Latest / Influencer Marketing with Fexingo: Creators, Sponsorships, and Modern Endorsement Deals / How Brand Safety Algorithms Are Killing Creator Revenue
Transcript
- Lucas: So there's this food creator I've been following — she makes sourdough tutorials, very chill, very baking aesthetic. Back in March, a sponsored video of hers making a basic boule got flagged by the brand's safety algorithm as 'inflammatory content.' Luna: Inflammatory? A sourdough boule? Lucas: Yeah. The algorithm apparently caught the word 'knead' and associated it with — I don't know, labor disputes? Or maybe the image of her punching down the dough triggered a violence classifier. The point is, the brand paused payment, the video was demonetized for two weeks, and she had to file an appeal with the platform and also with the brand's verification partner. Luna: That's not an isolated story. I've heard similar from creators in finance, health, even gardening. The algorithms are trained on broad semantic categories, and they're making high-stakes mistakes. Lucas: Exactly. And the scale of it is huge. There's a 2025 Creator Economy Benchmark report — not the one from SignalFire, a different one from a firm called CreatorIQ — that found that 42 percent of sponsored content now requires some form of manual whitelisting before it can run. That's nearly half. And whitelisting isn't automatic — it's a negotiation between the creator and the brand's compliance team. Luna: Right, so you're not just making the video, you're also getting pre-cleared by a human who may not even understand the content. That's a huge time sink. Lucas: And a revenue sink. Every day the video sits in limbo, the creator isn't earning. The brand might also be losing the window of relevance — especially if it's a topical campaign. Luna: So who are the main players here? The brand safety vendors? Lucas: Two big ones are Zefr and IAS — Integral Ad Science. They're the ones that scrape video metadata, audio transcripts, and on-screen text to score content on a 'brand safety' scale. If a video scores below a threshold, it gets flagged. The problem is their models are still pretty crude. They rely on keyword blacklists that haven't been updated for context. Luna: So 'knead' triggers a false positive. What about more obvious categories? Like, a health creator talking about weight loss — that's a minefield. Lucas: Yes. Many brands automatically block any content related to weight, diet, surgery, mental health — even if the creator is a licensed therapist. In the health vertical, our benchmark data showed that nearly 60 percent of sponsored posts from qualified practitioners were initially flagged or suppressed. Luna: That's a structural issue. It's not just a glitch — the system is designed to avoid risk at all costs, and creators are the ones paying for it. Lucas: And the costs are real. Let's talk about the financial impact. If a creator has a cost per thousand impressions of, say, $20, and their video gets demonetized for a week during a campaign, they could lose thousands of dollars in exposure value. Plus the brand might reduce the payment if the view count drops. Luna: Plus the creator's own analytics take a hit. Lower engagement on that video can suppress future algorithmic recommendations. It's a double penalty. Lucas: Which is why we're seeing a rise in what some call 'brand safe creator networks' — agencies that pre-vet creators and their content categories so that brands can buy media without running every video through a safety filter. The creator gives up a little control, but they get faster payment and higher fill rates. Luna: It's an interesting trade-off. But it also means the platform is ceding editorial judgment to a third-party vendor. Are there any examples of creators pushing back successfully? Lucas: A few. Some large creators have negotiated whitelisting contracts that guarantee a baseline payment even if a video is flagged. And platforms themselves are starting to build appeals processes that are faster — YouTube, for example, now has a 'creator safety review' team for sponsored content, but it's only available to partners with over a million subscribers. Luna: So if you're a mid-tier creator, you're still stuck. Lucas: Pretty much. And that's the core tension. Brand safety was supposed to protect advertisers from controversy. But it's evolved into a system that penalizes the very creators brands want to work with — especially the ones who cover nuanced, real-world topics. Luna: I think we're reaching a point where the industry needs to redefine 'brand safety.' Not as absence of risk, but as alignment of values. If a brand stands for something, they should be willing to have their content appear next to real conversations. Lucas: I agree. And some brands are starting to move that way — they're creating bespoke whitelists that are more context-aware. But it's slow, and it's not scaling fast enough for the millions of creators who need to make a living. Luna: Let's take a quick beat here. You know, conversations like this — where we dig into the actual mechanics of how money flows in the creator economy — are exactly why we keep this podcast ad-free. We don't have a brand safety algorithm telling us what we can talk about. If you find that valuable, and you want to help us keep it that way, we have a link: buy me a coffee dot com slash fexingo. No pressure, just a way to support the work. Lucas: Totally. It's listener-supported, and that keeps us independent. And we really appreciate anyone who chips in. Luna: So back to the algorithmic problem — what's the next frontier? Are we going to see ai powered moderation that actually understands context? Lucas: There are startups working on it — companies like Clarifai and Hive — that use large language models to analyze entire video transcripts in context. But they're not widely adopted yet. Most brands still rely on the keyword-matching systems because they're cheap and auditable. Luna: So the short-term fix is for creators to get really good at reading the blacklists and adjusting their language accordingly. Lucas: Which is exactly what many are doing. They're replacing words like 'diet' with 'lifestyle change,' or 'investment' with 'money moves.' It's a kind of algorithmic doublespeak. But that's a band-aid. The real solution is industry-wide standards for content classification that creators have a say in. Luna: And maybe a lawsuit or two to push the platforms into action. Lucas: Wouldn't be surprised. The legal theory is that demonetization without a clear, transparent appeals process is a form of unfair business practice. But that's a whole other episode. Luna: True. For now, I think the takeaway for creators is: know your brand's safety vendor, ask for their keyword block list upfront, and negotiate a whitelisting clause in your contract. Lucas: And for brands — maybe think about whether your algorithm is costing you the authentic voice you hired the creator for in the first place.