Price transparency and AI agents

Deferring costs until checkout is a long standing tactic that works reasonably well on human shoppers and inverts completely against agents. It is worth understanding why, because the instinct behind it is strong and the consequence is invisible.

An unknown is not an opportunity

A human who cannot see your shipping cost may click through, get invested, and accept it. That is what the tactic is built on.

An agent computing a delivered total for a budget constrained request cannot proceed on an unknown, and it does not resolve one optimistically. It has a competitor with a computable total, and that competitor is the safe answer.

So the store with genuinely cheap delivery and no published rule loses to the store with worse delivery and a stated number.

What has to be computable

• Item price matching what a shopper actually pays, including discounts applied at render.
• Shipping cost for your main market, as a value, with any free threshold stated as a number rather than shown in a graphic.
• Tax and duty position, said plainly, particularly for cross border.
• Anything else that appears between the product page and the payment total.

The failure that never surfaces

This is a ranking loss rather than a checkout loss, so nothing in your analytics records it. There is no abandoned cart, no bounce and no error. There is a comparison you were dropped from, and comparisons do not report themselves.

That is why this survives in stores that are otherwise carefully optimised. See how agents compare shipping costs.

The check

Take a product and, using only what is published before checkout, work out what a shopper in your main market pays to have it delivered. If you cannot, nothing else can either. See how to publish shipping costs machine readably.