When the one comparing prices is no longer your customer, but their AI agent

The buyer who never looks at your site

For twenty years we optimised for a hesitant human: a hooking headline, social proof, scarcity, an orange button. All that machinery assumes someone is looking at the screen. In a growing share of mid-consideration purchases, nobody is: the user asks an assistant to compare, and the assistant reads specs, price, delivery window and return policy across twenty stores before your banner has finished loading.

That reader is unmoved. It does not care about your lifestyle photography or your brand video. It cares whether the data it needs is present, specific and verifiable.

What makes an agent recommend you

• Complete, structured data. Dimensions, materials, compatibility, warranty, real delivery window by region. What is missing is not assumed: it is discarded.
• Visible price, no friction. A price that only appears after registration is a price the agent cannot read, and therefore a product that never enters the comparison.
• Consistency across sources. If your site says one thing, your marketplace listing another and your business profile a third, the agent lowers confidence in all of them.
• Returns and warranty written plainly. It is the field that weighs most when two options tie on price.

What this breaks in LATAM

Two things, both cultural before they are technical. The first is the habit of hiding price to force a WhatsApp conversation. That worked when human attention was the bottleneck; today it removes you from the list before any conversation exists.

The second is catalogue fragmentation. Many brands in the region keep real inventory in a spreadsheet, a stale public catalogue, and the true price only in the salesperson's head. An agent that finds three different versions of the same product does not pick the best one: it picks another brand.

What we would do first

Before touching any marketing, a data audit: every product carrying the same fields, with the same values, on the site, the marketplace and the business profile. It is boring work and it is the work that decides.

Then public pricing and real delivery windows by region, even when the window is bad. A long honest window competes; an absent one does not.

Finally, stop measuring traffic alone. Ask the assistants yourself about your category, in the words a customer would use, and note whether you appear, in what position and with which data. That manual check, once a month, tells you more than any traditional ranking report.

The advantage right now is the usual one with new channels: almost nobody is doing it, and the barrier is not budget but order.