How agent-ready is Confluence?

Confluence's public website scored 47 out of 100 on agentability across its homepage, pricing page (53), and documentation (40). This mid-range score indicates that while autonomous AI agents can interact with basic content, they encounter significant barriers when attempting to interpret system state, understand reasoning, or navigate non-standard interface patterns.

The platform performs adequately on technical fundamentals—machine readability (79) and control mechanisms (72) are solid—but falls short on principles that help agents understand context and decision-making. Transparency scored 0, meaning agents receive no insight into confidence levels, sourcing, or system reasoning. Shadow UI avoidance scored just 15, suggesting heavy reliance on client-side rendering or non-semantic interactive elements that agents struggle to parse. Defaults (33) also lag, forcing agents to supply parameters where reasonable assumptions could apply.

These gaps primarily affect agent workflows that require understanding why information is presented, tracing data provenance, or operating interactive features beyond simple navigation. Improving transparency and reducing reliance on opaque UI patterns would materially increase agent-readiness without requiring fundamental architectural changes.

To improve agentability, Confluence should focus on the following concrete changes:

• Add semantic HTML landmarks (<header>, <nav>, <main>, <article>, <section>, <footer>) in place of generic <div> and <span> elements to establish clear document structure.
• Implement a single <h1> per page and ensure heading levels follow a contiguous hierarchy without skipping (e.g., <h2> follows <h1>, not <h3>).
• Expose an activity log or audit trail in machine-readable format (such as a JSON event stream) with a visible 'why this result' affordance that records system actions and decisions.
• Attach source references, citations, or input identifiers to generated or dynamic outputs so agents can trace data provenance.
• Pre-fill form inputs and configuration fields with sensible defaults wherever a reasonable value exists, reducing the parameter burden on agents.
• Rephrase section headings as questions they answer (for example, 'How do I cancel my subscription?' instead of 'Cancellation') to improve navigability and intent matching.