AI agents don't like to be wrong. When they recommend a product and the user has a bad experience, that's a failure. Agents are designed to minimize failure.
Incomplete data creates uncertainty. Uncertainty creates risk. Risk makes agents recommend someone else.
The Certainty Score (0-100) measures how "buyable" a product is for an autonomous agent.
The Scoring Framework
Availability (20 points)
- Real-time inventory count (not just "in stock"): +10
- Clear in-stock/out-of-stock/backorder status: +5
- Estimated restock dates when relevant: +5
Pricing (20 points)
- Exact current price (not "starting at" or "from"): +10
- Currency explicitly specified: +5
- Sale/regular price clearly distinguished: +5
Fulfillment (20 points)
- Shipping time as integer days: +10
- Shipping cost as exact amount or clear threshold: +5
- Express/expedited options documented: +5
Returns (15 points)
- Return window in days (integer): +5
- Free returns or exact return cost: +5
- Policy clearly stated: +5
Attributes (25 points)
- All category-relevant specs present: +15
- Boolean/integer format (not descriptive language): +10
Example: Two Products
Product A: Score 94
{
"price": 79.99,
"currency": "USD",
"inventory_count": 234,
"ships_in_days": 2,
"returns_window_days": 30,
"returns_free": true,
"sweat_resistant": true,
"battery_hours": 8
}Product B: Score 52
{
"price": "From $69.99",
"availability": "In Stock",
"shipping": "Fast shipping",
"returns": "Easy returns",
"sweat_resistance": "Built for workouts",
"battery": "All-day battery"
}Product A is agent-ready. Product B is human-ready. In agentic commerce, Product A wins—even if Product B is objectively better.
Why Descriptive Language Fails
"Fast shipping" requires interpretation. How fast? Compared to what?
ships_in_days: 2 is deterministic. The agent can compare, filter, make guarantees.
"Built for workouts" is marketing. Does it handle sweat? Can it survive drops?
sweat_resistant: true and drop_tested_feet: 4 are facts the agent can use.
The Thresholds
Below 70: Risky for agents. They'll prefer higher-certainty alternatives. 70-84: Acceptable. May be recommended with caveats. 85-94: Agent-ready. Confident recommendations. 95+: Agent-optimized. Wins competitive categories.
How To Calculate Yours
- Pull your product feed as an agent would see it
- Score each attribute against the framework above
- Be honest about what's machine-readable vs. requires human interpretation
- Calculate the total for your top products
Most brands I've audited score 40-60. They have the information—just not in a format agents can use.
Quick Wins To Improve Your Score
+10 points: Replace "In Stock" with actual inventory counts +10 points: Replace "Fast shipping" with integer days +5 points: Add explicit currency codes +10 points: Convert descriptive attributes to booleans/integers
The gap between where you are and 85+ is the opportunity. Most competitors haven't closed it yet.