Influencer Self-Disclosure Practices on Instagram: A Multi-Country Longitudinal Study

The social media economy is changing dramatically. In their early history (the mid-2000s), social networks were predominantly used for peer-to-peer communication and user-generated content that reflected people’s daily lives [11]. At that point, business models and monetisation opportunities were still largely undeveloped. The advertising potential of harnessing human attention in social surroundings saw the rise of individual Internet entrepreneurs who have since received the label of social media influencers. In this study, we refer to influencers as social media content creators who monetise online content through native advertising [16, 41].

To deepen the understanding of the evolving ecosystem of content monetisation on social media, we conducted a longitudinal study spanning 2010-2020 of 400 Instagram influencers from four distinct countries. We present an analysis of the professionalisation of their advertising activities, focusing on the evolution of influencer self-disclosure practices. For this purpose, we collected and analysed a dataset of more than one million posts by 50 mega and 50 micro-influencers each from Brazil, Germany, the Netherlands, and the US. These countries were selected based on their different regulatory approaches in terms of legal standards and enforcement, which we will discuss in subsection 2.1.

Our contributions include (1) a methodological approach to longitudinal data curation in the context of influencer selection, laying the foundations for more reproducible and robust research in the field of influencer studies; (2) empirical observations of the rise and decline in the volume of posts by these influencers over the years and the effect on disclosed ads; (3) a comparative analysis of different disclosure practices for Instagram posts, likely correlated with national regulation; (4) a characterisation of monetisation strategies in terms of the timing of posts and partnership with global brands; (5) a thorough analysis of the impact of disclosures on engagement; and (6) an investigation of the prevalence of undisclosed ads in our dataset and their effect on engagement. Our study provides a foundation for updating regulatory guidelines and developing automated tools to enhance transparency and compliance in influencer marketing by offering insights into how disclosure practices and content strategies impact audience engagement. Ultimately, these observations can inform legislation and public policy on strategies that may lead to changes in practices related to ad disclosures on social media.

Disclosures are a crucial legal aspect of influencer marketing. This section presents a brief overview of the general legal framework for advertising in the four countries in our dataset and discusses work related to our research.

Because our objective is a longitudinal study of monetisation practices and disclosures, we collected our dataset using an influencer-focused approach. Publicly available Instagram datasets (such as the one by Kim et al. [22]) were collected for different purposes using a hashtag-focused approach and do not provide adequate temporal data for a longitudinal study; moreover, they are generally not curated for particular countries.

We first selected four countries with different regulatory frameworks to allow for a diverse sample of self-disclosure advertising practices: Brazil (BR), Germany (DE), the Netherlands (NL), and the United States (US). We used Heepsy¹¹1heepsy.com, one of the many platforms that have emerged to support influencer marketing, to select influencers from each of these countries based on the number of followers. We chose Heepsy because it offers a location-based selection filter.

We focused on two groups of influencers: those with more than $500K$ followers, which we call mega-influencers, and those with fewer than $500K$ followers, which we call micro-influencers. The choice of the $500k$ cutoff was inspired by a 2022 national oversight measure implemented by the Dutch Media Authority that focuses on video platform accounts with more than $500k$ followers²²2https://www.cvdm.nl/nieuws/commissariaat-voor-de-media-start-toezicht-op-video-uploaders/. While various other cutoffs exist in the literature, there is no standard definition of what a mega or micro-influencer is. Moreover, follower volume remains a random measurement for influencer studies, as the importance of size varies between platforms and industries. As monetisation models evolve and influencers diversify their strategies, traditional definitions around influencer categories, especially regarding follower count, become increasingly blurred–particularly considering the recent rise of nano and niche influencers [40, 27]. We chose the 500k follower cutoff to align with our study’s objective of understanding how influencers’ visibility impacts their adherence to regulations. Influencers with more than 500k followers are highly visible and likely face more scrutiny, which might drive them to adhere more strictly to advertising regulations. While influencers around the 500k mark are also significantly visible, we hypothesise that those surpassing this threshold experience a qualitative change in their professional practices and engagement with brands, thus attracting more regulatory attention. This hypothesis is supported by the focus of the Dutch Media Authority on accounts above this threshold, suggesting that such influencers are considered impactful enough to warrant specific oversight. Therefore, the 500k cutoff provides a meaningful distinction to study differences in compliance and professionalisation.

To ensure a comprehensive analysis, we initially aimed to select 400 accounts evenly distributed across the four countries and between mega- and micro- categories. Having established our criteria, we used Heepsy to search for accounts that matched our specifications. Heepsy returns, by default, a list of accounts sorted relevance; the details of this metric – how it is defined, calculated and applied in the ranking process – are unclear, as Heepsy offers no transparency regarding its algorithms. Bishop [5] highlights that influencer management tools, such as Heepsy, often incorporate concepts related to brand risk in their metrics, and these concepts could propagate and amplify biases related to sexuality, class, and race. By encoding subjective and potentially discriminatory criteria into their algorithms, platforms like Heepsy may privilege certain types of content over others, leading to a skewed and potentially biased representation of influencers. Therefore, to avoid the potential impact of such biases, we chose not to rely on Heepsy’s proprietary sorting algorithm and instead sorted the accounts in decreasing order of the number of followers. Starting from the largest accounts, we went down the list in descending order of followers. This method caused our data to skew towards micro-influencers approaching the 500k cutoff, as we began selecting from the top of the descending list at 500k.

Heepsy also uses proprietary algorithms to infer the location of influencers. During the data selection process, we noticed that inferred countries were sometimes unreliable. Therefore, we manually curated each account and selected only those with explicit mentions of location and consistency in the language of the posts; this is not a widespread practice given the work-intensive curation and annotation task. Previous studies focus on identifying influencer accounts at scale based on hashtag selection [21, 22, 45]. However, this approach may lead to inaccuracies that are difficult to map computationally. Although our approach may lead to the exclusion of outlier influencers not correctly identified by Heepsy, such a limitation exists across various fixed sampling methodologies. Systematically estimating the coverage of the data returned by Heepsy would require knowledge of every influencer based in a given country, which is not feasible given the current data access restrictions. However, our observations indicated that Heepsy correctly identified at least the most popular influencers from each country.

Given the methodological challenges and the inherent biases of the Heepsy platform, we acknowledge the propensity of our sampling to skew towards micro-influencers nearing our $500k$ cutoff and mega-influencers with substantially larger followings. Obtaining a truly random sample of Instagram influencers is impossible, as we cannot access a complete list of users on the platform. To compile the final list of influencers, we excluded apparent accounts of organisations, fictional characters, pets, and “repost accounts” (profiles focused on compiling content from other users or platforms, e.g., memes).

We collected all posts and their metadata from each curated account using CrowdTangle, along with account-level information such as follower counts. A complete description of the information provided by CrowdTangle is available online³³3help.crowdtangle.com/en/articles/4201940-about-us. We collected a total of $1,006,253$ posts spanning from October 2010 to September 2022. Of these, mega influencers contributed $568,915$ posts (56.5%). Breaking it down by country, Brazilian influencers posted the most with $462,746$ posts (46.0%), followed by the US with $294,599$ posts (29.3%), the Netherlands with $143,978$ posts (14.3%), and Germany with $104,930$ posts (10.4%). Table 1 presents descriptive statistics on the most relevant features of the dataset. Each column represents a country (BR, DE, NL, US) and size (Micro ($\mu$), Mega (M)) combination (we call that a setting); Table 2 presents the statistics combining data from all countries.

There was a time lapse between our initial data curation (when we selected the accounts) and the final data collection stage. This interval caused some fluctuation in the number of followers. Consequently, a few micro-influencers surpassed the 500k followers mark when we compiled the data. However, these fluctuations were generally minor, averaging a 2% deviation, and explain the average number of followers that exceeds the cutoff value in the case of DE and US in Table 1.

The average number of followers for micro-influencers is similar for all countries, with values close to $500k$, which matches the criterion used to select the influencers. The differences for mega-influencers are more significant and, at least partially, correlate to the countries’ population size. Influencers from the US are more likely to have larger international audiences due to language and media influence, resulting in almost four times as many followers as the global average. The Verified row in Table 1 denotes the percentage of accounts that have a verified badge, indicating that Instagram has confirmed the account’s authenticity. Most mega-influencers have verified accounts, which is expected given their high number of followers.

The total number of average posts per account varies greatly across settings and does not follow a clear pattern. Brazilian influencers (especially micro) are outliers, having a much higher post volume than the combined average. Similar differences have been identified on other platforms and are linked to the importance and pervasiveness of social media in Brazil and the Global South in general [35, 36]. The average number of Likes and Comments per post, a measure of user engagement, correlates to the number of followers—a larger audience entails more people seeing and likely engaging with a post. It also correlates negatively to the number of posts, as the low relative engagement for Brazilian micro-influencers indicates. A high post-frequency might discourage followers from engaging because there is less novelty, or followers feel overloaded.

To investigate the role of regulation in influencer marketing, we focus on disclosed posts because they directly impact how influencers frame their sponsored content. To identify self-disclosures, we compiled an initial list of hashtags and keywords frequently used to disclose ads based on previous work  [45, 22]. We expanded the original list to include related terms in other languages by analysing the co-occurrence of hashtags in a data sample. We also include a few non-standard spelling variations (e.g. typos). The final list contains #ad, #advertisement, #advertising, #advertisiment, #advertisment, #anzeige, #publi, #publicidade, #publipost, #werbung, #spons, #samenwerking, and #sponsored and their versions as keywords (without the # symbol). These disclosure terms are typically included in the self-regulatory guidelines discussed in subsection 2.1. Finally, we labelled posts containing any of those terms as sponsored. We also consider posts tagged with the paid partnership label as sponsored. This label is an Instagram feature released in 2017 to encourage disclosing branded content.

Table 3 presents statistics about sponsored posts aggregated by setting; Table 4 presents the statistics combining data from all countries.

We see that the number of sponsored posts varies greatly across countries. Germany comprises approximately 50% of the total, indicating that their strict enforcement of regulation must affect disclosures. Brazilian micro-influencers have a significantly lower amount of disclosed sponsored posts. Combined with the high number of posts, this shows a lower level of professionalisation of monetisation practices; these accounts might not be getting sponsorships despite their frequent activity, or they might not be disclosing them, which connects to the lack of clear regulation and enforcement. The numbers from the Netherlands are also below the global average despite these accounts having a high relative engagement, which is often the measure brands use to find and employ influencers.

Days First Sponsored in Table 3 represents the average number of days between the first post of an account and its first sponsored post. This number shows how long influencers take to monetise their content and thus can serve as a proxy for professionalisation. Finally, we define an account as Sponsored if it has at least one sponsored post, i.e., it shows whether the influencer has ever accepted paid sponsorships. The vast majority of influencers in all settings have sponsored posts. The number is consistent across sizes, indicating that micro-influencers appeal to brands as much as mega-influencers.

In order to investigate how content monetisation evolved on Instagram and to analyse trends over time, we conduct longitudinal analyses of different features of posts in our dataset. Figure 1 presents the number of monthly posts per country throughout the observation period. Instagram was launched at the beginning of October 2010.

The first posts in our dataset were from the US in October 2010, Brazil in December 2010, the Netherlands in February 2011, and Germany in September 2011. By mid-2011, the activity of the 100 accounts from Brazil significantly exceeded that of all other countries, with the US taking the second position. This ordering by the number of posts is maintained throughout the period and is also visible in Table 1, where the average Brazilian account in our dataset makes approximately 4.5K posts, 50% more than the average US account, and three times more than the average Dutch or German influencer account.

With the exception of Germany, all countries saw a peak of activity in late 2015 and a consistent decrease in the monthly number of posts since then. The sharpest decline, which began around mid-2016, coincides with the release of the Stories feature in August 2016. Since our dataset only contains posts, the decrease in overall activity could indicate a shift in posting strategy – from regular (permanent) posts to stories. The Brazilian accounts shrunk their activity fivefold in August 2022 from the peak in March 2015, while the German accounts only less than threefold. This difference highlights that although some activity might have shifted from posts to stories, the decline could at least partially be explained by platform migration – some influencers might have increased their activity on other social media platforms, such as TikTok. In the context of Brazilian influencers, this trend contrasts with recent findings by Martins et al. [28], which indicate a consistent increase in the number of disclosed ads. We acknowledge this discrepancy and consider that factors such as differences in data collection and sampling methodologies may account for these differing observations. Unlike our approach, Martins et al. [28] use a larger set of disclosure hashtags to categorise sponsored posts. Additionally, their criteria for selecting influencers differ from ours; they include accounts with over 1k followers that have posted disclosed content. Given our stricter definition, comparisons should be made with these differences in mind. We verified that the accounts in our dataset were still active at the end of our observation period; however, we could not investigate their activity on other platforms.

This pattern of activity is not noticeable in the time series of the percentage of disclosed sponsored posts shown in Figure 2. As early as October 2012, two years after the launch of Instagram, disclosed sponsored posts were visible in the German influencers’ activity, and they reached 35% of all posts by October 2018. None of the other countries comes close to this level of disclosure.

Although posts by German influencers comprise the majority of disclosures, the country has the least diversity in the number of accounts contributing to the total volume of disclosures. 17% of accounts from Germany contribute with 80% of the disclosures – the lowest among all countries. This proportion shows that a few influencers concentrate most of the sponsored content, or at least are the ones to adhere more strictly to the disclosure requirements. Similarly, the other countries in the dataset, especially Brazil, show a higher diversity in the number of influencers with sponsored posts, which might indicate that the monetisation market is more accessible.

Content creators use different strategies to disclose sponsored posts. Our dataset includes three types of disclosures: hashtags, keywords, and the paid partnership tag (AD tag). Different disclosure types impact users differently: Mathur et al. [29] show that the efficacy of disclosures on YouTube and Pinterest varies according to their type, and social media users particularly fail to understand nonexplanatory disclosures. We study how content creators interpret influencer marketing regulations and express their adherence to them through disclosures. We also investigate how different disclosure types impact user engagement.

To analyse how the use of disclosure types evolved over time, Figure 3 shows the frequency of the three individual disclosure types over the years.

Starting in 2017, the use of the AD tag grew rapidly among BR and NL content creators. Instagram states that content creators must use the paid partnership label to tag the brand with which they are working. However, US and DE posts still mostly use hashtags and keywords to disclose sponsored content. From a user-experience perspective, the paid partnership label is the most visible disclosure type: posts with the tag have a clear label above the photo or video indicating that they are ads. Hashtags are primarily included at the final of a caption, after the entire text. Instagram’s interface does not display long captions completely, so users must click to read the entire text and thus see the disclosure hashtags. Keywords could be placed anywhere within the caption.

Figure 4 presents the average position of the first disclosure in the captions as a proxy for the visibility of the disclosure types. The results are aggregated by country, with vertical lines indicating the standard deviation value.

On average, German content creators place disclosures at the start of the caption, highlighting a difference in positioning between keyword and hashtag disclosures.

Caption length also affects disclosure position, especially for hashtags – longer captions increase the position of hashtags at the end. Figure 5 presents the average caption length per country for sponsored and non-sponsored posts.

The trend for both sponsored and non-disclosed posts shows that the average caption length increased significantly over the years in all countries. Sponsored posts tend to be much longer than non-disclosed ones, with nearly double the average length. This finding is consistent with the literature on marketing strategies, highlighting that advertising posts tend to include more complex narratives to persuade followers to engage with partner brands [26]. Despite the variations, the average disclosure position remained low, indicating that disclosures in posts by German content creators are more effective and consistent.

In general, our analyses show the influence of regulation on how content creators disclose paid sponsorship. With a prevalence of keyword disclosures, sponsored posts by German influencers tend to have more visible disclosures at the start of the captions. Although the paid partnership label has gained traction in most countries in our dataset, it still falls behind other, less transparent, types of disclosure; one possible explanation for its lack of adoption is in its name: the wording implies a partnership involving only monetary transactions, failing to include other business models which are common in the context of influencer marketing, such as barter [17]. Analysing content creators’ use of disclosures helps evaluate regulation effectiveness and suggests ways to enhance disclosure visibility, ultimately improving transparency for content monetisation.

In addition to ad disclosure strategies, we also analyse post-timing patterns to better understand how sponsored content differs from regular posts. Content creators might strategically select the time of day they post to maximise engagement, especially for sponsored content. Figure 6 shows the post-frequency aggregate by hour of the day converted to the local timezone of the corresponding country; for countries with multiple timezones, we select the one that covers the most population. We compare sponsored and non-sponsored posts to verify whether influencers employ different strategies according to the type of content.

The trend for both types of posts is consistent during most of the day; however, there is a significant increase of sponsored posts in the late afternoon (between 5 pm and 9 pm), which coincides with prime time for legacy media advertising hours. This result confirms that there is a clear pattern for when content creators post most sponsored content, indicating that optimising post-timing is, in fact, (still) a marketing strategy. Understanding this strategy could be useful for law enforcement: focusing on analysing activity around this time window could lead to identifying more disclosed and undisclosed sponsored content.

We analyse the brands that influencers tag in their sponsored posts to understand what they advertise and gain more insight into their marketing strategies. We use the paid partnership with label, which requires influencers to include the sponsoring brand’s username, to obtain a reliable list of brands by compiling all these usernames.

In total, 1752 unique brand accounts were tagged in posts in our dataset using the paid partnership label; 1490 (85%) of them are also explicitly mentioned in the caption. This number comprises 33.4% of all users tagged in the disclosed posts. 39.4% of all disclosed sponsored posts tag at least one of these brands, indicating that we compiled a representative sample. The ten most common brands found in our dataset are Embelleze, Eudora, Gymshark, Adidas, Zalando, Natura, PrettyLittleThing, LBA Sunglasses Boutique (@lbdashop), Intimissi, and Huawei. These brands span a variety of sectors, including fashion, beauty, sports, and electronics. To delve deeper into which brands collaborate with micro and mega influencers, Table 5 presents the most frequently tagged brands for each influencer category.

Among the top ten brands, only Embelleze – a Brazilian cosmetics brand – appears in both micro and mega influencer lists. Taking into account the entire dataset, there is an overlap of 197 brands between the two categories, representing 11.24% of the overall distribution. This overlap highlights a substantial difference in most brand partnerships across influencer categories. This variation emphasises the unique brand strategies for each influencer group.

We observe significant differences in the distribution of brands across countries. The posts of Brazilian influencers tag the highest number of brands: 688 (46.6% of the total of our compiled list), while US, DE, and NL, respectively, tag 511 (34.3%), 433 (29.1%), and 207 (13.9%); mega influencers tag more brands than micro in all countries. The higher number of brands tagged by Brazilian content creators indicates more activity and diversity in the context of content monetisation, as more brands opt to partner with influencers for advertising. Table 6 shows the ten most frequently tagged brands by influencers from each country in our dataset.

To better understand which brands (e.g., local or international, type of industry) influencers partner with, we measure the general prevalence of brands across different countries. 241 brands (16.2%) occur in posts by influencers from two different countries; 87 (5.8%), from three; and 21 (1.4%) are tagged in all countries. Therefore, at least 23.4% of brands tagged in our dataset are international, and posts tagging these brands comprise 17.6% of the total volume of disclosed posts. These brands are predominantly from the fashion and beauty industries (e.g. Gucci, Dior, Zara, Lancôme). We analyse the distribution of these posts across mega and micro influencers to understand brand preference when selecting the influencers to collaborate with: 66.1% are by mega influencers and 33.9%, by micro, suggesting that international brands may prefer to partner with influencers with larger audiences. We also analyse the distribution by setting (country + size) and observe that mega influencers have more posts sponsored by global brands in all countries, with the highest prevalence in Germany; however, German micro influencers still have proportionally more posts than all other settings. This result reinforces the evidence for the presence of undisclosed sponsored posts from other countries, as the real distribution of disclosures should be closer to the German one because of its stricter regulations.

To further investigate the potential presence of undisclosed sponsored posts, we analyse the overall prevalence of tagged brands in the entire dataset, including posts without disclosures. In total, 51.6K posts have at least one of the brands tagged; only 13.1K of them (25.4%) are disclosed as sponsored. Therefore, 74.6% of these posts could be undisclosed ads. Moreover, 362 out of the 400 accounts in our dataset (90.5%) have at least one post tagging one of these 1490 brands, which shows that brands engaging in sponsored content have massive reach across all settings. We also analyse the distribution of undisclosed posts tagged with brands considering country and size: the majority (58.5%) are posted by BR influencers (31.4% mega and 24.6% micro), followed by US mega (20.8%), NL mega (10.9%), US micro (4.4%), NL micro (3.7%), DE mega (2.8%), and DE micro (1.4%). This result indicates that the stricter disclosure regulations for German influencers are effective, as most posts tagged with these brands are actually disclosed as ads. On the other hand, Brazilian content creators disclose the least. Although a post could mention a brand without necessarily being sponsored by it, our dataset likely contains a significant volume of undisclosed ads.

Research highlights that influencers often hesitate to disclose sponsored posts adequately, fearing that they will get less engagement, thus decreasing their value [8, 10, 2]. Ershov and Mitchell [15] concluded that after the strengthening of disclosure regulations in Germany engagement with disclosed posts fell substantially. Engagement metrics often determine an influencer’s value for a brand; therefore, reductions in engagement are likely to impact the monetisation potential of content creators.

We define engagement as the number of likes and comments averaged over all posts of the posting influencer. We do not normalise engagement by the number of followers because this value is only available when data are retrieved; thus, we do not have access to the number of followers an account had when each post was published. We conducted experiments to verify whether stricter disclosure requirements reduce engagement and to understand which influencers these reductions affect the most. We focus on two types of sponsored posts: disclosed (subsection 7.1) and undisclosed (subsection 7.2).

This paper presents a longitudinal characterisation of self-disclosed sponsored content posted on Instagram by 400 influencers from four countries from the platform’s release in October 2010 to September 2022. Our study revealed the following observations: First, our longitudinal analysis highlights the impact of adopting new features of platform monetisation in different countries. Specifically, we observe that the introduction of the paid partnership with feature in 2017 led to faster adoption in Brazil and the Netherlands and very low adoption in Germany, perhaps due to German case law and the adoption of pragmatic practices by law firms who advised influencers “to be on the safe side” [23].

We also observe a significant difference in disclosure strategies across the different countries, such as the length of captions for disclosed posts and the position of the disclosing term. German case law recommends that the disclosure terms be clearly visible and separate from a hashtag cloud, which might explain the overwhelming adoption of keywords in sponsored posts by German influencers and their positioning at the beginning of the caption. Therefore, we conclude that strict disclosure requirements lead to more disclosed posts and more visible disclosures.

Second, we show that the posting activity of the observed accounts increased significantly until 2016 and then decreased steadily to the current day. The start of this decline coincides with the release of the Stories feature. We hypothesise that influencers diversified their activity by posting more Stories than permanent posts. However, the proportion of sponsored content remained constant in about 5% of the posts in three of the countries we analysed, except for Germany, which saw a significant decrease in the percentage of disclosed posts after 2018. This result shows that influencers continued to use permanent posts to share sponsored content.

We also observed that Brazilian accounts post significantly more than influencers from other countries, and the share of sponsored content is spread across more accounts, showing that the market is more uniform. However, high posting activity leads to lower engagement on average; we observe that the volume of posts is inversely proportional to engagement.

These findings indicate the need for nuanced regulatory strategies that consider changes in platform characteristics and cultural differences in influencer behaviour. Additionally, these results underscore the methodological challenges in replicating and generalising findings over time, as ongoing changes in platform features and user behaviour may render past observations and trends non-representative.

Third, we identify several monetisation strategies. Regarding the timing of posting sponsored content, we see a significant increase in the late afternoon local time (between 5pm and 9pm), consistent with legacy media advertising hours. Therefore, the timing of posts can be a relevant feature for both machine learning and computational methods aimed at detecting sponsored content and for regulatory agencies monitoring these activities.

We thoroughly investigate the impact of sponsored content and disclosures on engagement. We identified that the engagement trend changed considerably after July 2018. Before this date, disclosed sponsored posts had higher engagement than non-disclosed posts. After July 2018, this trend reverses. We analyse the period in which the disclosed posts had higher engagement and conclude that popular mega influencers drive the average engagement up through a few posts with outlier performance; we identify coordinated marketing campaigns among these posts. This observation could be helpful for the enforcement of disclosure regulations in social media, as it suggests that monitoring spikes in posts related to specific brands or campaigns could lead to undisclosed sponsored content.

To analyse whether undisclosed ads impact the engagement trend with non-disclosed posts, we fine-tune a BERT-based model to detect sponsored content, including undisclosed ads. The model reaches a 0.90 F1 score for disclosed sponsored posts, with a 67.4% accuracy rate when identifying undisclosed ads. We identified 123.6k undisclosed ads, representing 12.3% of all posts. We compare the differences in engagement between disclosed and undisclosed sponsored posts and conclude that, generally, disclosures do not harm engagement.

Our observations highlight the impact of legal uncertainty on advertising disclosure. They also reflect the importance of clear mandatory guidelines for effective disclosure strategies. Without them, even regulations that are supposed to be harmonised at the European level seem to be perceived differently and applied in Member States such as Germany and the Netherlands. Moreover, disclosed content does not show less engagement than undisclosed content in most scenarios – this empirical observation could be used to encourage influencers and brands to follow disclosure regulations.

We acknowledge that the 400 selected influencers may not be a representative sample of the user account population with a very large following base. However, our qualitative data curation relies on clearly defined criteria for finding and selecting influencers, allowing for more systematic and reproducible experiments. Moreover, we highlight that our study, by design, focuses on descriptive and observational analyses contextualised within specific influencer categories and countries. This approach allows us to provide nuanced perspectives on disclosure practices across different social media landscapes. However, the complexity of external factors, such as seasonality and their potential impact on influencer behaviour, presents challenges in measuring statistical significance with robustness. Consequently, while our findings offer valuable contributions to understanding influencer marketing dynamics, they must be interpreted within the framework of these methodological constraints.

All the code relevant to the paper is publically available on GitHub. We provide a script for our experiments’ data collection and processing phases. This script is self-contained; researchers simply need to provide their CrowdTangle API key to initiate the data collection process in line with our methodology. The code necessary to reproduce all of our experiments is also publicly available. All resources can be found on https://github.com/thalesbertaglia/instagram-disclosure-trends.

This research was funded by Studio Europa, Maastricht University, and the ERC Starting Grant HUMANads.

The authors declare that they have no competing interests.

TB, CG, and JS collected and curated the dataset. TB, AI, and JS designed the experiments. TB implemented the experiments. All authors contributed to writing and reviewing the manuscript.