Digital pollution refers to the degradation of the digital information environment through the mass production and proliferation of AI-generated content, combined with the substantial environmental costs of the computing infrastructure required to generate it. The concept encompasses both the informational dimension — low-quality synthetic content overwhelming authentic human-created material — and the environmental dimension — the energy, water, and carbon costs of AI generation at scale. 1)
By 2025, AI-generated content had become pervasive across the internet, with the volume of synthetic text, images, and video growing exponentially as generative AI tools became widely accessible. 2) The term “AI slop” entered mainstream vocabulary to describe this flood, earning recognition as Merriam-Webster's 2025 Word of the Year.
The digital commons — shared online information ecosystems including search results, social media, Wikipedia, forums, and websites — faces contamination as synthetic material competes with and displaces authentic content. This creates a degraded information environment where distinguishing reliable from unreliable sources becomes increasingly difficult.
A single ChatGPT query uses approximately ten times the computing power of a standard Google search, and the data centers powering generative AI are consuming electricity at an accelerating rate. 3) Projections indicate that global data center energy consumption could quadruple by 2030, potentially matching the total energy consumption of Japan. In the UK alone, the average data center powers the equivalent of 5,000 homes, with 480 existing centers and over 100 additional facilities planned. 4)
AI data centers consume between 11 and 19 million liters of water daily for cooling, equivalent to the water needs of towns with 30,000 to 50,000 residents. 5)
When powered by fossil fuel grids, AI data centers generate significant greenhouse gas emissions. A report by Beyond Fossil Fuels found that only 26% of Big Tech's claims about using AI to benefit climate were backed by academic evidence, suggesting widespread greenwashing in the industry's environmental messaging. 6) Lifecycle emissions from data centers — including manufacturing and maintaining servers that may go unused — further amplify the environmental impact. 7)
AI-generated content pollutes information ecosystems by producing convincing but potentially inaccurate text, images, and video at scale. This material buries factual content under layers of synthetic noise, eroding public trust in online sources. 8)
Digital pollution creates a secondary effect: as AI-generated content contaminates the web, new AI models trained on web-scraped data inadvertently learn from synthetic material rather than human-generated sources. This contributes to model collapse, where successive generations of AI produce increasingly degraded and homogenized outputs. 9)
Bad actors leverage generative AI to produce disinformation at industrial scale, including climate denial campaigns and politically motivated propaganda. The low cost and high volume of AI generation makes such campaigns far more accessible than traditional disinformation operations. 10)
Researchers and policymakers have proposed several approaches to addressing digital pollution:
The ACM has published research on proactive mitigation of digital pollution's effects, emphasizing the need for systematic approaches rather than reactive measures. 11)