Turning Trust to Transactions: Tracking Affiliate Marketing and FTC Compliance in YouTube's Influencer Economy
| Authors | Chen Sun et al. |
| Year | 2026 |
| Field | AI / ML |
| arXiv | 2603.04383 |
| Download | |
| Categories | cs.CY, cs.CR, cs.IR, cs.LG |
Abstract
YouTube has evolved into a powerful platform that where creators monetize their influence through affiliate marketing, raising concerns about transparency and ethics, especially when creators fail to disclose their affiliate relationships. Although regulatory agencies like the US Federal Trade Commission (FTC) have issued guidelines to address these issues, non-compliance and consumer harm persist, and the extent of these problems remains unclear. In this paper, we introduce tools, developed with insights from recent advances in Web measurement and NLP research, to examine the state of the affiliate marketing ecosystem on YouTube. We apply these tools to a 10-year dataset of 2 million videos from nearly 540,000 creators, analyzing the prevalence of affiliate marketing on YouTube and the rates of non-compliant behavior. Our findings reveal that affiliate links are widespread, yet dis- closure compliance remains low, with most videos failing to meet FTC standards. Furthermore, we analyze the effects of different stakeholders in improving disclosure behavior. Our study suggests that the platform is highly associated with improved compliance through standardized disclosure features. We recommend that regulators and affiliate partners collaborate with platforms to enhance transparency, accountability, and trust in the influencer economy.
Engineering Breakdown
Plain English
This paper builds measurement tools to analyze affiliate marketing disclosure practices on YouTube at scale, examining 2 million videos from 540,000 creators over a 10-year period. The authors use NLP and web measurement techniques to detect affiliate links and assess compliance with FTC guidelines requiring disclosure of affiliate relationships. The key finding is that non-compliance and lack of transparency in affiliate marketing remains widespread on the platform, despite regulatory guidelines. By quantifying the prevalence and patterns of undisclosed affiliate relationships, the paper reveals the actual state of the problem that regulatory agencies face.
Core Technical Contribution
The core technical contribution is a scalable NLP-based detection system for identifying affiliate marketing relationships and disclosure statements in YouTube video metadata, descriptions, and transcripts. Rather than manual auditing, the authors developed automated tools leveraging recent advances in NLP to classify videos as containing affiliate marketing and assess whether proper disclosures were made. This enables analysis at population scale (2M videos) rather than sampling, providing the first comprehensive measurement of affiliate marketing non-compliance on YouTube. The novelty lies in operationalizing regulatory compliance detection as a machine learning problem, moving from anecdotal evidence to systematic measurement.
How It Works
The system processes YouTube videos through a multi-stage pipeline: first, it extracts structured data (titles, descriptions, metadata) and unstructured content (transcripts if available) from the video. Natural language processing models then identify affiliate marketing signals—patterns like affiliate links, product recommendations with monetization intent, and common affiliate disclosure phrases. In parallel, the system classifies whether explicit FTC-compliant disclosures are present (e.g., 'sponsored,' 'affiliate link,' 'this video contains affiliate links'). The final output is a classification for each video indicating: (1) presence/absence of affiliate marketing, (2) presence/absence of disclosure, and (3) compliance status. This enables aggregate statistics across creators, time periods, and content categories to quantify the ecosystem state.
Production Impact
For platform engineers building content moderation or compliance systems, this work provides a blueprint for automating regulatory compliance detection at scale. Instead of manual review teams auditing creators (infeasible at 540K creators), you could deploy a lightweight NLP classifier to flag videos for review, reducing auditor workload and improving detection speed. The approach integrates into existing video processing pipelines—scanning metadata and transcripts during upload or indexing. Trade-offs include: NLP models have false positive/negative rates (requiring human review of flagged content), transcripts aren't always available (limiting coverage), and affiliate marketing tactics evolve, requiring model updates. The system would run on existing infrastructure (CPUs for text processing, GPUs only if using transformer-based models) with moderate latency impact during indexing.
Limitations and When Not to Use This
The paper is limited by reliance on NLP models that may not catch sophisticated or coded affiliate language (e.g., 'check out this product I love' without explicit disclosure). Transcripts and captions aren't available for all videos, creating measurement gaps and potentially biasing results toward creators who upload transcripts. The approach assumes FTC disclosure guidelines are the ground truth, but may not capture emerging disclosure practices or regional variations in regulations. The dataset covers only a 10-year window and YouTube specifically—findings may not generalize to other platforms (TikTok, Instagram) with different creator ecosystems and disclosure norms. Follow-up work should explore multi-modal detection (analyzing video thumbnails, product placement visuals) and longitudinal modeling of disclosure behavior change.
Research Context
This paper bridges platform measurement research (building on YouTube-scale analysis work from prior web measurement studies) and NLP-based content classification (applying recent advances in text understanding to a compliance problem). It connects to broader research on platform transparency, creator accountability, and regulatory measurement in digital ecosystems. The work opens a research direction in automating regulatory compliance detection—applying ML to enforce FTC, GDPR, and other guidelines at scale rather than through manual audits. It also contributes to the growing field of computational social science, using automated tools to measure societal problems (undisclosed advertising, consumer harm) on digital platforms.
:::tip Subscribe Get weekly breakdowns of papers like this in AI Letters - the newsletter for engineers building production AI systems. :::