Visa partners with OpenAI to power AI agent payments in ChatGPT

What's happening

On June 10, 2026 at the Visa Payments Forum in San Francisco, Visa announced a strategic collaboration with OpenAI to plug the Visa network directly into ChatGPT, letting AI agents complete payments on a user's behalf. The same day, Visa unveiled Agent Scoring, an Agentic Registry, and a Large Transaction Model designed to make agent-initiated transactions auditable, scorable, and routable through the existing card rails.

The headline is consumer-facing — ChatGPT can now shop and check out for a user — but the practical effect for Amazon sellers is that one of the largest payment networks and the largest agentic platform have agreed on the plumbing for agent checkout. That makes ChatGPT-led discovery, comparison, and purchase a credible top-of-funnel for product demand that historically started with an Amazon search.

Why it matters for Amazon sellers

ChatGPT had agentic shopping flows already, but checkout was the missing link. Visa joining the loop solves that for the largest spend tier and gives merchants a reason to treat agent traffic as a real channel, not a curiosity. For Amazon brands, the bottom-of-funnel risk is the same as Rufus and Alexa for Shopping — a shopper completes the transaction inside the agent surface rather than the Amazon app or product page, so PDP sessions and Buy Box data underrepresent true demand.

The reverse opportunity is real too. Brands with clean structured data, consistent reviews, and accurate pricing will be more legible to an agent comparing options, and that legibility advantage shows up in both ChatGPT-driven sales and any future Amazon agent layer. The brands that win this cycle are the ones treating catalog hygiene as a paid-media input rather than a back-office chore.

What to monitor this month

1. 1.

   Direct traffic and dark social on the brand storefront

   An agent that completes a purchase off-Amazon still drives some shoppers back to the brand site for support, reorders, or post-purchase. Watch your direct-traffic baseline and brand-search volume — a rise without an obvious campaign is a strong signal that agent-led discovery is starting to move units. Pair it with the Amazon brand-search trend in day-to-day analytics.

2. 2.

   Pricing parity across marketplaces

   Agents compare price across surfaces in milliseconds. A 3 to 5 percent gap between Amazon and the brand DTC site that used to be invisible will start to leak units to whichever side is cheaper. Lock a parity rule, then read the impact on Amazon Buy Box and units in your SKU P&L.

3. 3.

   Catalog data quality on flagship ASINs

   Agents lean on structured data — title, bullets, attributes, A+ content, reviews. Run a hygiene pass on the top 20 SKUs by revenue: correct category, complete attribute table, clean variation parent, accurate compliance flags. This is the cheapest way to be legible to every agent that lands on your shelf in the next 12 months.

4. 4.

   PPC mix between branded and non-branded

   If agents intercept generic queries, non-branded Amazon search volume softens before the BSR moves. Watch the share of impressions on broad-match generic terms and rebalance toward branded defense and long-tail before competitors do. Mid-size brands tend to see this first in PPC analytics two to three weeks before category-level volume.

How Nova helps

Nova surfaces the early signals that matter when an off-Amazon agent layer starts to move demand. The winners and losers view flags units quietly slipping on flagship ASINs before the BSR alarms; the seller cockpit keeps pricing, sessions, and conversion in the same frame so a parity drift is obvious in one screen; and brand managers can hand finance a clean read on whether ChatGPT-driven shopping is incremental or cannibalizing the Amazon channel.