Online reviews have become a key source of data concerning customer experience for restaurants and are relevant for a business’s competitiveness and positioning. The present study analyzes users’ perceptions of the restaurants featured on the Catalan television show Joc de Cartes, broadcast by TV3, by examining both user-generated content (Google reviews) and brand-generated content (Instagram posts and comments).
The dataset includes over 24,000 reviews and 5,900 social media comments from 909 official posts. We use a mixed-methods approach that combines qualitative and quantitative analyses of sentiment and emotion, powered by large language models, to evaluate how viewers respond to different aspects of the dining experience, food, service, atmosphere and price, both before and after each episode.
Results show a general predominance of positive sentiment, but also a notable spike in negative emotions after episodes air, particularly in user comments. Posts with negative sentiment, although rare, tend to generate the highest engagement. These findings suggest that television exposure influences restaurant reputation but must be complemented by active digital communication strategies.
The findings highlight the need for restaurants to adopt proactive digital communication and active social media monitoring. For broadcasters, maintaining consistent online content and using emotionally engaging formats is key to sustaining audience interaction and protecting brand reputation.
The study offers practical insights for restaurants and broadcasters on managing online reputation and audience engagement.
1. Introduction
According to Li et al. (2025), the use and impact of social media are essential for creating a memorable customer experience, particularly in sectors such as restaurants. Users of social networks enhance the perceived quality, satisfaction, and value of products by sharing their opinions, emotions, and experiences (Martí-Ochoa et al., 2024). Social media platforms also serve as influential reference tools, playing a critical role in decision-making processes, such as consumers’ intentions to visit a restaurant (Neger and Aorna, 2025). In this scenario, cooking-related shows have become very popular TV programs worldwide (Anguera-Torrell et al., 2025), for example, shows where different restaurants compete to be crowned the best in the program. In these cooking shows, winning restaurants gain high visibility and audience attention, making them particularly suitable for examining the dynamics of brand-generated content (BGC), user-generated content (UGC), and post-broadcast online reactions. This research is focused on Joc de Cartes. Since its debut in 2017, Joc de Cartes has been broadcast continuously on the public regional television channel, TV3, in Catalonia, Spain. In each episode, four restaurants from the same region or with similar culinary concepts compete to determine which is the best. The restaurants (usually their owners) evaluate each other’s establishments, while the host provides the final deciding vote. Participants are assessed based on cuisine, menu, service, pricing, and atmosphere. The winning establishment receives a prize of 5,000 euros. The format is adapted from the German show My Restaurant Rocks, which first aired in 2013 (3 Cat, n.d.).
This restaurant-based reality television show has established itself as one of the most popular programs on the network, consistently achieving high viewing figures, with an average audience share of 22.3% and a 24.6% share among the key commercial demographic (ages 30–55) (3 Cat, n.d.). In 2022, it received the Ondas Award for Best Program Aired by a Regional Broadcaster or Channel. As of 2025, the show is entering its ninth season, having featured over 500 restaurants and covered more than 25,000 kilometers across the region (3 cat.cat, 2024).
Considering the above, previous research has analyzed the impact of cooking shows on audiences, but none have studied how diners perceive the criteria used to assess the winning restaurants. Therefore, the objective of this study is to analyze diners’ perceptions regarding the criteria evaluated in the winning restaurants on Joc de Cartes, broadcast by TV3. In this line, it is important to mention that this sampling strategy therefore aligns with the study’s objective of analyzing how televised exposure influences digital engagement and public perception.
2. Literature review
2.1 Television cookery shows and restaurants
Television programs serve as a credible source of information for potential customers (Sung et al., 2020) and function as a significant promotional tool for gaining a competitive advantage in the hospitality industry (Zopiatis and Melanthiou, 2019), as audiences tend to favor familiar themes and recognizable places (Wayne and Castro, 2025). At the same time, media exposure can shape opinions and behaviors through priming (Neyens and Smits, 2017), defined as the activation of mental constructs by external stimuli, either consciously or unconsciously (Weingarten et al., 2016). Visually appealing depictions of dishes or spaces can, for example, prime interest in specific foods or places. In that line, television cookery shows are not only a medium for demonstrating recipes but also a tool for fostering social connection with viewers (Vilani et al., 2015). This direct engagement can influence dining intentions (Gajdzik et al., 2023), reduce consumer uncertainty by showcasing hygiene and food safety practices (Ovca et al., 2024), and shape audience evaluations (Anguera-Torrell et al., 2025).
Additionally, the popularity of TV cookery shows, and restaurants’ consequent visibility, are shaped by the presence and influence of celebrity chefs who act as cultural intermediaries (Bourdieu, 1984). Their culinary cultural capital has a significant impact on various food-related issues, including traditional cuisine, emerging food trends, and ethical and sustainable practices (Giousmpasoglou et al., 2020).
2.2 Criteria used to evaluate the restaurants
One of the main objectives for the restaurant industry is to achieve memorable customer experiences (Noguer-Juncà and Fusté-Forné, 2024) since it is regarded as the measurement of restaurant quality, as perceived by customers (Jeong and Jang, 2011). The creation of such experiences involves the meeting of sensorial, emotional, cognitive, behavioral and other stimuli (Bonfanti et al., 2025). These stimuli, in turn, elicit cognitively, emotionally, and behaviorally favorable, neutral, or unfavorable customer responses. These stimuli are based on the customer’s prior experience and/or pre-formed expectations (Walter et al., 2010). That is, customer experience is a key source of competitive advantage in the restaurant industry (Junkrachang et al., 2021).
Mathayomchan and Taecharungroj (2020), in their analysis of the impact of core restaurant attributes on customer experience across rating levels, highlighted that food emerges as the key element driving higher ratings; service and atmosphere play a more critical role in shaping lower and mid-level ratings; alcoholic beverages and dietary-specific offerings positively influence customer satisfaction and ratings; and the impact of value for money on satisfaction varies depending on the dining context and type of restaurant. According to these authors, restaurant attributes are based on four primary dimensions: food, service, atmosphere, and value.
2.2.1 Food
Various scholars have demonstrated that taste, presentation, and ingredient quality significantly predict overall satisfaction and customer loyalty in restaurants (Namkung and Jang, 2007). According to Liu and Jang (2009), food quality encompasses attributes such as food quality, taste, variety, presentation, freshness and temperature, among others. Furthermore, the use of local ingredients reinforces the sense of place (Bessière, 2013; Noguer-Juncà et al., 2021), supports community ties (Minton et al., 2018; Crespi-Vallbona and Noguer-Junca, 2024), and showcases regional authenticity (Fusté-Forné and Noguer-Juncà, 2024). In this context, practices like “Kilometer 0” support community resilience and local economies (Noguer-Junca and Fusté-Forné, 2023), but also help to reduce food waste in supply chains and lower greenhouse gas emissions (Zhang et al., 2025). Consequently, in recent years, local sourcing has gained importance as a socially responsible practice (Talukder et al., 2024).
2.2.2 Service
Although service quality is subjective, it is shaped by personal experience and by measurable attributes (Louzao and Crespi-Vallbona, 2022). These include both tangible and intangible aspects such as food quality, hygiene, empathy, hospitality and billing processes (Vu et al., 2019).
Bitner (1991) identified premises, staff responses and materials as core dimensions shaping satisfaction and perceived quality. She later differentiated service quality into functional and technical components. Zeithaml et al. (1993) proposed five service quality dimensions: assurance, reliability, empathy, tangibility and responsiveness. In any case, the perception of service value plays a crucial role in consumers’ purchase decisions, particularly in food and beverage services (Crespi-Vallbona et al., 2023).
2.2.3 Atmosphere
Atmosphere encompasses both physical and sensory elements (Mathayomchan and Taecharungroj, 2020). It refers to the physical environment, including surroundings, facilities, ambiance, design, decoration, music and lighting. The atmosphere aims to trigger sensory perceptions and emotional responses that enhance purchase intentions (Şahin and Yazıcıoğlu, 2025) and encourage revisit intentions (Bichler et al., 2021). A positive dining atmosphere increases the perceived value of the food and improves the overall customer experience, which in turn raises willingness to pay and influences customers to stay longer (Meiayawowor and Kurniawan, 2025).
2.2.4 Value or price
Value for money is frequently used to analyze customer satisfaction (Gupta et al., 2023); it is also closely linked to product quality and variety (Steptoe et al., 1995) as well as trust and affordability (Donaher and Lynes, 2017). Mathayomchan and Taecharungroj (2020), note that value for money has a context-dependent effect, influencing satisfaction differently depending on the type of restaurant and the experience.
2.3 Social media in gastronomy
Technological innovations have rapidly transformed marketing practices in recent decades (Law and Chen, 2025). According to Internet World Stats (July 2025), global internet use reached 5.65 billion people (68.7% of the world population), an increase of 146 million users in one year. Of these, 5.24 billion were social media users (Statista, 2025). This widespread access to the Internet has changed how consumers search for and access information and how they share experiences and feedback with others. Previous research has analyzed this contribution of social media as a trustworthy marketing tool (Hanaysha, 2022).
Media outlets increasingly focus on areas that strongly influence public life, such as food and drink (Elshater and Abusaada, 2024). Kristensen and From (2012) remark that modern food coverage goes beyond simple guidance and now includes advice, recipes, reviews, and expressions of taste and lifestyle. Food is presented not only as cultural or gastronomic content, such as in restaurant reviews, but also as a lifestyle symbol, similar to fashion. As a result, social media act as both an influence and a source of inspiration and motivation (Marti-Ochoa et al., 2025a).
Exposure to positive content on social platforms enables individuals to evaluate the quality of products and services more effectively (Armutcu et al., 2024). Limited or no exposure, by contrast, reduces their ability to make informed consumption decisions (Yu and Lee, 2019). That is, customer satisfaction is a key predictor of behavioral intentions toward a restaurant (Jani and Han, 2011) which could strongly influence diners’ restaurant choices (Dinç, 2023) and allow diners to take an active role in co-creating their dining experience (Alotaibi, 2025).
Sung et al. (2020) have explored how information shared about restaurants featured in competitive television programs influences audience perceptions. They found that high information quality significantly enhances consumers’ attitudes toward restaurants, and that positive attitudes increase visit intentions. In the same line, Anguera-Torrell et al. (2025) created two fictitious versions of restaurants appearing on the Joc de Cartes television show to identify the causal effect of media exposure on online ratings. Their results indicate that the show increases the number of online reviews but does not affect the average rating, yielding critical implications for both researchers and practitioners. These researches shows that online engagement can increase consumer loyalty, satisfaction, trust, commitment, word of mouth, and value co-creation.
In parallel, recent advancements in AI-powered text analytics have enriched the methodological landscape for studying online reviews and social media in tourism and hospitality (Belal et al., 2023). Although machine learning and LLMs support scalable sentiment and emotion analysis, their use in empirical, multi-channel, and mixed-method studies remains limited. However, a notable gap remains in understanding the impact of television shows on online public opinion and its online brand management and positioning. This paper addresses that gap by analyzing diners’ perceptions of TV cookery shows regarding the criteria evaluated in the restaurants, using Joc de Cartes as a case study. The methodology follows a multi-channel approach that combines quantitative and qualitative analysis of Google Reviews and Instagram data.
3. Data and methodology
All data analyzed in this study were obtained from public sources, specifically Google My Business reviews and posts and comments from public Instagram profiles. No personally identifiable information (PII) such as names, user IDs, or IP addresses was collected or stored. Data collection was limited to publicly visible text available on the platforms, whether in the form of user-written reviews or captions and comments published on open accounts. No contact was made with the authors of the content, nor were any attempts made to identify them in any way. Therefore, individual informed consent was not required, as the research relied exclusively on publicly available content that was anonymized for analysis purposes. All data were handled solely for academic purposes, securely stored, and used in accordance with the terms of service of the digital platforms involved. The study posed no risk to users, as it did not involve direct interaction or the collection of private information.
3.1 Quantitative analysis
This study draws on two types of data. On the one hand, Google Reviews by reviewers were collected for the entire population of restaurants (n = 116) that participated in each of the eight seasons of the television show. Because the study aims to analyze the program’s specific impact, the dataset includes all participating establishments rather than a sample, ensuring complete representativeness of the televised phenomenon. Google Reviews was chosen over TripAdvisor because it accumulates a higher volume of recent and diverse reviews (Mellinas and Sicilia, 2024). In parallel, all the text content from the show’s official Instagram account, including user comments, was also downloaded. Instagram was selected as the primary social media platform for analysis because it is where the show has the largest digital following, with over 59,300 followers, significantly more than on Facebook, where it has approximately 20,000 followers. This makes Instagram the most representative channel for assessing the brand’s communication strategy and user engagement.
3.1.1 Google My Business reviews
The review data was downloaded in April 2025 using Octoparse web scraping software. The final dataset constitutes a census of the show’s participants, including a total of 116 restaurants and 24,120 customer reviews, spanning from 2013 to 2025, thus covering a 12-year period of analysis. This extensive timeframe was chosen to ensure a robust longitudinal comparison; by including data from 2013, the study establishes a baseline of the restaurants’ organic digital reputation prior to their media exposure. During the data cleaning process, 1,389 reviews were excluded from the temporal analysis because they lacked a verifiable publication date, which was essential for the pre- versus post-broadcast comparison. Consequently, the final comparative analysis was conducted on the remaining 22,731 reviews for which complete chronological data was available.
The downloaded dataset already contained three key variables: the date of the review, the text content, and the rating. Each restaurant was manually assigned to the corresponding season in which it appeared on the show. To isolate the impact of the television program, a temporal difference was calculated by subtracting the episode’s original broadcast date from the review’s publication date. Additionally, each review was classified as being written either before or after the broadcast of the episode in which the restaurant featured. This segmentation allows for a longitudinal comparison, evaluating the “pre-broadcast” organic feedback against the “post-broadcast” viewer-driven sentiment.
An automated sentiment and emotion analysis was conducted on each review using a workflow built on the low-code platform N8N. This tool enables users to design and automate processes visually with minimal coding, facilitating the integration of multiple services. In this case, Excel data files were connected to AI agents for processing through a visual interface. The setup reflects the shift from Robotic Process Automation (RPA) to Agentic Process Automation (APA). RPA focuses on automating repetitive, rule-based tasks using software “robots”, improving efficiency in routine operations. APA expands on this by incorporating AI-driven agents capable of handling both repetitive and cognitive tasks, including decision-making and learning. This allows more complex workflows to be delegated to digital systems. Additionally, low-code automation simplifies the handling of large-scale textual data and makes advanced AI techniques accessible to research environments with limited technical resources, enhancing overall organizational agility and performance (Kitsantas et al., 2024).
The workflow displayed in Figure 1 integrates various tools and is triggered automatically when the review processing flow begins. First, the data is retrieved from a Google Sheets document where all the reviews are stored. Then, a limit is set to define how many entries will be processed in each run, and a loop iterates through each review individually. Each review is sent to an AI agent that uses the Google Gemini 2.0 Flash model, hosted on the OpenRouter platform (OpenRouter, 2025). This model is responsible for interpreting and analyzing the content of the reviews and generating a response based on a predefined prompt:
The flowchart begins with a trigger labeled When clicking Test workflow. This leads to a step labeled Google Sheets 3, which reads a sheet. The next step is labeled Limit, followed by Loop Over Items 2. The loop processes items and can either continue looping or proceed to the next step. The next step involves an AI Agent that uses the OpenRouter Chat Model for analysis. The AI Agent can interact with memory and tools. If no operation is needed, the flow proceeds to No Operation, do nothing 2. If an operation is needed, it updates a sheet in Google Sheets 1. The flowchart includes conditional paths and loops, with clear labels and annotations for each step.Automated workflow for review data processing and AI analysis. Source: Authors’ own work
The flowchart begins with a trigger labeled When clicking Test workflow. This leads to a step labeled Google Sheets 3, which reads a sheet. The next step is labeled Limit, followed by Loop Over Items 2. The loop processes items and can either continue looping or proceed to the next step. The next step involves an AI Agent that uses the OpenRouter Chat Model for analysis. The AI Agent can interact with memory and tools. If no operation is needed, the flow proceeds to No Operation, do nothing 2. If an operation is needed, it updates a sheet in Google Sheets 1. The flowchart includes conditional paths and loops, with clear labels and annotations for each step.Automated workflow for review data processing and AI analysis. Source: Authors’ own work
Instructions:
You are a text analysis specialist. Your task is to analyze the following text and identify:
The sentiment polarity (positive, neutral, or negative),
A sentiment score ranging from −1 to +1, where −1 represents extremely negative sentiment, 0 represents neutral sentiment, and +1 represents extremely positive sentiment, and
The predominant emotion expressed in the text.
The predominant emotion must be selected from the following list only: neutral, fear, sadness, anger, disgust, surprise, joy.
Output format:
Return the result in a single line using the following structure:
[sentiment polarity], [sentiment score], [predominant emotion]
Example output:
positive, +1, joy.
Do not include explanations, comments, or any additional text. Return only the result.
Depending on the output, the workflow either updates the Google Sheet with the processed information or, if there are no changes, performs no additional actions. Following the automated analysis, a manual inspection was performed to verify the consistency of sentiment and emotion tags and to ensure interpretative alignment with the platform’s conversational and expressive language. Subsequently, the data was processed using SPSS statistical software.
3.1.2 Instagram content
To strategically utilize the brand content associated with the TV show, we employed a specialized tool for data collection. Apify (2025), a web scraping tool, has also been used in other studies analyzing social networks (Marti-Ochoa et al., 2025b; Yeung et al., 2023) to obtain all the information from the posted text.
The “Instagram Profile Scraper” developed by the Apify platform was employed to extract data from the official Instagram account of the television show @JocdeCartestv3. Data was collected on July 1st, 2025, and included all posts published up to that date. For each post, the following variables were retrieved: post text, format (image, carousel, or video), video duration, publication date, number of likes, number of comments, and the profile’s follower count at the time of extraction. In addition, the same API was used to obtain all user comments associated with each post. In total, 909 posts with 4,950 comments were retrieved. The posts span from 2017 to 2023, when the final publication on social media marked the end of the sixth season. An engagement rate was calculated for each post using the following formula: (likes + comments) / followers × 100 (see Ryhänen (2019), and the results were included in the dataset for further analysis.
The workflow shown in Figure 2 follows the same structure as in the case of the review analysis; the only difference lies in the natural language model used, which in this case is ChatGPT-5. The same prompt was applied as in the review comments, and the output was exported to an Excel file for further processing using pivot tables and ANOVA analysis.
The flowchart begins with a Schedule Trigger and a manual trigger labeled 'When clicking Test workflow'. It then proceeds to a Code block, followed by a Google Sheets component labeled 'read sheet'. The data flows into a Loop Over Items block, which splits into a Filter and an OpenRouter Chat Model. The Filter directs data to a Basic LLM Chain, which processes the data and sends it to a Structured Output Parser. The parsed data is then sent to another Google Sheets component labeled 'appendOrUpdate sheet'. The flowchart illustrates the automated process of reading data from a Google Sheet, processing it through various AI models, and updating the Google Sheet with the results.Automated workflow for social media data processing and AI analysis. Source: Authors’ own work
The flowchart begins with a Schedule Trigger and a manual trigger labeled 'When clicking Test workflow'. It then proceeds to a Code block, followed by a Google Sheets component labeled 'read sheet'. The data flows into a Loop Over Items block, which splits into a Filter and an OpenRouter Chat Model. The Filter directs data to a Basic LLM Chain, which processes the data and sends it to a Structured Output Parser. The parsed data is then sent to another Google Sheets component labeled 'appendOrUpdate sheet'. The flowchart illustrates the automated process of reading data from a Google Sheet, processing it through various AI models, and updating the Google Sheet with the results.Automated workflow for social media data processing and AI analysis. Source: Authors’ own work
3.2 Qualitative analysis
To analyze the impact of Joc de Cartes on public opinion and online brand management, a qualitative study was conducted by two authors and validated by the rest of the team. The analysis followed a qualitative content approach, examining language through thematic categorization (Neuendorf, 2002). It was based on key evaluation criteria identified in the literature review: food, service, atmosphere, and price. The coding scheme was refined iteratively through constant comparison between data and categories, allowing core themes to emerge. An intentional sampling strategy guided by predefined criteria was applied. Data collection and analysis continued until thematic saturation was reached, defined as the point at which no new codes or meaningful themes appeared (Guest et al., 2006).
Beyond offering a descriptive account of the featured restaurants (e.g. types of products and dishes), the analysis also uncovers the constructed relationships and symbolic meanings attached to TV show cooking as a marketing instrument.
4. Results
4.1 Reviews
As shown in Table 1, the distribution of sentiment categories in Google Reviews differs slightly before and after the broadcast of the television program, while maintaining a generally positive overall evaluation.
Sentiment distribution and mean rating
| Reviews | Mean rating | SD rating | Mean polarity | SD polarity | |
|---|---|---|---|---|---|
| Before | |||||
| Positive | 82.11% | 4.7 | 0.53 | 0.86 | 0.12 |
| Neutral | 2.20% | 3.37 | 0.72 | 0.12 | 0.03 |
| Negative | 15.70% | 1.88 | 1.02 | (-0.65) | 0.22 |
| Total | 100,00% | 4.01 | 1.21 | 0.57 | 0.57 |
| After | |||||
| Positive | 79.72% | 4.77 | 0.49 | 0.87 | 0.12 |
| Neutral | 2.53% | 3.49 | 0.77 | 0.11 | 0.02 |
| Negative | 17.75% | 1.99 | 1.09 | −0.64 | 0.22 |
| Total | 100,00% | 4.25 | 1.24 | 0.59 | 0.6 |
Before the broadcast, most reviews were classified as positive (82.11%), with a high mean rating of 4.70 (SD = 0.53) and a strong positive mean polarity of 0.86, indicating a clear alignment between favorable textual sentiment and numerical ratings. Negative reviews accounted for 15.70% of the total, with a substantially lower mean rating of 1.88 (SD = 1.02) and a negative polarity score of −0.65, confirming consistency between negative sentiment and low ratings. Neutral reviews accounted for only 2.20% of the sample and showed intermediate values, with a mean rating of 3.37 and a polarity score of 0.12. Overall, the average rating before the broadcast was 4.01, with a mean polarity of 0.57, reflecting a generally positive perception prior to the program’s airing.
After the broadcast, positive reviews remained dominant, although their proportion decreased slightly to 79.72%. These reviews showed an even higher mean rating of 4.77 (SD = 0.49) and a similar polarity score of 0.87. The proportion of negative reviews increased modestly to 17.75%, with a mean rating of 1.99 (SD = 1.09) and a polarity score of −0.64. Neutral reviews continued to account for a very small share of the dataset (2.53%), with a mean rating of 3.49 and a polarity of 0.11. Importantly, the overall mean rating increased to 4.25 after the broadcast, along with a slightly higher mean polarity of 0.59, suggesting an improvement in the overall evaluation of the restaurants following their appearance on the television program.
As shown in Table 2, a chi-square test of independence was conducted to examine whether the distribution of emotions expressed in Google reviews differed depending on whether they were written before or after the television program was aired. The analysis revealed a significant association between emotion and time of review, χ2(6, N = 22,731) = 20.93, p < 0.05. This indicates that the emotional content of the reviews was unevenly distributed across the two time periods. In particular, the categories Sadness (row contribution = 11.01) and Disgust (row contribution = 4.31) contributed most to the overall chi-square statistic, suggesting that these emotions deviated most from the expected values based on independence. Thus, the findings reveal that the emotional tone of audience reviews on Google was influenced by whether they were posted before or after the broadcast of the TV program, especially with regard to negative sentiment.
Chi-square test results for the relationship between the emotions of the review written before and after broadcasting
| Emotion | Observed (Before) | Observed (After) | Total | Expected (Before) | Expected (After) | Contribution (Before) | Contribution (After) | Row χ2 contribution |
|---|---|---|---|---|---|---|---|---|
| Anger | 287 | 1,154 | 1,441 | 288.76 | 1152.24 | 0.01 | 0.0 | 0.01 |
| Disgust | 164 | 782 | 946 | 189.57 | 756.43 | 3.45 | 0.86 | 4.31 |
| Fear | 6 | 15 | 21 | 4.21 | 16.79 | 0.76 | 0.19 | 0.95 |
| Joy | 3,754 | 14,539 | 18,293 | 3665.68 | 14627.32 | 2.13 | 0.53 | 2.66 |
| Neutral | 80 | 366 | 446 | 89.37 | 356.63 | 0.98 | 0.25 | 1.23 |
| Sadness | 264 | 1,317 | 1,581 | 316.81 | 1264.19 | 8.8 | 2.21 | 11.01 |
| Surprise | 0 | 3 | 3 | 0.6 | 2.4 | 0.6 | 0.15 | 0.75 |
| Total | 4,555 | 18,176 | 22,731 | 4,555 | 18,176 | 16.73 | 4.19 | 20.93 |
Note(s): χ2(6, N = 22,731) = 20.93, p < 0.05. There is a significant association between emotion and time of emission
4.2 Social media
An examination of engagement metrics by sentiment and emotional categories, as detailed in Table 3 (individual emotional subcategories omitted where non-significant), highlights a clear dominance of positively framed content on the official Instagram profile of the television show, with 79.29% of posts categorized under the emotion joy. This type of content not only constitutes the majority of posts but also generates a relatively high average engagement rate (1.12) and a strong positive sentiment polarity (0.61). Similarly, user comments reflect a generally favorable reception, with 46.81% of them associated with joy and an even higher polarity score (0.64), indicating a coherent and effective communication strategy that resonates well with the audience.
Sentiment, emotional analysis and engagement Metrics in Instagram Content
| Official profile | Comments | ||||
|---|---|---|---|---|---|
| Sentiment/emotions | Engagement | Polarity | Sentiment/emotions | Polarity | |
| Negative | 4.10% | 1.35 | −0.47 | 28.81% | −0.58 |
| Anger | 1.22% | 0.82 | −0.47 | 17.88% | −0.59 |
| Disgust | 1.66% | 1.90 | −0.48 | 7.60% | −0.57 |
| Fear | 0.55% | 1.51 | −0.40 | 0.24% | −0.30 |
| Sadness | 0.66% | 0.80 | −0.50 | 3.09% | −0.53 |
| Neutral | 16.61% | 1.21 | 0.06 | 24.38% | 0.01 |
| Neutral | 14.17% | 1.15 | 0.00 | 23.88% | 0.00 |
| Surprise | 2.44% | 1.59 | 0.40 | 0.51% | 0.32 |
| Positive | 79.29% | 1.12 | 0.61 | 46.81% | 0.64 |
| Joy | 79.29% | 1.12 | 0.61 | 46.81% | 0.64 |
| Total | 100.00% | 1.15 | 0.48 | 100.00% | 0.13 |
Interestingly, although negative content represents only 4.10% of brand-generated posts, it registers the highest engagement rate (1.35), suggesting that emotionally charged negative content may stimulate greater user interaction despite its limited presence. Notably, negative emotions such as anger and disgust appear more frequently in user-generated comments (17.88% and 7.60%, respectively) than in official posts (1.22 and 1.66%), with consistently low polarity scores of around −0.57.
This emotional gap between brand-generated content (BGC) and user-generated content (UGC) suggests a potential dissonance between how the show is perceived and how it is promoted. While the brand maintains a predominantly joyful tone, audience responses reveal more critical or emotionally negative reactions that may influence the overall reputation of the program. Finally, neutral content appears less frequently in both posts and comments and is associated with lower engagement rates, reinforcing the idea that emotionally expressive content, positive in particular, tends to drive greater interaction on social media platforms.
Table 4 presents the results of one-way ANOVA tests conducted to examine differences in the average engagement rate across three dimensions: content type, predominant emotion, and sentiment polarity. Content type proved to be a significant factor (F = 4.24; p < 0.001), with video posts generating the highest average engagement (M = 1.24), followed by carousels (M = 1.12) and static images (M = 0.96). In contrast, no statistically significant differences were found in engagement when comparing the predominant emotions detected in the post text (F = 1.48; p = 0.193) or sentiment polarity (F = 0.78; p = 0.457). These results suggest that emotional tone alone does not systematically determine engagement levels in this context. Although negatively framed posts showed higher interaction rates, the ANOVA results indicate that sentiment polarity and predominant emotion were not statistically significant predictors of engagement. Instead, structural factors such as media format and temporal evolution appear to exert a stronger influence on audience interaction patterns. This finding highlights the complexity of engagement dynamics, where affective intensity may attract attention in specific cases but does not operate as a consistent explanatory variable.
One-way ANOVA results for engagement rate across years, content types, emotions, and sentiment polarity
| Variable | Dimension | Mean | N | Std. Deviation | F | p-value |
|---|---|---|---|---|---|---|
| Type | Image | 0.96 | 265 | 1.24 | 4.24 | <0.001 |
| Carousels | 1.12 | 113 | 0.74 | |||
| Videos | 1.24 | 531 | 2.06 | |||
| Emotion | Anger | 0.82 | 11 | 0.24 | 1.48 | 0.193 |
| Disgust | 1.90 | 15 | 4.11 | |||
| Fear | 1.51 | 5 | 3.11 | |||
| Joy | 1.12 | 716 | 1.53 | |||
| Neutral | 1.15 | 128 | 1.49 | |||
| Sadness | 0.80 | 6 | 0.28 | |||
| Sentiment | Negative | 1.35 | 37 | 2.32 | 0.78 | 0.457 |
| Neutral | 1.21 | 150 | 2.23 | |||
| Positive | 1.12 | 716 | 1.53 |
Before proceeding with post-hoc comparisons, we applied Levene’s test for homogeneity of variances, which indicated that this assumption was violated. Therefore, the Games–Howell procedure was applied for the year of post variable. The results, shown in Table 5, confirm that engagement levels were significantly higher in more recent years, particularly in 2022 (M = 2.15) and 2023 (M = 2.39), compared to earlier periods such as 2017 or 2018. This upward trend may reflect the effects of digital audience consolidation, increasingly effective content strategies, or stronger organic interaction within the program’s online community.
Games–Howell post-hoc comparisons of engagement by years
| Year of the post (i) | Year of the post (j) | Mean (i-j) | Std.Error | Sign. |
|---|---|---|---|---|
| 2017 | 2018 | (−0.25) | 0.12 | * |
| 2019 | (−0.66) | 0.14 | * | |
| 2020 | (−0.89) | 0.17 | * | |
| 2021 | (−1.03) | 0.13 | * | |
| 2022 | (−1.89) | 0.12 | * | |
| 2023 | (−2.13) | 0.23 | * | |
| 2018 | 2017 | 0.25 | 0.12 | * |
| 2019 | (−0.41) | 0.13 | * | |
| 2020 | (−0.64) | 0.15 | * | |
| 2021 | (−0.78) | 0.11 | * | |
| 2022 | (−1.64) | 0.1 | * | |
| 2023 | (−1.88) | 0.22 | ** | |
| 2019 | 2022 | (−1.24) | 0.13 | * |
| 2023 | (−1.48) | 0.23 | * | |
| 2020 | 2022 | (−1.00) | 0.15 | * |
| 2023 | (−1.24) | 0.25 | * | |
| 2021 | 2022 | (−0.86) | 0.12 | * |
| 2023 | (−1.10) | 0.23 | * | |
| 2022 | 2017 | 1.89 | 0.12 | * |
| 2023 | (−0.24) | 0.22 | * | |
| 2023 | 2017 | 2.13 | 0.23 | * |
Note(s): * sig: <0.001; ** sig: <0.05
4.3 Qualitative content analysis
The initial content analysis shows that the most prominent words include food, service, place, and menu. Other frequent terms are good, excellent, recommended, price, friendly, delicious, quality, atmosphere, and staff. This reflects that users mainly comment on food, service, atmosphere, and value for money.
Focusing on these most frequent words, it is observed that food is used 10,444 times; service is used 8,482 times; atmosphere is used 2,084 times; and price is used 4,358 times. Food is the most mentioned word in positive reviews, but it also appears frequently in negative reviews. Service is greatly present in negative reviews. This suggests that it is a critical factor linked to unsatisfactory experiences. Atmosphere is mentioned more in positive reviews. This indicates that it usually contributes to a good experience. Price appears in all types of sentiment. It stands out in negative reviews, possibly due to perceptions of overpricing (see Figure 3).
A bar graph compares the frequency of the most used keywords according to content analysis. The horizontal axis is labeled Palabra clave and includes the categories food, service, atmosphere, and price. The vertical axis is labeled Frecuencia and ranges from 0 to 8.02k. The graph features three vertical bars for each keyword category, representing Positive, Neutral, and Negative sentiments. The color scheme includes blue for Positive, orange for Neutral, and green for Negative. For food, the frequencies are 8k for Positive, 1.6k for Neutral, and 335 for Negative. For service, the frequencies are 6.7k for Positive, 1.4k for Neutral, and 331 for Negative. For atmosphere, the frequencies are 1.6k for Positive, 428 for Neutral, and 36 for Negative. For price, the frequencies are 2.6k for Positive, 572 for Neutral, and 127 for Negative.Frequency of the most used keywords according to the content analysis. Source: Authors’ own work
A bar graph compares the frequency of the most used keywords according to content analysis. The horizontal axis is labeled Palabra clave and includes the categories food, service, atmosphere, and price. The vertical axis is labeled Frecuencia and ranges from 0 to 8.02k. The graph features three vertical bars for each keyword category, representing Positive, Neutral, and Negative sentiments. The color scheme includes blue for Positive, orange for Neutral, and green for Negative. For food, the frequencies are 8k for Positive, 1.6k for Neutral, and 335 for Negative. For service, the frequencies are 6.7k for Positive, 1.4k for Neutral, and 331 for Negative. For atmosphere, the frequencies are 1.6k for Positive, 428 for Neutral, and 36 for Negative. For price, the frequencies are 2.6k for Positive, 572 for Neutral, and 127 for Negative.Frequency of the most used keywords according to the content analysis. Source: Authors’ own work
However, the use of ATLAS.ti for a systematic coding and categorization process resulted in five main themes of analysis or code families, reflecting the dominant dimensions of the customer experience: food quality; service and customer care; atmosphere and place; perceived value and price; and finally, loyalty and recommendation. Code frequencies, densities, and co-occurrence patterns were examined to identify central themes and relationships among them (Table 6).
Code families, number of codes, quotations and density
| Code family (theme) | Number of codes | Number of quotations (n) | Percentage of total quotations (%) | Code density |
|---|---|---|---|---|
| Food quality | 5 | 9,842 | 40.8 | 0.78 |
| Service and customer care | 5 | 6,214 | 25.7 | 0.74 |
| Atmosphere and place | 4 | 3,126 | 13.0 | 0.46 |
| Perceived value and price | 3 | 2,041 | 8.5 | 0.51 |
| Loyalty and recommendation | 3 | 2,877 | 11.9 | 0.69 |
| Total | 20 | 24,100 | 100 | – |
The Food quality code family showed the highest frequency and density, highlighting its central role in participants’ narratives. Codes related to taste, dish quality, and menu evaluation appeared consistently and often co-occurred with themes such as Service quality and Will return, indicating that food quality is the core evaluative dimension of the overall experience. Analysis of reviews mentioning food revealed that 87% were positive, emphasizing quality, flavor, and presentation. Only 8.5% were neutral and 4.2% negative. Negative mentions are very scarce and usually relate to service or hygiene issues. As some users pointed out, the experience can range from highly positive (“I love it. A very correct menu. Good value for money. Good food. Good service”) to extremely negative (“At one in the afternoon, they didn’t even have a cold beer. It smells like fried food”).
Most complaints relate to high prices and poor value for money. While food generally evokes positive emotions, price remains more contentious and attracts negative criticism. Common negative terms include expensive, slow, cold, dirty, and rude, reflecting key issues such as high prices, poor service, food problems, and hygiene concerns.
The Service and customer care family showed high frequency and strong co-occurrence with food-related codes. Participants highlighted staff friendliness, professionalism, and attentiveness, with service acting as a catalyst that enhances perceived value and loyalty. Negative service reviews focused on waiting times and staff attitude. Positive comments often linked good service with fair pricing, suggesting it improves cost perception. In contrast, poor service was associated with complaints about high or unjustified prices, especially when delays or lack of attention occurred. For instance, one reviewer expressed frustration stating: “What an experience … This restaurant needs more organization. The pizza was very hard, the service was very slow, and I wanted to speak with the boss, and [he/she] had disappeared.” Overall, service quality directly shapes how price is perceived, even when food quality is high.
The Perceived value and price family emerged more implicitly, often in conjunction with references to menu quality. Co-occurrence analysis showed moderate links between perceived value and food-related codes, suggesting that customers assess price fairness primarily through the lens of quality received rather than cost alone. Service has many more positive mentions than negative ones, indicating that it is usually valued favorably. Price also has more positive mentions, but with a higher proportion of negative ones than “service”. This suggests that price is a more controversial or sensitive issue.
Codes related to Atmosphere and place showed moderate frequency and density, functioning as contextual elements that complemented the core dining experience. References to the physical environment, comfort, and location contributed to overall evaluations, though they were less central than food and service. The term atmosphere appeared positively in many reviews, with verbatims such as “The atmosphere is warm and welcoming” and “Lovely atmosphere with relaxing music.” Negative mentions were less frequent and typically linked to discomfort or unmet expectations, as reflected in comments like “The atmosphere was too loud and chaotic” and “Unpleasant atmosphere, not what we expected.” Overall, atmosphere carries a strong positive connotation, often reflecting not only the physical setting but also customers’ emotional responses, such as feeling relaxed, welcomed, or inspired. It plays a meaningful role in shaping the overall impression of the restaurant and influences behavioral intentions, frequently associated with statements like “we will come back.” This highlights its relevance for customer loyalty.
Finally, the Loyalty and recommendation family displayed high density despite a lower number of codes. Expressions of intention to return and recommendation co-occurred strongly with both food and service quality codes, indicating that loyalty emerges as an outcome of consistently positive core experiences. Table 7 summarizes these code families, individual codes, and frequencies.
Code families, individual code, and frequencies
| Code family (themes of analysis) | Code | Quotations (n) |
|---|---|---|
| Food quality | Food quality – positive | 3,214 |
| Good food | 2,487 | |
| Excellent dishes | 1,964 | |
| Taste | 1,327 | |
| Menu quality | 850 | |
| Service and customer care | Service quality | 2,341 |
| Friendly staff | 1,529 | |
| Good attention | 1,118 | |
| Professional service | 812 | |
| Well organized | 414 | |
| Atmosphere and place | Nice place | 1,204 |
| Good atmosphere | 846 | |
| Comfortable space | 613 | |
| Location | 463 | |
| Perceived value and price | Good value for money | 1,021 |
| Suitable price | 654 | |
| Menu value | 366 | |
| Loyalty and recommendation | Will return | 1,342 |
| Recommendation | 978 | |
| Top restaurant | 557 |
To sum up, food quality occupies a central position, reinforced by service quality. Together, these dimensions lead to overall satisfaction, which is in turn associated with intention to return and recommendation. Atmosphere and perceived value function as supportive contextual factors. The high frequency and density of food-related codes and their strong co-occurrence with loyalty-oriented codes, reinforce the notion that culinary excellence functions as the foundation upon which the overall dining experience is constructed.
4.3.1 Brutal (in English, amazing, awesome)
The term brutal in Catalan is the slogan of the TV cooking show. The word has become a catchword used by the presenter and celebrity chef, Marc Ribas, at the end of each recipe. Thus, this recurring exclamation contributes to the show’s identity and audience engagement.
The word “brutal” appears 281 times, confirming that it is a very frequent expression in reviews, generally with a positive meaning. Some examples: “Simply BRUTAL. exquisite dishes, splendid service …” “The brutal pizza”.
The word brutal is used as an emphatic compliment, especially to describe food, service, or memorable experiences. Thus, within the total positive reviews (82.39%), there are many compliments for dishes such as seafood rice, noodle dishes, escargots, croquettes, French toast, and homemade desserts. The service is also rated as brutal. The friendliness and professionalism of waiters and waitresses are highlighted. Some are mentioned by name, such as Cristian, Miriam, and Héctor. The atmosphere is also classified as brutal. Cozy, romantic, and family-friendly places are especially appreciated. Sea-facing terraces and rustic decor are often mentioned. Finally, several comments describe the overall experience as “brutal,” “spectacular,” and “unforgettable.”
5. Discussion and conclusions
5.1 Conclusions
The findings of this study highlight the major role of digital media in shaping public perceptions of dining experiences, reinforcing previous research on the influence of media narratives and online engagement (Dinç, 2023). Television cookery programs function not only as platforms for demonstrating recipes but also as a promotional tool (Zopiatis and Melanthiou, 2019) and as an instrument for fostering social engagement with audiences (Vilani et al., 2015). The effects of non-fictional programs (e.g. Joc de Cartes) are contingent upon the cultural proximity between the audience and the content, with viewers showing a preference for familiar themes and spaces (Wayne and Castro, 2025). At the same time, media exposure can modulate perceptions and behaviors through priming (Weingarten et al., 2016), influencing individuals’ motivations, expectations, and prejudices prior to dining at a restaurant. In addition, engagement levels varied significantly by year, with a marked increase in 2022 and 2023. This suggests that the program’s digital presence and content strategies have become more effective over time, potentially benefiting from audience consolidation and greater algorithmic visibility on social media platforms.
Posts on the program’s official Instagram profile that emphasize negative or controversial aspects of the restaurants achieve the highest engagement rates (1.35), despite accounting for only 4.10% of the total content. The majority of posts (79.29%) are framed positively and aim to showcase the strengths of the participating establishments. This confirms earlier evidence that emotionally charged information arouses stronger reactions from audiences (Sung et al., 2020), and suggests that critical content may function as a catalyst for user interaction in competitive culinary contexts.
In contrast, user-generated comments present a more polarized emotional landscape. Positive comments account for 46.81% of the total, while negative comments represent a substantial 28.81%. This gap between the positivity of official posts and the sentiment diversity expressed by users supports the view that social media platforms operate as marketing channels, as highlighted by Hanaysha (2022), but also as spaces for public discourse, where audiences articulate independent evaluations that may diverge from official narratives (Marti-Ochoa et al., 2025b). These findings highlight the importance of active social listening and adaptive digital communication strategies, two practices highlighted in recent studies on online co-creation and consumer engagement (Alotaibi, 2025).
The qualitative analysis deepens this understanding by identifying the evaluative dimensions most frequently referenced by users: food, service, atmosphere, and price. These criteria are widely recognized as being central to customer assessment in hospitality research (Mathayomchan and Taecharungroj, 2020). Food quality is a determinant of customer satisfaction and loyalty in restaurants (Namkung and Jang, 2007). In the same line, service value and quality amplify the positive effects of core food attributes (see Louzao and Crespi-Vallbona, 2022). Atmosphere and place codes are also consistent with Bitner’s (1992) concept of servicescapes, which posits that physical environments influence customer perceptions primarily through emotional and contextual mechanisms. However, as reflected in previous studies dealing with restaurant settings (Ryu and Jang, 2008), our results show that atmosphere contributes to pleasure and satisfaction but rarely outweighs core factors such as food and service quality. Finally, as Mathayomchan and Taecharungroj (2020) concluded, perceived value has a context-supportive effect, influencing satisfaction according to the type of restaurant and the experience.
Moreover, the frequent use of expressions such as “brutal” to describe highly positive experiences, as evidenced by Giousmpasoglou et al. (2020), highlights the relevance of celebrity chefs as cultural intermediaries who shape how users articulate their perceptions. This phenomenon suggests an emotional and symbolic connection with the program that goes beyond merely rating a restaurant. Therefore, online engagement is essential for reinforcing consumer satisfaction and loyalty (Mubdir et al., 2025), but also as an influence and a source of inspiration and motivation (Marti-Ochoa et al., 2025a).
5.2 Theorical implications
The findings of this study offer several theoretical contributions. Its primary strength lies in the integration of advanced methodological tools, particularly LLM-based analysis, within a multichannel mixed-method design that moves beyond single-source data. This approach enables a comprehensive examination of how televised cookery programs influence public opinion, digital reputation, and the online brand management of featured restaurants. Moreover, the hybrid analytical framework, combining quantitative summaries and rigorous statistical reporting with qualitative evidence from verbatim excerpts and word-cloud visualizations, allows for the simultaneous identification of broad patterns and nuanced customer experiences. From a theoretical perspective, the study advances the understanding of how User-Generated Content (UGC) and Brand-Generated Content (BGC) interact in shaping restaurant positioning and competitive advantage within digitally mediated marketing environments. This paper contributes to the literature by evidencing the growing influence of media exposure for businesses, especially those related to gastronomic experiences. The study explores the unique theme of capturing viewers’ perspectives on restaurants featured in a reality show; however, it is important to note that consumers’ online comments may differ from their actual dining experiences.
5.3 Practical implications
From a practical perspective, this study offers actionable insights for both participating restaurants and television content producers. For the establishments featured in Joc de Cartes, systematically analyzing online reviews and user-generated content is crucial for identifying strengths and areas for improvement. Such analysis enables restaurants to refine their positioning, enhance service quality, and adapt their value proposition, particularly given that online reviews shape perceived experiences and significantly influence consumers’ decision-making processes. For these establishments, it is essential to develop a parallel digital communication strategy to the television broadcast, engaging in active social media listening and anticipating potential waves of online reviews and user-generated content. Active and strategic engagement on social media, particularly on Instagram, can help manage public perception and mitigate the reputational risks associated with the emotional reactions triggered after the episodes air. For content producers and public broadcasters such as TV3, prioritizing video content and maintaining an active presence beyond the broadcast season aligns with empirical results indicating that audiovisual formats generate higher interaction levels and contribute to sustained brand loyalty. Maintaining consistent brand-generated content after the season concludes can strengthen long-term audience engagement and preserve the show’s relevance, while a dynamic, video-oriented publishing strategy remains central to effective digital communication.
5.4 Limitations and future research
One of the most notable limitations of this study lies in the fact that the analysis of the program’s official Instagram profile was limited to data collected until 2023, coinciding with the final post of season 6. Although six videos related to Joc de Cartes were identified on TV3’s general Instagram account, the absence of continued communication through the program’s dedicated profile may indicate an underutilized strategic channel. For future research, it is recommended to expand the analysis to other platforms where the show maintains a residual presence, such as Facebook and YouTube, and to incorporate visual and audio analysis techniques to explore multimodal content. Furthermore, focusing this research on reviews of the winning restaurants ensures a controlled and comparable set of cases. However, this sampling strategy may limit the generalizability of the findings to all participating restaurants. Future research could therefore examine reviews of non-winning participants and compare online reactions to winning versus non-winning establishments. Additionally, a limitation concerning the engagement rate should be noted: follower counts were retrieved on a single extraction date (July 1, 2025) and therefore applied uniformly to all posts regardless of their year of publication. Lastly, comparing this case with other gastronomy programs from different regions or countries would allow for broader generalizations and enrich the academic discussion on the relationship between media, emotions, and gastronomic reputation.
