Guide
How to make your business readable by AI agents
Agent readability starts with the unglamorous work: clean offers, clear policies, structured data, and safe handoffs.
Updated June 18, 2026
Readable means more than crawlable
Search engines can index a web page that is confusing to an agent. An agent needs to know what the business does, who each offer is for, what it costs, what actions are possible, what policies apply, and where to route support or quote requests.
A readable business surface should be friendly to both humans and machines. The human page explains value. The machine-readable surface carries IDs, categories, prices, terms, support routes, risk notes, and contact paths.
The minimum public discovery layer
The first layer should be public-safe. It should contain only information the business is comfortable publishing and maintaining. It should not expose private topology, customer data, credentials, raw regulated records, or unsupported claims.
- A current `llms.txt` file for AI-readable site orientation.
- A structured product or offer feed with stable IDs.
- Pricing and quote-path pages that say what can and cannot be bought directly.
- Support and dispute pages that route requests clearly.
- Policy pages that state approval and human-review boundaries.
The first internal layer
Once public discovery is clean, the next layer is an internal tool inventory. This is not a marketing asset. It is a safety document that lists candidate agent actions, owners, data classes, required scopes, approval gates, and audit requirements.
Every tool should be classified before it is built. Read-only catalog lookup is different from creating a support ticket. Preparing an evidence packet is different from submitting it to a regulator or customer.
What DID checks
DID scores agent readiness across discovery files, product feed quality, pricing clarity, policy clarity, support clarity, quote-path clarity, structured data, tool inventory, approval gates, and audit logging.
The output is a practical implementation map, not a vague AI strategy document. It should identify what can be launched quickly, what needs review, and what should stay off limits.