The Multiplication Factor
HUMAN Security's 2026 report reveals a number that changes the infrastructure math entirely: AI agents visit approximately 5,000 pages per task, compared to 4-5 for a human performing the equivalent search. That's a 1,000x multiplication factor on server requests.
When an AI research agent answers a question about enterprise software options, it doesn't visit 5 vendor pages like you would. It crawls 5,000. When an AI shopping agent compares prices, it hits every retailer with a product listing. When an AI content agent summarizes industry news, it reads every publication in the vertical.
The Framework Math at 1,000x
A WordPress page: 2-4MB average weight (HTTP Archive). Serve 5,000 of those to an AI agent: 10-20GB of data transferred for one task. Most of that data is plugin-injected JavaScript, theme overhead, and framework noise the AI agent discards. You're paying to transmit waste.
An Astro page: 30-100KB average weight. Serve 5,000 of those: 150-500MB. Same content, 40x less data. The AI agent gets the same information — faster, cleaner, cheaper. At machine-scale traffic volumes, the framework's output weight is the dominant factor in your infrastructure cost.
The Hidden Cost Reframe
WebPulse's True Cost of WordPress analysis calculated $4,200 to $38,000 per year per site in maintenance costs. That analysis assumed human-scale traffic patterns. At 57.4% bot traffic with a 1,000x page multiplication factor, the bandwidth and compute costs of serving WordPress pages to AI agents could exceed the maintenance costs.
This isn't speculative. Hosting providers are already reporting increased bandwidth consumption from AI crawlers. Sites that allow AI crawling — as WebPulse recommends for discoverability — absorb the machine-scale traffic their framework generates. A lean framework generates lean traffic. A bloated framework generates bloated traffic. At 1,000x scale, lean wins.
