Different refugees,different portrayals. A GDELT-based comparative analysis of the mediatic portrayal of Syrian refugees in 2015 and Ukrainian refugees in 2022 across seven European countries

Abstract

This study examines how European media portrayed Syrian refugees in 2015 and Ukrainian refugees in 2022, across seven national contexts. Drawing on the GDELT database of news articles and its automated sentiment analysis, the study compares the tone and salience of coverage surrounding two highly symbolic mediatic events: the death of Syrian child Alan Kurdi in 2015 and the outbreak of the war in Ukraine in 2022. Results reveal stark disparities: Ukrainian refugees were consistently represented more positively than Syrian refugees and non-Syrian/non-Ukrainian refugees. These findings provide quantitative evidence of how cultural, racial, and geopolitical proximity shape refugee representations in European media, thereby confirming previous literature. While some prior scholarship has demonstrated these dynamics, this article contributes to existent literature by offering a large-scale, cross-national, and event-based comparative analysis.

Introduction

The aim of this article is to analyse whether Syrian and Ukrainian refugees were portrayed differently by European media during two significant migratory influxes: the 2015 inflow of Syrian refugees escaping the civil war, and the 2022 flight of Ukrainian refugees following the Russian invasion. The comparison becomes particularly relevant when we consider the contrasting political and social reactions to the arrival of these two displaced groups in Europe. In 2015, the so-called ‘refugee crisis’ – a term widely debated in the literature (Chouliaraki and Stolic, 2017; Krzyżanowski et al., 2018; Şahin-Mencütek and Triandafyllidou, 2024) – was construed as an emergency, triggering the securitisation of European borders, externalisation policies, and disputes over asylum quotas (Caballero Vélez and Krapivnitskaya, 2020; Ceccorulli, 2018). In contrast, in 2022 Ukrainian refugees received an unprecedented level of support at the policy level, with the first-ever activation of the EU's Temporary Protection Directive since its creation in 2001 (Council of the European Union, 2023; Eurofound, 2022).

This article contributes to existing literature by providing a computational comparative analysis of mediatic representations of the two refugee groups across seven European countries and in the global media sphere. Using the GDELT database and leveraging its automated multilingual translation and sentiment analysis capabilities, it examines changes in the tone and number of articles regarding the two specific refugee groups and also non-Syrian/non-Ukrainian refugees (in 2015 and 2022 respectively) during a four-month period surrounding two crucial dates: the death of Alan Kurdi on 3 September 2015, and the outbreak of war in Ukraine on 24 February 2022. It focuses specifically on non-Syrian/non-Ukrainian refugees before and after the events, and on Syrian/Ukrainian refugees after the events. The study does not claim to capture the full complexity of media narratives, as it does not consider visual framing or fine-grained linguistic nuances. Instead, it focuses on measurable proxies – namely, tone and article count – as indicators of media representation patterns.

The structure of the article is as follows: we first provide an overview of previous studies of comparable scope, and then focus on the article's contribution to the existing literature. Then, the methodology section details the characteristics of the GDELT database and the procedures of data collection and analysis. Next, the findings section offers empirical evidence of differences in the portrayal of the refugee groups analysed, through graphs and tables that illustrate variation in tone and article count. This is followed by a nation-specific discussion of the results to highlight key patterns and trends. Finally, the conclusion integrates the main findings, reflects on the study's limitations, and proposes avenues for future research.

Refugees in the media

Significant scholarly attention has been directed towards understanding the role of the media during global crises, particularly armed conflicts and humanitarian emergencies (Bennett, 1996, 2003; Bennett and Entman, 2001; Deuze, 2020; Entman, 1993; Ibañez Sales, 2023; Iyengar and Simon, 1993). In such contexts, media coverage is pivotal in shaping public perceptions and influencing policy responses (Eberl et al., 2018; van Dijk, 2018). Within migration studies, media portrayals of refugees have been shown to frame displaced populations either as social threats or as humanitarian victims, with profound consequences for public sentiment and political action (Boomgaarden and Vliegenthart, 2009; Chouliaraki, 2006; Gabrielatos and Baker, 2008; Gallant, 2022; Georgiou and Zaborowski, 2017; McCann et al., 2023).

Empirical studies provide nuanced evidence of these dynamics. Gabrielatos and Baker (2008), for instance, found that U.K. newspaper coverage from 1996 to 2005 largely framed asylum seekers as a threat by employing dehumanising metaphors such as ‘sneaking in’, ‘swarms’, or ‘floods’. Similarly, Boomgaarden and Vliegenthart (2009) demonstrated that both the volume and negative tone of immigration-related news in Germany were strongly correlated with increased anti-immigration attitudes among the public. While occasional high-profile events, such as the death of Alan Kurdi, temporarily shifted coverage towards humanitarian frames, these shifts were generally short-lived and failed to alter prevailing negative trends (Raja and Alotaibi, 2018). Comparative analyses across Europe have also shown that negative portrayals – such as burden and security frames – predominate in media coverage, although humanitarian frames are also present, revealing cross-national variations in framing practices (Joris et al., 2018).

Moreover, multiple studies stress that the notion of a ‘refugee crisis’ is often a discursive construct rather than an objective fact, shaped by media framing and political discourse to amplify public anxiety and justify restrictive policies (Chouliaraki and Stolic, 2017; Krzyżanowski et al., 2018; Şahin-Mencütek and Triandafyllidou, 2024). Additionally, Smets and Bozdağ (2018) emphasise that mediatic representations of refugees should not be examined in isolation from their impact on public opinion or the role of media literacy. Negative portrayals, they argue, risk reinforcing stereotypes and fuelling exclusionary attitudes, underscoring the need for audiences to critically engage with media narratives.

A number of recent scholarly articles analyse the framing of Syrian and Ukrainian refugees. Gallant (2022) identifies a stark contrast in how these groups were depicted in one of Germany's most widely read superregional newspapers. Syrian refugees were often framed as security threats, a portrayal shaped by Islamophobic stereotypes, while Ukrainian refugees were predominantly depicted as ‘good refugees’, benefiting from their racial and geopolitical proximity to European societies. Similarly, Roman et al. (2024) provide a comparative analysis of British, Russian and Polish news coverage regarding Ukrainian refugees, finding no negative portrayal of Ukrainian refugees in Poland and the United Kingdom. Furthermore, in a comparative analysis of U.K. mediatic coverage from the BBC and The Sun, Ajana et al. (2024) state that Ukrainian refugees were overwhelmingly framed in humanitarian terms, while Syrian and Afghan refugees were often portrayed as cultural and security threats. This mirrors the findings of Wildemann et al. (2023), whose multilingual analysis of tweets from 565 different European news outlets reveal a significant shift in mediatic sentiment about migration during the Ukrainian crisis, with terminology moving from ‘migrant’ to ‘refugee’, signalling a more positive stance of the media towards Ukrainian refugees compared to other groups. El-Nawawy and Elmasry (2024) expand this perspective in their examination of elite American newspapers, uncovering stark contrasts between the humanitarian framing of Ukrainian refugees and the securitised portrayals of Syrian refugees, thereby underscoring the significance of geopolitical proximity. Hoffmann and Hameleers (2024) also conduct an automated content analysis of newspapers from Germany, Spain, the United Kingdom, and Switzerland, demonstrating how framing practices are consistently more favourable towards Ukrainian refugees in diverse national contexts.

While existing studies have brought to light disparities in media portrayals of Syrian and Ukrainian refugees, several critical gaps persist, particularly in terms of methodological approaches, temporal analysis, and the integration of multilingual datasets. Eberl et al. (2018) highlight the scarcity of large-scale, comparative studies analysing mediatic portrayal of different refugee groups across multiple national contexts. Their review underscores the necessity for research that transcends national boundaries so as to improve understanding of cross-national variations and general trends in the mediatic portrayals of refugees. Similarly, Ibañez Sales (2023) emphasises the importance of comprehensive analyses that incorporate sentiment analysis and multilingual datasets to address the gaps in existing research.

This study makes a distinctive contribution to the literature by directly addressing the methodological and comparative gaps long identified in the research on mediatic representations of refugees. Whereas previous scholarship has typically focused on single countries or on a more limited number of media outlets, our design leverages the GDELT database to conduct a large-scale, cross-national, cross-temporal comparison. In so doing, it addresses Eberl et al.'s (2018) call for comprehensive comparative research capable of highlighting both national variations and systemic biases. Methodologically, the article advances the field by combining automated sentiment analysis with large-scale dataset analysis, offering scalable yet conceptually grounded evidence of tone and salience. By integrating daily article counts with GDELT's Global Content Analysis Measure (GCAM) tone indicators, we produce large-N measures that complement (rather than replace) qualitative frame analyses, thereby contributing systematic, replicable evidence of mediatic bias (Ibañez Sales, 2023; Hoffmann and Hameleers, 2024; Ongenaert and Soler, 2024). Finally, we recognise the text-centric orientation of the study as both a strength and a limitation. While automated English-language processing cannot capture all linguistic nuances, and visual framings are excluded from this analysis, we integrate insights from scholarship on visual and multimodal communication (Bozdag and Smets, 2017; Martikainen and Sakki, 2024; Mortensen and Trenz, 2016) to justify the event selection and to frame avenues for future research. In this way, the article not only provides robust empirical evidence of systematic differences in the portrayal of Syrian and Ukrainian refugees, but also offers a replicable methodological framework for studying media representation of social groups across events, contexts, and time.

Methodology

This study uses a mixed methodology – derived from Weisser (2023) – that combines a quantitative analysis performed by the authors (the extraction and analysis of news article datasets) with the qualitative analysis performed automatically by the GDELT database on the articles it collects. All the non-English news articles collected every 15 min by the GDELT database are machine translated from 65 different languages into English through the GDELT Translingual platform (becoming the largest real-time streaming news machine translation deployment in the world), before being processed and coded. The qualitative component of the processing and coding is provided by the GDELT deep learning algorithms that automatically codify each article into a series of different parameters, including date of publication, the URL of the source, the actors involved, and the average sentiment. It employs the Conflict and Mediation Event Observations Event and Actor Codebook (CAMEO) to designate events and actors through specific codes. It also employs the GCAM algorithm, a GDELT-specific algorithm that relies on many different sentiment dictionaries to assign a tone value to the article, with scores ranging from −100 (extremely negative) to +100 (extremely positive) (Saz-Carranza et al., 2020). In this research, we report and calculate period averages as proxies for the overall evaluative tone used towards a group. GDELT has been acknowledged as a reliable source on migration events by Tjaden (2021) and has been used for different research purposes (Boudemagh and Moise 2017, Carammia et al. 2022, Weisser 2023).

We first extracted specific datasets from the GDELT databases on Google BigQuery using Structured Query Language (SQL), focusing on articles from the selected contexts and timeframes in which refugees from specific geographic locations or nationalities are identified as primary actors (according to the CAMEO taxonomy). For each country (plus the global context), eight datasets were extracted: four regarding non-Syrian/non-Ukrainian refugees in the two months before and after the events, and two regarding Syrian or Ukrainian refugees in particular in the two months following the events. As noted earlier, we analysed the periods before and after two crucial dates (3 September 2015 and 24 February 2022), selected for their association with highly symbolic media events. On the one hand, the picture of Alan Kurdi's lifeless body arguably became a powerful visual icon that created an ‘impromptu moral spectatorship’ (Mortensen and Trenz, 2016) – although some literature argues that it did not fundamentally change how refugees were discussed (Bozdag and Smets, 2017). On the other hand, photographs of Ukrainian refugees crossing European borders were deployed as humanising symbols, fostering solidarity and a sense of proximity to them (Martikainen and Sakki, 2024). Thus, we analysed articles on general refugees ¹ for the two months before (02 July 2015–02 September 2015) and after (02 September 2015–02 November 2015) Alan's death, as well as for the two months preceding (24 December 2021–24 February 2022) and following (24 February 2022–24 April 2022) the Russian invasion of Ukraine. In the post-event period, instead, we solely focused on Syrian and Ukrainian refugees. This choice stems from a comparability constraint, as there was virtually no news coverage of Ukrainian refugees before the Russian invasion.

Specifically, to ensure we collected articles exclusively from a specific national context, we applied restrictions on the top-level domain of the source URL (Saz-Carranza et al., 2020). This approach approximates national media coverage, but it may omit national outlets using different domain conventions and can capture internationally hosted outlets. Therefore, we treated national results as indicative rather than exhaustive. Once such queries were performed on the dataset via SQL language, the corpus of extracted datasets was uploaded and analysed in RStudio, in order to calculate tone averages and total article count (see Appendix 1 for further details).

Regarding the choice of countries, a diverse range of relevant media landscapes was selected. Italy, France, Germany and Spain were analysed due to their demographic and political weight in the European Union, while Hungary and Poland were taken into consideration due to their shift from an antimigration stance in 2015 to an open borders policy for Ukrainians in 2022. Moreover, the United Kingdom, despite its departure from the EU, continues to be a major source of English-language news, while consideration of the global context allowed us to gauge the global trend.

Finally, to analyse the variations in tone across different countries, we conducted a one-way ANOVA test. First, we calculated the mean tone for each country in each period. Then, ANOVA assessed whether tone variations across these countries were due to random fluctuations, or whether there was a systematic difference attributable to the geographical dimension at each time (see Appendix 2). To further interpret the magnitude of these differences, we calculated the effect size (η²), which quantifies how much of the variance in tone scores can be explained by temporal categories. Finally, after running a Levene test for homoscedasticity, we ran post-hoc Games-Howell tests to identify which specific countries differed from each other. We then repeated the analysis, comparing the average tone in each period by country (Appendix 3).

Findings

Figure 1 shows the quantity of articles related to various refugee groups, analysed distinctly within each national context. We have to bear in mind that GDELT is based on the English language and applies automated translation into English to categorise the articles, which explains why the United Kingdom is the context with the largest number of articles. Figure 2 gives a clear idea of the mediatic salience of each refugee group in each context, allowing us to also appreciate the impact of the two events. Across all countries, while refugees in general receive similar media attention, coverage of Syrians and Ukrainians is far more prominent – at least twice that of all other migrant groups combined. However, national specificities emerge: while France, Germany, Spain, and the UK show similar coverage of both nationalities, Italian and Polish outlets prioritise Ukrainians, whereas Hungarian ones prioritise Syrians (see below and Annex 3).

Figure 1.

Number of articles for each refugee group and country in absolute terms.

Figure 2.

Average tone of articles for each country and refugee group in absolute terms.

Tone of articles

The tone used in the articles analysed is uniformly negative, with values ranging from −5.4 (British tone for refugees after Alan's death), to −1.14 (Italian tone for Ukrainian refugees). Considering the eight scenarios examined (seven countries and the global context), statistically significant differences are observed between the countries in each period. However, country differences account for only 2–4% of the variance in tone across all refugees, and for 11–25% of the variance when considering Syrian and Ukrainian refugees, respectively. Therefore, the effect size (η²) is small: the country has a limited impact on tone variation across all refugees and a moderate impact for Syrian and Ukrainian refugees, with most of the variation occurring within countries (see Appendix 2 for further analysis).

Figure 2 illustrates the average tone. Each context is represented by six bars corresponding to a specific refugee group subjected to analysis: ‘Ref’ denotes non-Syrian/non-Ukrainian refugees, ‘Syr Ref’ signifies Syrian refugees, and ‘Ukr Ref’ indicates Ukrainian refugees. Ukrainian refugees (dark grey bars with dashed border) receive the most favourable (i.e. least negative) portrayal among all groups, while refugees from the period before Alan's death (white bars with solid border) score the lowest tone values (i.e. most negative), except in Italy, where refugees from the period after Alan's death (white bars with dashed border) score lowest. The statistical analysis gauging the differences in tone across different time periods is provided in Appendix 3. This analysis indicates that the average tone differs significantly over time (p-value < .001), confirming statistically significant differences between the groups. Between 14% and 40% of the variance in tone (η²) is explained by these time-based differences, reflecting a medium effect size and a notable influence of the scenario on tone variation.

The data on the articles’ tones are also displayed in Tables 1 and 2 via heatmaps, to facilitate visualisation from a broader, less country-specific perspective, in order to assess general trends and peculiarities. A grey gradient is used to represent these data, with dark grey signifying the lowest average tone value and light grey signifying the highest. For example, Table 1 shows that the representation of refugees in 2022 seems to be generally more positive, as the last three columns present ligheter gradients.

Table 1.

Heatmap with average tones for each refugee group (in absolute terms).

Table 2.

Heatmap with differences in average tones among the different refugee groups (in absolute terms).

In turn, Table 2 displays the differences between different refugee groups in the same year. This heatmap helps to assess the effect of the two historical events, as it draws a direct comparison between refugees (before and after the event), and Syrian or Ukrainian refugees in particular. For instance, by noting the generally lighter grey in the fifth and sixth columns, we can observe the higher average sentiment of articles about Ukrainian refugees compared to other refugees in the same year. By calculating the difference, we can also get a more precise idea of the impact of such events on each context and the evolution of tone values. For example, by observing the interval between the highest and lowest values in every row, we can obtain a general picture of the fluctuation of mediatic representations of refugees for every context. We can thus observe that the global context exhibits the lowest variation among groups, resulting in a more consistent and coherent portrayal compared to the other contexts.

Nation-specific analysis

France

The French data comprise a high volume of articles for both Syrian and Ukrainian refugees (notably peaking at over 1800 and 1600 respectively), accompanied by strongly negative average tones (around −4 across groups and time). Syrian refugees are the group with the highest number of articles. Other refugees before Alan's death show the lowest tone value (−4.33) and, while Alan's death slightly increased the tone of both Syrian and non-Syrian refugees thereafter, the tone values are still below −3.5. In 2022, refugees both before and after the invasion receive negative tone values (−3.98 and −3.29 respectively), while Ukrainian refugees receive the highest tone value among all groups (−2.09). The overall negative tone of articles regarding refugees in 2015 could plausibly be explained by France's particular exposure to terrorist attacks in that year: those of January (Charlie Hebdo) and November (Paris) likely intensified discussions on securitisation and terrorism, while also contributing to the spread of Islamophobia (Cohu et al., 2016; Najib, 2020). The tone seems relatively constant also in 2022, suggesting that the French media generally depicts refugees quite negatively, while the high volume of articles on both Syrian and Ukrainian refugees suggests a notable impact of the two mediatic events on the mediatic discourse.

Germany

Germany played a central role in dealing with the refugee influx to Europe in both years analysed. The data shows a remarkably similar trend to France, with Syrian refugees collecting the highest number of articles, followed by Ukrainian ones. Tone values are also fairly similar, with refugees before Alan's death scoring the lowest tone value (−4.88) and Ukrainian refugees the highest (−2.02). However, it is interesting to note that the overall sentiment in 2015 was remarkably negative, even lower than France. Indeed, in 2022 refugees score over one point more compared to their counterparts in 2015 both before and after the event (−3.45 and −3.34, compared to −4.88 and −4.53). The time trend is not influenced by the two events if not marginally (0.35 for Alan's death and 0.11 for the Ukrainian invasion). This might indicate that both caused a selective increase in the average tone only for Syrian (1.15) and Ukrainian (−1.43) refugees. Overall, the consistent overall negative tone values seem to stand in contrast to the ‘Willkommenskultur’ expressed by the German government with the ‘Wir schaffen das’ response to the migratory influx, but might also be connected to the political backlash that fuelled the rise of far-right parties (Conrad and Aðalsteinsdóttir, 2017).

Hungary

Data from Hungary show a series of peculiar results. It is the only context where the number of articles regarding Syrian refugees greatly exceeds the volume of all other refugee groups. In a similarly unique fashion, it is the only country where Syrian refugees score a lower tone value compared to other refugees after Alan's death. It would seem, therefore, that the tone of coverage improved more markedly for non-Syrian than for Syrian refugees following Alan's death. This dynamic could be linked to the Orbán government's centralised media control and its strategic deployment of migration as a political tool during and after 2015 (Bolonyai and Campolong, 2017). The construction of the border fence and the ‘Stop Soros’ campaign might have contributed to an ideologically consistent and hostile portrayal of Syrian refugees. On the other hand, the Russian invasion seems to have had a remarkably positive impact on the portrayal of both Ukrainian refugees and refugees in general, which score the two contextually highest tone values (−2.27 and −2.61, respectively). It is also interesting to note that the salience of refugees in the Hungarian media was greater in 2015 than in 2022, with a much higher total article count (Figure 1). In general, however, the relatively modest article counts might reflect both a more limited mediatic environment and a general disinterest towards refugees, but we should also bear in mind possible translation issues due to the complexity of the Hungarian language.

Italy

Italy stands out for producing a remarkably high number of articles about Ukrainian refugees (more than 2500), while registering a very limited number of articles for Syrians and for refugees in general. Italy also surprisingly stands out for its more positive tone values overall. It shows the highest tone value in the sample for refugees in general before Alan's death and after the Russian invasion, and for Ukrainian refugees. The fact that, in 2015, Alan's death seems to have paradoxically decreased the average tone value for both Syrian refugees and refugees in general is remarkably unique. On the other hand, the Russian invasion seems to have had a notably positive impact, greatly increasing the tone values also for non-Ukrainian refugees. Italy's long-standing position as a key entry point into Europe via the Central Mediterranean route, which would tend to normalise migration as a recurrent ‘emergency’ in Italian media coverage, might explain the paradoxical mediatic effect of Alan's death. This counterintuitive outcome might suggest that the tragedy, while generating global compassion, may have reinforced existing narratives of crisis, chaos, and the state's incapacity to manage arrivals by sea, reflecting a saturation effect. In contrast, the outbreak of the war in Ukraine in 2022 may have reactivated public attention, with the novelty of the crisis, together with its geographical and geopolitical proximity, producing a surge in both the salience and tone of coverage surrounding Ukrainian refugees. The supposed cultural and religious proximity between Italians and Ukrainians might have facilitated more sympathetic portrayals, especially compared to Syrian or African refugees (Melotti et al., 2024).

Poland

Poland stands out for its very limited coverage of every refugee group analysed, with the exception of a massive surge in reporting on Ukrainian refugees after the Russian invasion. The low number of articles on Syrian refugees could be understood as a reflection of Poland's geographical and political distance from the main migratory routes: asylum applications were negligible, and the refugee issue might have been framed largely as a Western European problem. In this context, we observe markedly negative tone averages in 2015 (ranging from −5.06 to −3.54), as migration was likely portrayed as a threat to Polish sovereignty, cultural identity, and Catholic values, especially following the French terrorist attacks. The drastic increase in both article count and tone positivity for Ukrainian refugees might reflect a striking instance of selective humanitarianism. Ukrainian refugees, unlike Syrians or Afghans, could be perceived as culturally and religiously closer, belonging to the same Slavic and Christian civilisational sphere (Mandelc, 2025). Moreover, the war in Ukraine was understood in Poland as part of its own geopolitical struggle against Russian expansionism – a struggle deeply rooted in historical memory given Poland's long history of Russian domination. In this sense, welcoming Ukrainian refugees would not only be a humanitarian gesture but also a patriotic and geopolitical act, aligned with Poland's foreign policy interests. The markedly more positive tone towards Ukrainians would reflect, therefore, not a general openness to refugees but a highly conditional one, dependent on ethnicity, religion, and geopolitics.

Spain

Spain shows some interesting trends. On the one hand, it is the context where Syrian refugees score the highest tone value by a remarkable margin. On the other, post-invasion refugees receive the lowest tone value by an equally remarkable margin. Both events seem to have caused a surge in media salience for the specific refugee groups, but other refugees seem to total a very low article count. This configuration would suggest that Spanish coverage of refugees is highly event-driven and shaped by episodic bursts of salience rather than by sustained attention. The unusually positive tone towards Syrian refugees might be interpreted in relation to the fact that Spain was less directly affected by large-scale arrivals. With fewer Syrians actually reaching Spanish territory, coverage would have been less dominated by crisis-management and securitisation frames, and more influenced by the global humanitarian moment triggered by Alan's death. In this context, Spanish media may have been more inclined to reproduce international narratives of compassion, which would explain why the average tone for Syrians is significantly higher than in other countries. While Ukrainian refugees were largely portrayed in positive terms across Europe and produce a growth in tone for all refugees, in Spain the other refugees are portrayed more negatively, mirroring an increasing concern for migration from Africa and irregular border crossings in Ceuta and Melilla. The militarised and spectacular nature of these events would have reinforced highly negative depictions (Ali, 2024; Terrón-Caro et al., 2022), producing the lowest tone averages for this group. This tension between humanitarian internationalism and securitised nationalism could explain why Spain produces contrasting tone outcomes: unusually positive coverage of Syrians, alongside extremely negative framing of refugees in the post-invasion period.

United Kingdom

Predictably, the United Kingdom generally presents the highest article count. Also in this context, we can observe a large surge in articles regarding Syrian and Ukrainian refugees after the events in question. Interestingly, we can also deduce that Alan's death increased the overall salience for all refugees, as the number of articles on non-Syrian refugees in the two following months is higher than the number of articles on all refugees in the two months before his death. Moreover, the Russian invasion seems to have monopolised mediatic attention, as the number of articles on non-Ukrainian refugees in the two following months is remarkably lower than the number of articles on refugees in the two preceding months. The United Kingdom also presents the lowest tone value for refugees before Alan's death and the second-lowest tone value after the invasion. The sharp decline in the number of articles about refugees after the Russian invasion, despite the surge in coverage of Ukrainians, would indicate that the U.K. media discursively separated Ukrainians from other refugee groups (Sambaraju and Shrikant, 2023). This would mirror political rhetoric in which Ukrainians were framed as European allies fleeing a clear aggressor. Other refugee groups, by contrast, would have been increasingly linked to the ongoing debate on ‘small boats’ crossing the Channel. These results can also be interpreted in light of the Brexit context. From 2015 onwards, migration was one of the most divisive issues in the referendum campaigns, with hostile tabloid narratives reinforcing public fears of ‘losing control’ of borders. Even humanitarian spectacles such as Alan Kurdi's death were swiftly reframed in this context as evidence of chaos, danger, and the need for stricter controls (Cooper et al., 2020). After Brexit, the entrenchment of restrictive asylum policies, including the Rwanda deportation plan, would further consolidate this negative framing.

General discussion

First of all, the data reveal a time trend that has not been described before: between 2015 and 2022, the tone towards refugees generally improved across contexts, with the exception of Spain. The change in tone is either nonsignificant or positive and significant (Appendix 3) when comparing coverage of refugees in 2015 and 2022 – ether before the invasion (globally, and in Germany, Italy, Poland, and the United Kingdom) or after the invasion (in Hungary). This suggests that the anti-migration sentiment prevalent around 2015 has subsided, resulting in a generally less negative tone.

The key question driving this research is whether Syrian and Ukrainian refugees were cast in a different light in different mediatic landscapes. The evidence collected underscores a systematic difference in how the analysed refugee groups were depicted. Making use of both macro-level statistical analyses and automated sentiment analysis for each article, the research provides robust empirical evidence that reveals patterns of biased mediatic representations. We have gauged that Ukrainian refugees had the least negative/most favourable sentiment scores across all contexts analysed, with notable improvements compared to general refugees in 2022 (both before and after the invasion) and Syrian refugees in 2015. This confirms findings from previous literature, highlighting the prominence of cultural, religious and racial factors in shaping the mediatic portrayal of different refugee groups (Ajana et al., 2024; el-Nawawy and Elmasry, 2024; Gallant, 2022; Georgiou and Zaborowski, 2017; Hoffmann and Hameleers, 2024; Ibañez Sales, 2023; McCann et al., 2023; Ongenaert and Soler, 2024; Roman et al., 2024; Wildemann et al., 2023).

Both high-profile events caused a notable surge in mediatic salience, with Syrian refugees even collecting more articles compared to Ukrainian refugees in France, Germany, Hungary, and Spain. However, the Russian invasion appears to have caused a fairly homogeneous and consistently positive shift in the mediatic representation of Ukrainian refugees in all contexts. It monopolised the mediatic discourse on refugees, leaving other refugees with little or no coverage, highlighting a pattern of selective attention. By contrast, the surge in media coverage following Alan's death was not accompanied by a proportionate increase in positive sentiment towards Syrian refugees. Before Alan's death, these obtain a significantly higher tone compared to other refugees only in Spain and the United Kingdom, while they get a lower score compared to other refugees before Alan's death in Italy and after Alan's death in Hungary. Furthermore, while coverage of non-Syrian refugees also expanded after Alan Kurdi's death, this is consistent with notions of cultural and geopolitical proximity, which suggest that refugees perceived as culturally and/or geopolitically closer to host societies may be framed in a more empathetic manner (Ajana et al., 2024; McCann et al., 2023).

Conclusions, limitations and avenues for future research

The findings of this study demonstrate that mediatic representations of refugees tend to be overwhelmingly negative but not uniform. Instead, they appear to adapt to the perceived cultural, religious, and geopolitical proximity of specific refugee groups. By leveraging the GDELT database, applying large-scale statistical methods, and employing automated sentiment analysis, this research has revealed that Ukrainian refugees are consistently depicted in a more favourable light compared to Syrians and other groups. Alan Kurdi's death, while briefly boosting the salience of Syrian refugees in the media agenda, did not significantly alter the underlying framing, which remained considerably more negative than that of Ukrainians. These results suggest that discursive hierarchies are at play, where proximity – whether geographic, cultural, or political – conditions the level of empathy afforded to displaced populations, ultimately shaping both public debate and policy outcomes.

Despite these contributions, we acknowledge that the study faces several important limitations. Firstly, the reliance on machine-translated content introduces a degree of interpretive risk. Translation algorithms can obscure or flatten linguistic nuances, metaphors, and idiomatic expressions that often carry crucial discursive weight in original-language reporting. This risk is particularly salient when analysing sentiment, as subtle rhetorical devices such as irony, sarcasm, or culturally specific references may be lost or misinterpreted in translation. In this sense, tone values may not fully reflect the complexity of media discourse, especially in languages structurally distant from English. Secondly, automated sentiment analysis itself presents inherent constraints. While it allows for the systematic processing of very large datasets, it tends to operate on lexical cues and struggles to capture higher-order meanings, frames, or narratives. Media portrayals of refugees often rely on imagery, metaphors, or intertextual references that exceed the capacity of algorithmic sentiment scoring. Thirdly, structural biases in the GDELT dataset must be acknowledged. As an English-speaking country, the United Kingdom appears to be disproportionately represented, with a higher number of articles than other contexts. This reflects both the dominance of Anglophone media sources within GDELT and the reduced need for translation, which may artificially inflate the United Kingdom's visibility in the dataset. Conversely, in non-English contexts, particularly in Central and Eastern Europe, smaller media ecosystems or less comprehensive coverage in GDELT could lead to underrepresentation. Fourthly, the temporal design of the study, focused on key ‘critical events’, while analytically productive, may also simplify more gradual or long-term dynamics in refugee coverage. Media representations evolve over extended periods in response to shifting political agendas, electoral cycles, or broader social debates. Finally, the analysis is limited to tone and article volume, which serve as proxies for media representation. While these indicators capture broad trends, they cannot fully account for other discursive dimensions.

Building on these limitations, future research could move in several directions. Expanding the temporal scope to include other refugee crises and geopolitical conflicts would allow for a broader historical perspective. Incorporating additional national contexts, particularly outside Europe, would also test the extent to which cultural and geopolitical proximity influences media coverage globally. Future studies would benefit from mixed-method approaches that combine computational analysis with qualitative techniques such as frame, critical discourse, or visual analysis, allowing for a richer understanding of how refugees are represented across different contexts. For example, a visual analysis of images scraped from the articles collected in the research could supplement meaningfully to the findings, bringing to light possible differences in the visual depiction of the two refugee groups. Thus, triangulating GDELT-based analyses with other datasets, or complementing quantitative sentiment measures, would increase robustness and validity.

In conclusion, this research confirms the persistence of selective and uneven portrayals of refugees in European media. These portrayals not only shape public opinion but also influence policymaking, reinforcing hierarchies of deservingness that privilege certain groups while marginalising others. By exposing these patterns, this study underscores the urgent need for more balanced and responsible reporting. The methodological framework applied here, while not perfect, offers a valuable methodological blueprint for subsequent investigations. In highlighting both the strengths and the limitations of computational approaches, it encourages future research to build more nuanced, multi-layered accounts of how media discourses construct refugeehood, and how these constructions ultimately feed into broader struggles over identity, solidarity, and exclusion.

Supplemental Material

sj-docx-1-ejc-10.1177_02673231251394694 - Supplemental material for Different refugees, different portrayals. A GDELT-based comparative analysis of the mediatic portrayal of Syrian refugees in 2015 and Ukrainian refugees in 2022 across seven European countries

Supplemental material, sj-docx-1-ejc-10.1177_02673231251394694 for Different refugees, different portrayals. A GDELT-based comparative analysis of the mediatic portrayal of Syrian refugees in 2015 and Ukrainian refugees in 2022 across seven European countries by Matteo Bassoli and Tommaso Greppi in European Journal of Communication

Supplemental Material

sj-sav-2-ejc-10.1177_02673231251394694 - Supplemental material for Different refugees, different portrayals. A GDELT-based comparative analysis of the mediatic portrayal of Syrian refugees in 2015 and Ukrainian refugees in 2022 across seven European countries

Supplemental material, sj-sav-2-ejc-10.1177_02673231251394694 for Different refugees, different portrayals. A GDELT-based comparative analysis of the mediatic portrayal of Syrian refugees in 2015 and Ukrainian refugees in 2022 across seven European countries by Matteo Bassoli and Tommaso Greppi in European Journal of Communication