Structured Context
Information an agent needs to make a decision should be available as typed, structured data with explicit units, relationships, and constraints.
Why it matters
Text written for humans relies on shared context: a reader knows that "Ships in 3–5 days" probably means business days, from the warehouse, after payment. An agent has to guess, and different agents will guess differently.
Structured context removes guessing from the parts of a decision that matter most: price, availability, eligibility, timing, and policy.
Examples
Product data includes price with currency, stock as a number per variant, and delivery estimates as a date range with time zone.
{
"sku": "TRAIL-RUNNER-42-BLK",
"price": { "amount": 129.00, "currency": "EUR", "includesTax": true },
"availability": { "status": "in_stock", "quantity": 7 },
"delivery": {
"earliest": "2026-10-09",
"latest": "2026-10-11",
"timeZone": "Europe/Berlin"
},
"returnPolicy": { "windowDays": 30, "condition": "unworn" }
}The price is rendered as an image, the currency is implied by the domain, and the return window appears only in a footer link.
Implementation guidance
- 1Identify the facts that drive decisions in your domain and give each a typed field.
- 2Prefer established vocabularies such as schema.org where they fit, and document your own extensions.
- 3Keep structured data and visible content generated from the same source so they cannot drift.
- 4State units, currencies, and time zones every time, even when they seem obvious.
Evaluation criteria
- Decision-critical facts are available without parsing prose.
- Structured values match what a human sees on the page.
- No field depends on locale guessing to interpret.