68% of engineers document for Claude, not teammates

According to a survey of 2,400 engineers published on Plover's blog [Plover Blog], 68% of respondents regularly produce API docs, inline comments, or usage guides explicitly to help Claude generate better completions, while only 31% say they do the same for colleagues. The survey asked participants to rate how often they document for three audiences: AI assistants (Claude, Copilot, Gemini), internal peers, and external users. Responses revealed a 2.2-fold higher frequency for AI-focused documentation [Plover Blog].

The same cohort reported a 12% drop in code-review comments over the past year, a trend the authors link to the rise of AI-assisted review tools [Plover Blog]. The Plover post also cites the 2025 Stack Overflow Developer Survey, which recorded a 9% increase in developers who consider “AI-generated suggestions” a primary source of learning new APIs [Stack Overflow Survey]. That survey noted 54% of respondents now rely on AI for onboarding, up from 45% in 2024 [Stack Overflow Survey].

When engineers tailor docs for Claude, they prioritize machine-readable patterns (e.g., exhaustive type annotations) over narrative explanations that help teammates understand design intent [Plover Blog]. Over time, the tacit knowledge that bridges code and business logic erodes, raising maintenance risk. The 12% decline in peer comments suggests that AI tools are supplanting human feedback loops [Plover Blog]. While AI can catch syntactic bugs, it lacks the contextual judgment needed for architectural decisions, potentially leading to hidden technical debt. Companies that reward “AI-ready” documentation may inadvertently discourage collaborative knowledge sharing, as engineers who feel their documentation is valued by AI report 18% lower job satisfaction than those whose docs are praised by peers [Stack Overflow Survey].