningful distinction between technical accessibility and effective accessibility. The former is relatively well established at the level of procedures and available tools; the harder challenge lies upstream, with those audiences who cannot even reach the digital channels through which institutional communication is now predominantly distributed:

For people who, for whatever reason, do not have access to digital media […] I think this is something of a weak point in our communication (id. 1).

The more institutional communication concentrates on digital environments, the more it risks leaving behind precisely those citizens with fewer resources, fewer skills, or fewer connectivity opportunities. A second critical dimension concerns the quality of representation. Here, the risk of tokenism — diversity formally present but poorly rooted in context — is widely acknowledged among communications professionals. Some describe it as a display of inclusivity that at times borders on ostentation.

Interviewees express a clear rejection of diversity washing in their accounts, inclusive communication at the surface level is of little value when it finds no correspondence in the organisation's internal culture. In this regard, the practice of pre-testing images and materials through local focus groups, to verify the plausibility of visual and narrative choices against specific cultural contexts, assumes particular importance.

Alongside this growing attention, however, an organizational tension emerges. In some contexts, DEI appears to have consolidated itself as a professional habit, embedded in the everyday practices of communication. One interviewee observes that it has almost become a “kind of habit”, something done as a matter of course, regardless of its actual meaning (id. 14).

Yet the informal nature of this habit also reveals its fragility: if attention to DEI depends on the sensitivity of individual professionals as opposed to formalized procedures and stable verification mechanisms, the continuity of inclusive practices becomes vulnerable to changes in staff or leadership. This tension between individual commitment and institutional embeddedness runs through the training dimension too, as the analysis in the following section shows.

European institutions operate in an increasingly polarized external environment, in which linguistic and representational choices can become the object of controversy and organized backlash. As some European Commission officials note, contemporary societies are markedly polarized, and issues of this kind consistently provoke strong reactions.

The reference is to a concrete case: a campaign on the Roma community, the largest ethnic minority in Europe, designed to overturn dominant stereotypes by showing Roma as active professionals in society, and distributed also in Romani to reach local communities. The campaign generated large numbers of explicitly racist comments in which the object of contestation was not the institutional message but the community itself. The institutional response, however, did not involve withdrawing the content: reactions were instead analyzed, interpreted and managed, with hate speech reported to the platforms. Backlash thus entered the communicative process as a variable to be governed, not as a reason to retreat from inclusive objectives.

The interviews show, however, that external pressure can produce different outcomes. In several cases, guidelines on inclusive communication are described as initiatives that have sometimes stalled under political pressure and controversy, yet without the practices developed disappearing entirely. According to one interviewee, many of the criteria elaborated continue to be applied in day-to-day professional work, even in the absence of the original formal framework. Within the General Secretariat of the Council of the European Union, the task is described as not always straightforward, particularly in relation to certain Member States, and the challenge of DEI tends to be perceived as external to the institution itself (id. 11).

Backlash can thus generate institutional resistance, adaptation or formal retreat, without a single uniform outcome.

More broadly, this pressure shapes communicators' choices in advance. Terminological reformulation, the use of more neutral imagery, and in some cases the use of English as a kind of refuge language are ways in which institutions seek to reduce their exposure to controversy: potentially divisive terms may appear less politically marked when kept in English, whereas their translation into national languages tends to place them more directly within local cultural and political conflicts. According to professionals interviewed, a paradox emerges: precisely while inclusion is recognized as an institutional goal, the fear of backlash may encourage processes of attenuation or neutralization of the message, reducing the visibility of certain differences or depoliticizing particular formulations. Inclusion thus configures itself as an unstable equilibrium between recognition of social plurality and management of the institution's reputational vulnerability.

Superimposed on this external pressure is a second tension, linked to the logics of digital performance measurement. KPIs of reach, engagement and visibility tend to favour audiences that are already connected and receptive, while the most marginal or institutionally distant publics risk being left behind, becoming also the least cost-effective from a metrics perspective. Van Dijck's (2014) concept of datafication is instructive here: KPIs are not neutral measurement tools, but mechanisms that quietly shape what counts as success and who counts.

Inclusion thus emerges not merely as a normative principle to be translated into content, but as a terrain of permanent negotiation between institutional values, political and reputational pressures, platform logics and the actual conditions of audience reachability. The fundamental tension concerns the capacity to represent social plurality without retreating in the face of polarisation, and to reach marginal subjects without subordinating inclusion to quantitative performance.

What appears to vary across the institutions is less the intensity of the pressure than where it is located, and what that location does to inclusive intent. In Commission and Council accounts the threat tends to be placed outside — hostile publics in one case, Member State positions in the other — so that inclusion has to be defended, calibrated or absorbed each time it is exercised. In Parliament accounts it appears to be located inside, in the routinization of inclusive communication as professional habit. In both cases inclusion rests on people and not on procedure: on those who defend it where it is contested, and on those who practice it where it has become routine. What separates the three institutions may therefore lie in their organizational arrangements and political mandates, while the vulnerability these produce appears to be shared.

4.2.3 Embedding DEI in organizational culture

Across the interviews, the conviction that DEI works best when it sits at the foundation of organizational culture itself emerges most clearly. This means building it into training, plural teams, professional competence and shared procedures. But it also means sustaining an internal architecture, visible in the roles set aside for it, the standing those roles are given, and the communicative competence.

One of the most tangible signs of European institutions' commitment to DEI is the breadth and ambition of their training offer. Programmes range from women's leadership and unconscious bias to neurodiversity, gender bias in data, and easy-to-read language, a comprehensive portfolio that reflects genuine investment in building organizational competence. Crucially, this investment in formal training is often accompanied by a commitment to reflexive practice: speaking directly to the people being represented, testing assumptions with colleagues who share relevant lived experience, and treating peer knowledge as a legitimate form of expertise. As an official declares:

The most important thing for us is always to speak to the person being represented, so if we know we have a colleague with a disability, we ask them […] I know these are my strengths, but I also know they are my biases… so sometimes we test this out on colleagues too (id. 7).

Alongside these structured offerings, forms of internal peer-teaching are emerging — contextually grounded, practically oriented, and often more effective than external consultancy. Individual officials who have developed specific expertise are increasingly acting as informal catalysts for cultural change, carrying knowledge across units and helping to make abstract principles operational.

The challenge, as several interviewees acknowledge, lies in turning this breadth of provision into genuine cultural transformation. In some institutions, the sheer volume of available modules — dozens on inclusivity, over two hundred on diversity — risks producing a kind of saturation effect. When training becomes a box to tick and not a practice to internalize, its impact on everyday behavior is limited. This resonates with recent syntheses of the diversity-training literature, which caution that the proliferation of, and investment in, such programmes has outpaced the evidence of their effectiveness in changing everyday behavior (Devine and Ash, 2022).

We'd effectively waste a lot of time if we had to do them all [the courses]… the reality is that there are far too many training courses on diversity (id. 19).

A further complexity concerns the uneven distribution of mandatory participation. Outside of recruitment, representative or political functions, training remains voluntary — even for staff responsible for setting strategic guidelines. The case of anti-discrimination training for Members of the European Parliament, which became compulsory in the wake of the #MeToo movement is instructive: it shows what it actually takes for individual sensitivity to become organizational obligation, external pressure and reputational accountability. One interviewee argues that:

There is one exception: training on verbal and sexual discrimination is compulsory for MEPs […] since the #MeToo movement reached the European Parliament, the rules have changed and now, if MEPs do not complete the training within six months of their election, this is published on their website and in newspapers […] so they all do it (id. 20).

Outside these cases, training often remains an option for personal development, encouraged but not imposed. Where mandatory frameworks are absent, the system often relies on the commitment of individual officials, frequently working in conditions of under-staffing and functional isolation. As an interviewee declares:

We are not obliged to do so, but as part of our personal development we are encouraged to draw up a sort of training plan for ourselves (id. 2).

The institutional difference lies along the voluntary/mandatory spectrum. Within the Commission, the embedding of DEI relies on individual commitment: reflective practice, training as a form of voluntary personal development, and informal peer teaching. The Council represents the governments of the Member States, and inclusive choices are calibrated to national sensitivities, so that DEI is experienced as a constraint arriving from outside the Secretariat. The Parliament is directly elected and publicly accountable, and it is the only institution in which an obligation has taken hold: anti-discrimination training became compulsory for MEPs once non-compliance could be made public. Interviewees report that the obligation followed the #MeToo movement, and it is reputational exposure that explains its hold: the requirement took root where failure to comply could be seen.

A common and well-intentioned practice is to involve colleagues with disabilities or from minority groups in the validation of communication campaigns and initiatives. Yet this practice, however good its intentions, reveals a structural fragility: rather than building systematic in-house expertise, it delegates the accuracy of the institution's communication to the occasional availability of colleagues who belong to the groups being represented, effectively casting them as “guinea-pig” witnesses, called upon case by case to validate the work of others. The reliability of this expertise, in this model, depends less on codified knowledge than on a network of personal relationships, one that remains only as stable as the individuals who hold it together. This dynamic exemplifies what recent scholarships term identity taxation is, whereby the representational labor of ensuring inclusive communication is disproportionately devolved onto minoritized staff themselves (Johnston and Yarrow, 2022).

This is, ultimately, the central tension of DEI as organizational culture: the gap between the scale of the aspiration and the slenderness of the structural support. The training investment is real; the commitment, at the individual level, is often remarkable. What remains open is the question of how to move from a model built on personal dedication to one anchored in shared and durable organizational practice.

4.2.4 Artificial intelligence and DEI risks

The final frontier — and in some ways the most unsettled — explored in the interviews concerns artificial intelligence. As European Institutions navigate the rapid adoption of generative tools, a distinctive institutional logic has emerged: one that prioritizes control, accountability and technological sovereignty over speed and scale, and that links AI governance explicitly to the values of inclusive communication. The picture that emerges from the interviews is not one of resistance to innovation, but of a carefully bounded engagement with it, grounded in the principle that responsibility cannot be delegated.

A prevailing stance of caution runs through the institutions analyzed, grounded in the need to maintain control over data, tools and the outputs produced. AI is progressively integrated into working practices, but within defined organizational boundaries, a preference for proprietary or in-house developed tools, such as GPT@EC, and a strict limitation on external systems when sensitive or confidential information is involved. The choice, as argued by several interviewees, is institutional, not technical. And it comes with a clear principle of accountability: responsibility for the output remains with the person, not the system.

We do not use ChatGPT […] we have our own internally developed tool […] it always comes with the responsibility that you are accountable for the output (id. 2).

This caution is even more evident in external communications and, in particular, in the production of images, where the ban on AI-generated imagery is explicitly linked to the desire to show real people:

We are not allowed to use AI for images. For external communication. […] We want to show real people because we want our audiences to identify with the communication products we produce (id. 6).

This is a significant alignment: the choice to show real people responds not only to requirements of control or reliability, but to the desire to foster identification, recognition and authenticity in relations with the public — values that sit at the core of any genuinely inclusive communication. The result is a model of selective adoption: AI is used where it can be kept within a framework of security and oversight; restricted where it risks compromising the authenticity of representation. Compared with previous digital communication tools, generative AI introduces both specific opportunities and a qualitatively different set of representational risks. On the opportunity side, interviewees point to cases where AI actively supports inclusion: AI tools, used in combination with English as a working language, can help neutralize gender markers across the EU's twenty-seven official languages: by enabling rapid drafting and cross-lingual communication at scale, they reduce the need to navigate gendered grammatical structures in each national language — a practical advantage that earlier digital formats could not offer in the same way (id. 2). At the same time, generative systems produce text, images and audiovisual material at scale without the staged human decisions required by earlier digital formats, making the identification of stereotypes and exclusions both more urgent and more difficult to catch. This challenge extends to synthetic images, manipulated visual content and deepfakes — risks already flagged in the literature (Lovari and De Rosa, 2025; Luttrell et al., 2025) — which raise fundamental questions about authenticity, trust and the integrity of inclusive representation in AI-mediated institutional communication.

Yet a second level of risk, less visible and less closely monitored, concerns the DEI quality of the outputs themselves. According to those interviewed, AI-assisted content generally passes through the standard review and approval chain, without a dedicated checkpoint for representational bias, gender or ethnic stereotypes, or other distortions that generative systems may produce or amplify. Awareness of this risk is present, however. As an interviewee points out:

The general guideline is not to use [AI] for any images […] and, if it is used for writing or editing, to always review everything, obviously to avoid bias and to avoid hallucinations (id. 9).

Bias and hallucinations are recognized as concrete risks. As some interviewees observe, “non-delegable responsibility” is necessary to control them. Institutions rely on the same kind of human review applied to any content before publication. This review is designed to catch factual errors, tone problems or oversights, not to check whether an AI-generated text or image reproduces stereotypes or excludes someone. A tension therefore runs between policy and practice. The 2026–2030 cycle identifies artificial intelligence as a vector of gender bias and, at the level of strategy, moves towards impact assessment of AI systems; yet interviewees report that at the level of daily production no dedicated representational checkpoint exists. AI-assisted content passes through a clearance chain designed to catch factual errors, tone and legal exposure, not to detect whether a generated content reproduces stereotypes or renders someone invisible. The risk is named in policy and acknowledged by professionals, but the safeguard has not been operationalized: representational governance appears, in the units sampled, to lag behind the technological governance that is already explicit and codified. Access to AI tools is regulated not only by policy but by resource availability and computational cost, creating a multi-level governance structure that is regulatory, organizational, and material all at once. The challenge ahead lies in moving from general oversight to more explicit forms of DEI monitoring, capable of integrating the assessment of inclusion risks into approval procedures. The robustness of technological governance should not be allowed to obscure the relative weakness of representative governance, that is where the next frontier lies.

5 Discussion

The discussion addresses the two research questions: how EU professionals interpret and apply DEI policies in their daily communication practices (RQ1), and how this approach is perceived, with the alignments, ambivalences and communicative clashes that emerge from it (RQ2).

Over the past decade, the EU has extensively codified DEI policies through a multilevel approach that has made inclusive communication a policy priority, embodied in the participatory principle of Nothing about us without us, through soft law instruments, guidelines, strategies, and roadmaps.

It is precisely the robustness of this system that makes the findings of the interviews significant: according to the interviewees' perception, this commitment in terms of policies has not translated into institutionalization, that is, into the incorporation of the principle into the organization's routines, roles, and evaluation metrics. The result is what can be defined as a state of entrusted inclusion: a commitment upheld on a daily basis by individuals, such as officials who are particularly sensitive to the issue and colleagues belonging to the represented groups who are called upon to validate the inclusive processes. The risk, therefore, is that this approach may falter due to the turnover of the people who support it. In response to RQ1, this indicates that professionals interpret and apply DEI less as a codified organizational routine than as a discretionary practice sustained by individual commitment.

It is in this gap between stated intent and daily practice that communicative clashes arise—that is, practices that interviewees describe as forms of tokenism and diversity washing, in which the message appears to contradict, or at the very least undermine, the principle it was intended to affirm. This inevitably leads to tensions and ambivalence that the interviewees perceive as structural and linked to the complex workings of the EU. In response to RQ2, DEI is thus perceived by communicators through recurrent ambivalences and communicative clashes and not through stable alignment. The predominantly female composition of DEI-related roles compounds the problem, risking the confinement of inclusion to a specialized niche and fail to become the cross-cutting competence that mainstreaming presupposes.

A parallel vulnerability resides in the nature of organizational proficiency. Whilst the availability of training is vast and personal dedication is frequently authentic, institutional routines often default to the incidental participation of staff from marginalised communities. This reliance on the ad hoc availability of minoritized colleagues substitutes for structured, institutionalized expertise. Ultimately, the robustness of this paradigm hinges upon informal social webs more than institutionalized professional standards, rendering these processes only as permanent as the specific personnel who sustain them.

Another factor, moreover, concerns external pressure and the difficulty European institutions face in communicating with twenty-seven national contexts characterized by profoundly different political and cultural sensibilities. Inclusion remains the stated goal, but the risk of backlash requires a constant balancing act between recognition and reputational caution, with outcomes that are not always predictable: as described by one interviewee, the Roma community campaign illustrates how backlash can be managed without abandoning the goal of inclusion, while in other instances, political controversy has led to the stalling of initiatives that had already been launched.

This aspect is also linked to the use of artificial intelligence, an area in which European institutions have established a technically sound framework for protection that is consistent with their stated values.

The interviews reveal a clear orientation: a preference for internally developed tools, restrictions on external systems when sensitive information is involved, and a ban on AI-generated images in external communications. Responsibility for the output cannot be delegated to the system but remains with the person who produces and validates it. It is precisely this strength, however, that creates a blind spot.

The system is designed to ensure security, privacy, and technological sovereignty, not to verify the representativeness of what is produced—which carries the risk of creating and amplifying gender, ethnic, or intersectional biases. In the units sampled, technological governance is explicit and codified; representational governance remains implicit, left to the discretion of those conducting the review.

6 Conclusions

This study contributes the notion of entrusted inclusion to describe a setting in which robust DEI commitments are sustained less by institutional routines than by the discretionary effort of individual professionals, and it identifies the communicative clashes through which inclusive intent is diluted as it moves through validation. Across the four thematic areas, the interviewees converge on the same tensions but read them through institutionally distinct registers. On decision-making, inclusion, and organizational culture, Commission respondents describe inclusive intent as diluted by a long administrative clearance chain; Parliament respondents frame it in party-political terms while Council respondents locate the constraint outside the institution, calibrated to Member State positions and perceived as external to the Secretariat.

On artificial intelligence, by contrast, the three seem to converge, sharing a common stance of caution, in-house tools and non-delegable responsibility. The AI Act is the one binding instrument the Union has adopted in this field, yet Article 50 exempts from disclosure any text that has undergone human review and editorial control. Institutions are thus accountable for declaring how content was produced, not for what it represents, and the review that exempts them is the same one that, on their accounts, contains no representational checkpoint.

This raises several questions that the research leaves unanswered. Inclusive communication currently holds because individual professionals decide to make it hold. What would have to change — in job descriptions, in written procedures, in performance criteria — for it to hold once those communicators move on? Responsibility for AI-generated content stays with the official who approves it, yet nothing tells that official how to recognize a stereotyped image or a skewed text: the competence is not required at recruitment and not covered by training. Who should set that standard, and where in the approval chain should it be applied?

These results should be interpreted in light of certain limitations. Although the sample spans three institutions, it provides more substantial empirical evidence for the European Commission than for the Council and the Parliament, and it primarily reflects the perspectives of professionals already involved in roles related to DEI policies. The behavior of senior validators is reconstructed largely from the accounts of professionals who work below them in the hierarchy. A fuller triangulation, including interviews at management level and organizational documentation, is envisaged for the subsequent phases of the research. Future research should extend the investigation to different levels of decision-making and integrate the analysis with the dynamics of public reception, a dimension that this study was able to address only indirectly, through the perceptions of institutional actors.

Author's Note

The author is a Serra Húnter Fellow.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher's note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Raffaele Lombardi, Libera Università Maria SS. Assunta, Italy

Reviewed by: Chendi Wang, VU Amsterdam, Netherlands

Sara Pane, University of Turin, Italy