Most merchants can now see some of the visits AI assistants send them, and few can follow those visits to an order. PYMNTS Intelligence's October 2026 report "The Agent Advantage: How Merchants Plan to Win AI Shopping" drew on a survey of 60 senior ecommerce decision-makers at U.S. merchants with $50 million to $1 billion in annual revenue. Among those selling through their own site or app, 37% could tell a visit came from an AI platform and lost the trail before the purchase, 51% attributed AI visits inconsistently, and 12% could not tell AI traffic from human traffic. None could fully and consistently connect AI-driven visits to completed purchases, PYMNTS reported on November 3.
Persa Flores, a family-owned florist in Spain, followed the trail with data its Shopify store already kept and found ChatGPT behind one in five of its online orders since September 2026. José Pérez Sánchez of Persa Flores described the method and the monthly numbers to Practical Ecommerce, which published them on November 1, 2026.
Shopify records the first and last visit before each order, including the referring site, the full landing page URL and the date. Pérez Sánchez counted an order as coming from ChatGPT when either visit came from ChatGPT or OpenAI. Every one of those orders arrived on a link tagged utm_source=chatgpt.com, a tag ChatGPT adds itself, and only about one in eight also showed ChatGPT as the referrer. Browsers and analytics tools drop the referrer because of privacy settings, site policies or a switch from HTTPS to HTTP, and the visit then reads as direct or unknown. A report built on referrers alone would have shown the florist about one in eight of its ChatGPT orders.
ChatGPT's share grew in every month the florist measured. It accounted for 4% of online orders in July 2026, 12% in August, 21% in September and 23% from October 1 to 27. Its share of online revenue went from 2% to 11%, then 18% and 26%. Pérez Sánchez said total online orders grew over the same months, so ChatGPT took a larger share of a business that was itself growing.
Most of the growth came before the florist did anything aimed at ChatGPT. Pérez Sánchez said the company started that work in mid-October, after the increase. Its city pages, each about sending flowers to one Spanish town such as Marbella, Segovia or Gijón, were already in place in Spanish and English. Of the florist's ChatGPT customers, 84% first arrived on one of those city pages, compared with about a third of customers who came from Google.
The ChatGPT orders came from new customers. Every one was a first order from that email address or account, most were placed within a day of the shopper's first recorded visit, and about half were paid from outside Spain, usually to send flowers to someone in the country. Average order value was about the same as for other online orders.
Tying orders to the assistant that sent them is also what makes AI channels comparable on cost. Nearly all the merchants PYMNTS surveyed expected to pay to take part in AI shopping, through commissions, subscriptions or other fees, and the report names connecting referrals to orders as the way to weigh those fees against the business they bring. It also notes that a shopper may ask an assistant, visit several stores and come back directly to buy, so a tagged visit shows an assistant took part in a sale without proving the sale was extra business.
Any Shopify store can run the florist's count from the first and last visit on each order. Filter for landing pages tagged utm_source=chatgpt.com as well as a ChatGPT or OpenAI referrer, and compare the share month by month.
Common questions
How many merchants can connect AI visits to sales?
None of the merchants selling through their own site or app in the survey behind PYMNTS Intelligence's October 2026 report could fully and consistently connect AI-driven visits to completed purchases. Thirty-seven percent identified AI visits but lost the trail before purchase, 51% attributed them inconsistently, and 12% could not tell AI traffic from human traffic.
How did Persa Flores count orders from ChatGPT?
It used the first and last visit Shopify records for each order and counted an order as coming from ChatGPT when either visit came from ChatGPT or OpenAI. Every one of those orders arrived on a link tagged utm_source=chatgpt.com.
Why did the referrer field miss most of these orders?
Only about one in eight of the florist's ChatGPT orders also showed ChatGPT as the referrer. Browsers and analytics tools can drop the referrer because of privacy settings, site policies or a switch from HTTPS to HTTP, and the visit is then labelled direct or unknown. The utm_source tag on the landing page URL stays with the visit when the referrer is dropped.
How fast did ChatGPT's share grow?
ChatGPT's share of the florist's online orders went from 4% in July 2026 to 12% in August, 21% in September and 23% from October 1 to 27. Its share of online revenue rose from 2% to 26% over the same months.
Which pages did ChatGPT send shoppers to?
Of the florist's ChatGPT customers, 84% first landed on a city page written about sending flowers to one Spanish town, compared with about a third of customers from Google. The pages existed before the ChatGPT traffic grew.