Bob Belderbos: Give Your AI Reviewer the Spec, Not Just the Diff

Original Article Summary
A second AI reviewer with no session context only checks the code is sound. Hand it the spec and it checks you built the right thing.
Read full article at Belderbos.dev✨Our Analysis
Bob Belderbos's recommendation to give AI reviewers the full specification instead of only a diff “hand‑off” detail reshapes how automated code checks operate, emphasizing context‑rich analysis over superficial line‑by‑line validation. For website owners, especially those hosting API docs, OpenAPI specs, or downloadable SDKs, this shift means AI bots—like GitHub Copilot’s reviewer or emerging “spec‑aware” linting services—will begin requesting entire spec files (e.g., *.yaml, *.json) rather than just the changed source. Expect a measurable uptick in traffic to endpoints that serve these specifications, and potentially higher request rates from bots that need the full context to evaluate whether a PR meets the intended behavior. Ignoring this pattern can overload servers, skew analytics, and expose proprietary design details to unintended crawlers. **Actionable tips:** 1. Add explicit entries to your llms.txt that allow reputable spec‑aware reviewers (e.g., User‑Agent: Copilot‑Reviewer/1.0) while blocking generic crawlers that request spec files indiscriminately. 2. Implement rate‑limiting on *.yaml and *.json endpoints and use llms.txt to serve a custom 429 response that directs legitimate AI reviewers to a throttled API key. 3. Deploy real‑time bot tracking (via llmscentral.com) to flag sudden spikes in spec file requests, enabling you to adjust llms.txt rules before performance degrades.
Track AI Bots on Your Website
See which AI crawlers like ChatGPT, Claude, and Gemini are visiting your site. Get real-time analytics and actionable insights.
Start Tracking Free →
