Ten decisions about AI agents

The ten decisions

• Which AI bots and agents to let inFour kinds of AI visitor, four separate choices, one robots.txt that matches your bot protection.
• Whether to charge AI crawlers and agentsRevenue against reach: what charging for access can cost you in AI answers.
• Whether prices and policies are readable as textWhich prices, fees and policies are plain text on the page an agent reads.
• Whether an agent may complete a purchaseWhether a customer's agent may finish the purchase or must hand back the last step.
• Whether to publish an API or an MCP serverWhether agents finish tasks through an interface you publish, or through your web forms.
• Whether to accept agent paymentsWhich agent payment routes you accept, and through which provider.
• How to tell verified agents from unknown botsWhether agents that prove who runs them are treated differently from unknown bots.
• What to tell legalWhat your terms and consent rules say about agents acting for customers.
• What to measureTask completion and mentions in AI answers, tracked as two separate lines.
• What to budgetWhich line pays for agent experience, and what the first spend buys.

How to use them

Read the decision and its options with the people who own the outcome. Each page names who decides, lists three options with the consequence of each, and shows the dated facts behind them. Settle one decision per meeting rather than all ten, and write the answer on one page with a review date.

The sentence at the side of each decision is written for your web team. All ten are collected on ask your web team, together with the technical checks from the self-check.