A major study released in late April 2026 has confirmed that the “Dead Internet Theory” is moving closer to reality. Researchers from Stanford University, Imperial College London, and the Internet Archive found that 35% of all new websites created since 2022 are now AI-generated or AI-assisted.
This transformation marks the fastest shift in the history of the internet, with the composition of the web changing more in the three years since the launch of ChatGPT than in the three decades prior.
1. The Composition of the “AI Web”
The study utilized the Internet Archive’s Wayback Machine and advanced detection software (Pangram v3) to analyze the trajectory of new site launches from mid-2022 to mid-2025.
- Fully AI-Generated: Roughly 20% of new sites are identified as “fully autonomous,” meaning the site structure, code, and content were produced entirely by LLMs with zero human editing.
- AI-Assisted: The remaining 15% utilize AI for high-volume content generation (blogs, product descriptions) while maintaining human-designed layouts.
- Zero to Hero: Before November 2022, AI content on the web was statistically negligible (near 0%). Crossing the 35% threshold in under four years is being called the “Generative Era (GE)” shift.
2. Key Findings: The Quality Paradox
The research tested several hypotheses about how this “flood” of machine-written content is affecting the user experience. Surprisingly, the results were mixed.
| Theory | Finding | Impact |
| Factuality Decay | False | AI sites were not found to be significantly less accurate than human ones; many use real-time retrieval to cite facts. |
| Artificial Positivity | True | The web is becoming “sanitized.” AI content tends to be overwhelmingly upbeat and avoids controversial or unique “human” viewpoints. |
| Semantic Density | True | Content is becoming “fluffier.” AI sites are often less verbose but also less diverse in vocabulary, leading to a “stylistic monoculture.” |
| Citations & Links | True | “Epistemic islands” are forming. AI sites provide answers but include fewer outbound links, reducing the interconnected nature of the web. |
3. The Economic Driver: 77% Volume Increase
For businesses, the move toward AI generation is purely a matter of efficiency.
- Content Velocity: Implementing an AI content engine allows companies to increase their monthly publishing volume by 77% within six months.
- Cost Reduction: The cost to produce an SEO-optimized article has dropped by an average of 42% since 2024.
- Performance: Data from Semrush in early 2026 shows that AI content performs almost identically to human content in search rankings, with 57% of AI text reaching the top 10 results.
4. How to Spot “Ghost Sites” in 2026
As detection becomes harder, experts suggest looking for these four markers of an AI-generated site:
- High-Frequency Posting: A site with a small team (or no listed staff) posting 20+ long-form articles a day.
- The “Vibe” of Neutrality: Writing that feels excessively balanced, artificially cheerful, and lacks personal anecdotes or “I” statements.
- Low Link Density: A lack of contextual links to other reputable websites or sources within the body text.
- Generic Bylines: Authors with AI-generated headshots (often perfectly symmetrical) and generic names that don’t have a presence on LinkedIn or other social platforms.