AI Slop Statistics 2026: The Numbers, Sourced
A reference roundup of the most-cited statistics on AI-generated "slop" — its prevalence in feeds, how people feel about it, how platforms are responding, and the energy footprint behind it. Every figure below links to its original published source. This page is a synthesis of work done by others (eMarketer, Gartner, Sprout Social, the IEA, journalists, and independent analysts), not original research by us. See the methodology & honesty note before you cite anything.
If you want context on what the term even means, see What is AI slop? and our narrative companion, AI slop by the numbers.
Prevalence: how much slop is actually out there
How common is AI-generated content in the feeds people use every day? The clearest figures available focus on YouTube recommendations.
| Statistic | Figure | Source |
|---|---|---|
| YouTube recommended videos that are low-quality AI slop | More than 1 in 5 (>20%) (Kapwing analysis of Social Blade data) | emarketer.com |
| YouTube recommendations to new users that are AI slop | ~21% | financialcontent.com |
| Additional share of "brainrot" recommendations to new users (late-2025 study) | ~33% more | financialcontent.com |
| Facebook's AI-generated spam problem | "Worse than realized" (reporting, not a single metric) | rollingstone.com |
Two independent lines of analysis landing near the same ~20–21% mark for YouTube recommendations is the closest thing to a consensus figure in this space — though both are estimates of a moving target, not a census.
Sentiment & trust: how people feel about it
This is where the data is richest. The throughline: awareness is high, enthusiasm is falling, and disclosed AI content carries a trust penalty.
| Statistic | Figure | Source |
|---|---|---|
| US adults who would use social platforms less or quit entirely if AI content increased in feeds | ~49% | emarketer.com |
| Consumer enthusiasm for AI-generated creator content | Fell from 60% (2023) to 26% (2025) | emarketer.com |
| People who see AI slop on social often / very often | 56% | emarketer.com |
| People who see AI slop on social at least sometimes | 83% | emarketer.com |
| People less likely to engage with / trust content known to be AI-generated | ~62% | sproutsocial.com |
| Consumers who distrust AI-powered search results (Gartner) | 53% | gartner.com |
The 60%→26% enthusiasm drop over two years is the single most striking trend here. Pair it with the ~62% engagement/trust penalty for known AI content, and the implication is plain: disclosure matters, and audiences punish what they can identify. For the digital-wellness angle on this, see Reclaim your feed.
Platform response: what the platforms are doing
| Statistic / development | Detail | Source |
|---|---|---|
| YouTube CEO Neal Mohan named managing AI slop a 2026 priority | Public statement | cnbc.com |
| YouTube juggling its own GenAI tools against an AI-slop crackdown | Reporting on the platform's dual position | emarketer.com |
| Facebook's AI-generated spam | Reported as worse than previously understood | rollingstone.com |
Worth noting the tension in the YouTube reporting: the same platform shipping generative-AI creation tools is also the one naming slop a priority to manage. Platform-level controls and disclosure labels are improving, but they're partial — which is why some people layer on their own filtering. Background on doing that yourself: How to block AI videos on YouTube and Hide AI posts in your Facebook feed.
Energy & infrastructure: the footprint behind the output
| Statistic | Figure | Source |
|---|---|---|
| Global data-centre electricity use | ~485 TWh (2025) → ~950 TWh (2030), with AI the biggest driver | iea.org |
No-greenwashing caveat. This figure describes the energy cost of producing and serving AI content at the infrastructure level. It is not a claim that filtering changes it. Hiding AI content in your own feed does not reduce data-centre electricity use — the compute was already spent when the content was generated and indexed. Any tool (ours included) that suggests otherwise is overstating its impact. Filtering is about your attention and feed quality, not energy savings.
Methodology & honesty note
- This is a synthesis, not a study. Every statistic on this page was published by an external organization or journalist. We did not run a survey, analyze a dataset, or commission research. There is no "we analyzed N videos" claim here because we didn't.
- Figures reflect their original sources. Definitions of "AI slop," sample sizes, dates, methodologies, and margins of error all belong to the original publications. Click through before citing — the source's framing is the authoritative one, not ours.
- Estimates are estimates. The prevalence numbers in particular (the ~20–21% YouTube figures) are analyst estimates of a fast-moving target, derived from third-party data such as Social Blade. Treat them as directional.
- Some entries are reporting, not metrics. The Rolling Stone and "2026 priority" items are journalistic findings and public statements, not survey statistics. They're included for context and labeled as such.
- We'll correct errors. If a source updates a figure or we've mischaracterized one, the fix is to match the source.
If you cite this page, citing the original source directly is better — this roundup just gathers them in one place.
A quick note on tools
If, after reading the trust numbers, you want fewer AI-generated videos and posts in your own feed, Unslop is a local, no-account browser extension that filters declared/disclosed AI content on YouTube and Facebook. It's not the first to market (see how it compares), it reads declared signals rather than analyzing pixels or audio (so undisclosed AI can slip through), and it's Chromium-only — but the core is free. That's the whole pitch; the statistics above stand on their own regardless of what you use.
Last reviewed: 2026-06-14. Sources are linked inline; figures belong to their publishers.
Want a private, local filter for YouTube + Facebook?
Try Unslop free