How agent-ready is Grammarly?

Grammarly's overall agentability score of 44/100 (averaged across homepage, pricing, and documentation) indicates that autonomous AI agents will encounter moderate to significant friction when attempting to parse and operate the platform. The product performs adequately on defaults (83/100) and status visibility (65/100), but falls short in transparency (18/100) and shadow UI avoidance (15/100)—two principles critical for machine interpretation of user actions and system outputs.

The low transparency score reflects a lack of structured metadata around generated content: agents cannot easily trace suggestions back to their source rules, understand confidence levels, or access machine-readable audit logs. Similarly, weak shadow UI avoidance suggests heavy reliance on client-side rendering or dynamic interface elements that are difficult for agents to discover and interact with programmatically. Mid-range scores in machine readability (59/100) and chunking (59/100) point to incomplete semantic HTML structure and headings that do not clearly signal their purpose.

Documentation scored lowest at 36/100, indicating that agents seeking integration guidance or product capabilities will struggle. To improve agent-readiness, Grammarly should prioritize structured metadata, semantic HTML, and clearer exposure of system reasoning and state changes.

To improve agentability, Grammarly should implement the following changes:

• Add JSON-LD structured data using schema.org vocabularies to describe page entities, enabling agents to extract product features, pricing, and capabilities programmatically.
• Replace generic
  and elements with semantic HTML5 landmarks (

  ,
• Surface confidence scores or certainty labels on grammar and style suggestions, either as a field in API responses or as visible 'verified' vs 'best-guess' indicators in the UI.
• Expose an activity or audit log in machine-readable format (such as a JSON event stream) that records suggestions made, user actions, and document changes, with a visible UI representation.
• Rephrase documentation headings as direct questions (e.g., 'How do I cancel my subscription?') to help agents quickly locate relevant procedures.
• Attach source references—such as rule IDs, style guide links, or grammar principles—to each generated suggestion so agents can trace reasoning and validate outputs.