How AI coding agents access your docs (and break your analytics)
If you run a developer portal or publish technical documentation, a meaningful chunk of your traffic is now AI agents, not humans. A new paper published on arXiv studied exactly what that looks like at the HTTP layer, and the implications for your analytics are significant.
What the Study Did
Researcher Oleksii Borysenko fingerprinted HTTP requests from nine AI coding agents and six AI assistant services hitting a live documentation endpoint. The coding agents studied were Aider, Antigravity, Claude Code, Cline, Cursor, Junie, OpenCode, GitHub Copilot in VS Code agent mode, and Windsurf. The assistant services were ChatGPT, Claude, Google Gemini, Google NotebookLM, MistralAI, and Perplexity.
Each tool left identifiable behavioral signatures across HTTP runtime environments, pre-fetch strategies, User-Agent strings, and header patterns. In other words, you can tell them apart if you know what to look for.
The Analytics Problem
Here is the part that should make any developer portal team pay attention. AI agents compress multi-page documentation navigation into a single request or two requests. A human working through your docs might visit eight pages, spend twelve minutes on site, and follow a recognizable click path. An AI agent consumes the same information in one HTTP call.
That makes your standard engagement metrics unreliable as indicators of actual documentation consumption. Session depth, time-on-page, click path, and bounce rate all read as disengagement when the agent is actually doing exactly what you want: ingesting your docs and using them.
What the Paper Recommends
The study points to several practical adaptations for teams maintaining developer documentation:
- Tokenomics-aware documentation design: Structure docs with AI consumption in mind, not just human navigation.
- Machine-readable standards: Adopt emerging formats including
AGENTS.md,llms.txt,skill.md, andagent-permissions.json. - MCP server-based feedback channels: Use Model Context Protocol servers to create structured feedback loops between agents and documentation systems.
- AI referral traffic instrumentation: Build analytics specifically designed to track AI agent access rather than relying on session-based metrics built for humans.
The full paper is available at arXiv under identifier 2604.02544. If your team maintains any public technical documentation and you care about whether AI agents can find and use it, this is worth reading in full.
I’m founder of BizStack at Cagri Sarigoz LLC and a passionate advocate for entrepreneurs.
With over 14 years in tech, marketing, and AI, including my role as Head of SEO at CitizenShipper and co-founder of TaleBot at Intale AI, I’m dedicated to sharing genuine, useful product insights and tips.
At BizStack, I aim to cut through the digital noise to provide clear, actionable advice.
And more than all else, I’m a father to a (always) little girl and a husband.
Contact me at [email protected] for assistance.
