How agent-ready is Google Analytics?

Google Analytics achieved an agentability score of 41/100, indicating moderate difficulty for autonomous AI agents attempting to navigate and operate its public website. The product performs well on machine readability (91/100), meaning its markup is clean and parseable, but struggles significantly with transparency (0/100), control (19/100), and shadow UI avoidance (15/100).

The low transparency score reflects a complete absence of mechanisms that help agents understand system confidence, trace outputs to sources, or access audit logs. The control score of 19 indicates that agents face challenges managing destructive actions, canceling long-running operations, and reversing completed changes. These gaps make Google Analytics difficult for autonomous agents to operate safely and predictably, even though the underlying page structure is technically sound.

To improve agentability, Google Analytics should focus on the lowest-scoring principles—transparency and control—by implementing the following changes:

• Add aria-live regions or role=status attributes around status updates so agents can monitor progress without polling the DOM.
• Gate destructive actions behind an explicit, distinctly-labelled confirmation step rather than using same-styled buttons that agents may activate unintentionally.
• Provide labelled pause, cancel, or stop controls for long-running actions with accessible names like 'Cancel import' instead of relying on unlabelled icons.
• Expose an activity or audit log that records system actions in machine-readable form, giving agents a visible activity feed and a structured JSON event log.
• Attach source references or input citations to generated outputs so agents can trace results back to their underlying data.
• Support undo or override capabilities on completed actions, allowing agents to reverse destructive changes or edit results when necessary.