AI-Generated Algorithmic Virality

2025-08-01

AI Summary

This study investigates the viral dissemination of AI-generated content (AIGC) on social media and its threat to the information ecosystem. Drawing on 13 political and public-issue hashtags across TikTok and Instagram in Spain, Germany, and Poland, we develop a classification framework for “proxy AI accounts” and conduct an empirical analysis of 30 top-search results and 153 AI-driven accounts, employing cross-lingual human annotation and manual content analysis. We find AIGC dominates multiple topical domains, while platforms largely lack standardized AI disclosure mechanisms—enabling widespread proliferation of low-quality, deceptive “AI slop.” Key contributions include: (1) the first systematic identification and definition of proxy AI account behavioral patterns; (2) empirical evidence that algorithmic recommendation systems amplify low-fidelity and misleading AIGC; and (3) a novel cross-platform, cross-lingual analytical paradigm for AIGC diffusion, providing actionable empirical foundations for regulatory policy and platform governance.

Technology Category

• Application Domains: Misinformation & Fake News
• Humans and AI: Game Design — Procedural Content Generation & Storytelling
• Philosophy and Ethics of AI: Accountability, Interpretability & Explainability

Application Category

• Social Networks and Social Media: Generative AI / large language models and their impact on social systems
• Web Mining and Content Analysis: Web data quality in the era of algorithmically-generated content
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web

Abstract

There is a growing discussion about social media feeds being increasingly filled with AI-generated content. Due to its visual plausibility, low cost, and fast production speed, AI-generated content is said to be highly effective in "gaming the algorithm" and going viral. Popularly referred to as "AI slop," this phenomenon arguably leads to the presence of sloppy and potentially deceptive content at a scale unseen before. This investigation offers a systematic analysis of AI-generated content and its labelling in TikTok's and Instagram's search results across 13 hashtags (see Appendix) in three European countries (Spain, Germany, and Poland) over the course of June 2025. We manually annotated and analyzed the 30 top search results on political (#trump, #zelensky, #pope) and broader topics (e.g.,#health, #history) to understand the relation between synthetic (content that is partially or entirely made using generative AI) and non-synthetic content across languages and countries. We then explored the emerging phenomenon of accounts producing generative AI content at scale by analyzing 153 accounts and proposing a new categorization schema of what we termed Agentic AI Accounts. Our main findings are:

Problem

• Analyzing AI-generated content prevalence in social media feeds
• Investigating algorithmic virality of synthetic vs non-synthetic content
• Categorizing Agentic AI Accounts mass-producing deceptive content

Innovation

• Manual annotation of top search results
• Analysis of synthetic vs non-synthetic content
• Categorization of Agentic AI Accounts