Study scrutinizes hidden marketing relationships on social media

Federal regulators require social media personalities to alert their viewers to promotional payments for products and gadgets shown on their channels, but an analysis by Princeton University researchers shows that such disclosures are rare.

The study focused on affiliate marketing, in which companies pay a commission to social media figures for driving sales. Content creators who produce videos, photos and commentary are rewarded when their followers purchase products after clicking on affiliate marketing links included in their social media posts.

Researchers in Princeton’s Department of Computer Science extracted affiliate marketing links from randomly drawn samples of about 500,000 YouTube videos and 2.1 million Pinterest pins. They found 3,472 YouTube videos and 18,237 Pinterest pins with affiliate links from 33 marketing companies — the first publicly available list of this size. The researchers found the links by identifying characteristic patterns in the URLs that marketers use to track readers’ clicks.

The researchers then used natural language processing techniques to search for disclosures of affiliate marketing relationships within the videos’ and pins’ descriptions. Disclosures were present in around 10 percent and 7 percent of affiliate marketing content on YouTube and Pinterest, respectively. These disclosures fell into one of three categories:

•  “affiliate link” disclosures, which use wording such as “Disclosure: These are affiliate links”;
•  “explanation” disclosures, e.g., “I am an affiliate with Amazon, which means I get a small commission when you buy through my links”; and
•  “channel support” disclosures, such as “Shop using these links to support the channel.”

The first type — “affiliate link” disclosures — were the most common, “and these are exactly the kinds of disclosures the Federal Trade Commission says people shouldn’t be using” because their meaning is not always clear to users, explained computer science graduate student Arunesh Mathur, the study’s lead author. “That was a very surprising finding.”

The researchers also conducted a user study of nearly 1,800 participants that revealed the relative effectiveness of different types of disclosures. Only half of the participants correctly interpreted the meaning of “affiliate link” disclosures on Pinterest, while 65 percent understood these disclosures when paired with YouTube videos. When presented with “explanation” disclosures, nearly 95 percent of users on both platforms were able to explain that the content creator would be paid when a product was purchased through an affiliate link. “Channel support” disclosures, which only appeared on YouTube, were correctly interpreted by 85 percent of participants.

While the FTC has issued warnings about marketing disclosures to some top social media influencers, the researchers said it has not fully exercised its authority to prosecute violations. The researchers propose that regulators take broader legal action against affiliate marketing companies for failures to disclose, and recommend that social media platforms make it easier for content creators to disclose marketing relationships in a standardized way.

Mathur and his colleagues are also developing a web browser extension that would automatically flag some types of paid content. In addition, they are working on computational methods to detect other types of hidden advertising on social media, including sponsored content and product giveaways, which are less straightforward to identify than affiliate marketing.

Other coauthors from Princeton’s Department of Computer Science include Associate Professor Arvind Narayanan and Research Scholar Marshini Chetty. The team received a best paper award from the Association for Computing Machinery’s Conference on Computer-Supported Cooperative Work and Social Computing, where they presented the work on November 7 in Jersey City, New Jersey. The research was supported by the National Science Foundation.


Substack subscription form sign up