Subscribe
Share
Share
Embed
This week asks whether sponsored posts, likes, and popularity scores can be trusted without disclosure, provenance, and transparent audit trails.
From 100 million tweets to Chinese video-platform popularity metrics, this episode follows how measurement breaks when disclosure, data access, and algorithms stay opaque.
Covers 2026-08-26 to 2026-09-02; 5 free papers from 40 selected papers.
This Week in Media Measurement tracks research on how media, platforms, and marketing are measured, from social media and web analytics to campaign evaluation, audience behavior, AI-driven content, and privacy-preserving methods.
Episode covers 2026-08-26 – 2026-09-02.
Themes: social media, digital media, digital marketing, elementary education, consumer behavior, media framing, mental health, political communication
Methods: survey, quantitative, qualitative, cross-sectional, content analysis, quasi-experimental
Premium also covers 10 related news stories, including measured.com — How to Measure TV and CTV Incrementality: 2026 Methods ..., news.broadcastmediaafrica.com — BMA Unveils Survey To Transform Advertising And Audience Measurement Across Africa - Broadcast Media Africa, and digiday.com — ANA continues attempt to standardize RMN measurement.
The premium version of this podcast covers all 40 research articles and 10 news stories selected for the episode. Subscribe to the premium podcast.
Generated by paperboy.fm.
This Week in Media Measurement tracks research on how media, platforms, and marketing are measured, from social media and web analytics to campaign evaluation, audience behavior, AI-driven content, and privacy-preserving methods.
Subscribe for the premium version of this podcast: https://paperboy.fm/podcasts/media-measurement/subscribe
Jenny: When you see someone raving about a product online, how do you know if it is a real recommendation or an ad?
Davis: I want to say you can feel it, but honestly, my feed is useful half the time, and that's what makes the too-perfect recommendation dangerous.
Jenny: Exactly, because if the sponsorship is invisible, then we're not just misreading one post, we're measuring paid persuasion as if it's ordinary enthusiasm.
Davis: So if researchers can scan more than 100 million tweets and find that over 96 percent of sponsored posts weren't disclosed as ads, the trust problem isn't a vibe, it's a measurement problem...welcome to This Week In Media Measurement on paperboy.fm.
Jenny: This week the funnel starts big: 2,811 search hits, 109 qualified papers, 326 unique authors, and 45 countries. So the headline isn't scarcity. It's trust. Which platform signals are sturdy enough to survive the cut?
Davis: And that cut got tighter in a funny way. Qualified papers slipped from 112 to 109, down 2.7 percent, even while the raw search pile got much larger. That tells me more media-measurement work is appearing, but less of it is landing cleanly inside the show’s scope.
Jenny: The search pile jumped from 1,768 to 2,811, up 59 percent, while the semantic shortlist stayed fixed at 200. Plainly, that's the model’s closest-match pile after the broad search. So my question is: did social media flood the results, or did measurement language get used more loosely across venues?
Davis: The theme sweep points that way. Social media leads with 26 papers, then digital media and digital marketing at 7 each. Surveys show up 34 times, quantitative work 28 times, and qualitative work 19 times, which means a lot of this week's evidence is still self-report or coded interpretation, not platform-side audit logs.
Jenny: The geography changed more than the paper count. Countries rose from 26 to 45, up 73.1 percent, with Indonesia at 11 papers, China at 7, and India and Britain at 4 each. That's broader coverage, but it also makes comparability harder unless the same platform metric means the same thing in Jakarta, Beijing, and London.
Davis: The author mix is unusually balanced too: 121 first-time authors, meaning first-ever paper by the metadata, plus 104 emerging authors and 101 experienced authors. That's roughly 37, 32, and 31 percent. Good for fresh measurement problems, but it raises the same audit-trail question: who can reproduce the signal after the post, dashboard, or API changes?
Jenny: Alright, let's get into the papers with a trust shocker. Daniel Ershov, Yanting He, and Stephan Seiler have a twenty-twenty-six paper called Quantifying the Nondisclosure of Influencer Advertising on Social Media, and it's about paid social posts that look ordinary because nobody labels them as ads.
Jenny: The headline is blunt: over ninety-six percent of sponsored posts they identified were undisclosed. They analyzed more than one hundred million tweets, so this isn't a tiny scrape of celebrity posts; it's a large attempt to measure how much paid persuasion is hiding inside normal-looking social chatter.
Davis: If almost all sponsored posts are hidden, what happens to every engagement metric built on top of that content? A like or repost looks like audience enthusiasm, but it may also be reacting to an ad the audience didn't know was an ad.
Jenny: That's exactly the measurement problem. The authors built a text-based method to spot sponsored posts that didn't disclose their sponsored nature, which means they weren't just counting hashtags like ad or sponsored; but the evidence is still centered on tweets, so I'd be careful about treating this as the rate for TikTok, Instagram, or YouTube.
Davis: The practical takeaway is pretty severe: disclosure detection has to come before influencer analytics, not after. This fits the Trust Needs Audit Trails thread, because if the label is missing, the whole chain downstream gets fuzzy: reach, engagement, conversion, and even whether regulators can see the market they're supposed to police.
Davis: That hidden-ad point makes me stare harder at the humble like, because this next paper treats likes as the visible trail campaigns leave behind. Philipp Darius, Wiebke Drews, Andreas Neumeier, and Jasmin Riedl have a twenty twenty-six paper called Radical populist parties receive greater audience support on social media, and it looks at the twenty twenty-four European Parliament election across all twenty-seven EU member states.
Davis: The plain finding is that radical populist parties got more visible audience support online than other parties, especially on TikTok, YouTube, and Facebook. The team tracked four hundred one parties across Facebook, Instagram, TikTok, X, and YouTube, and their main engagement signal was likes on posts, meaning the easiest public tap of approval or attention that platforms expose.
Jenny: But are likes a real measure of political support, or are they just the easiest signal platforms make visible? I can imagine a voter hate-liking, irony-liking, or just boosting a clip because TikTok served it up three times.
Davis: That's the right caution, and the authors don't claim likes equal votes. They link platform trace data with expert surveys, so party traits like Euroscepticism, emotional appeals, and anti-elitist communication get compared with actual post engagement; the strongest pattern is that more Eurosceptic and more emotionally anti-elite parties tended to draw more likes across several platforms, but the study measures post engagement, not vote choice or offline persuasion.
Jenny: So the evidence is strong for the platform story, because four hundred one parties and five platforms is a big monitoring design, but it's not proof that someone changed their ballot. This is exactly the Engagement Is Not Impact thread: don't mash TikTok, YouTube, Facebook, Instagram, and X into one blended social score when the political dynamics are this uneven.
Jenny: That likes-versus-votes caution is exactly where Fabiano Couto Corrêa da Silva's Datificación social y auditabilidad informacional lands, but from the other direction. Instead of asking whether platform signals mean support, he asks whether Brazil has the public audit trail needed to treat platform-produced data as evidence at all.
Jenny: The plain finding is pretty blunt: Brazil looks stronger on formal transparency rules than on the machinery that lets people verify what happened. The study says normative transparency is relatively solid, meaning laws and stated obligations exist, but traceability, preservation, scientific access, and cognitive justice are fragile; cognitive justice just means different communities can help define what counts as knowledge, not only platforms or state agencies.
Davis: So what would have to be publicly checkable before platform data could count as evidence, not just a dashboard someone asks us to trust?
Jenny: He builds an auditability matrix with eight dimensions and thirty-two documentary indicators, then applies it to forty sources about Brazil, including regulations, international instruments, Meta, TikTok, and X policies and reports, foundation-model documentation, and scholarly work from two thousand to twenty twenty-six. Each indicator gets coded as present, partial, or not publicly verifiable, but the big limitation is that this is exploratory and descriptive, not a validated measurement scale.
Davis: That makes it useful as a checklist, not a final grade for Brazil, and the checklist is the real contribution. In our Trust Needs Audit Trails thread, this is the most literal version so far: if platforms make the evidence, public systems need documentation, traceability, access, and preservation before researchers, regulators, or journalists can inspect the claim.
Paperboy.fm: This is the free version of the podcast. Subscribe at paperboy.fm to access a dozen different paper review podcasts for five dollars a month.