ChatGPT's Shop at buttons sent 56% of shoppers to big-box stores and about 5% to the brand

A shopper asks ChatGPT for the best unscented laundry detergent for a baby's clothes. ChatGPT names three products, writes a few lines on each, and since October 10, 2026, it may put a button under each one: Shop at Walmart, View at Target, Buy at Amazon. The brand gets the recommendation. Someone else gets the click.

That is the short version of a study published on October 14 and updated on October 16 by WISLR, a company that tracks how AI assistants answer shopping questions. It counted where ChatGPT's new buttons pointed for four product brands across 250 shopping prompts. The brand's own store came last by a wide margin.

None of the four brands in the study is a grocery or household brand, and the sample covers four days. So read the exact percentages as provisional, which is how WISLR describes them too. The direction, though, matches what OpenAI itself says about how it ranks merchants, and it is the direction that matters for a CPG team planning the holidays.

What changed on October 10

Before October 10, fewer than 1 in 1,000 ChatGPT answers WISLR tracked carried a retailer button. On the October 10 run, 11% of answers to shopping prompts had one. From October 11 to 13 the share held between 16% and 20% a day.

Day	ChatGPT	Perplexity	Google AI Overviews
Sept 20 to Oct 9	under 0.1%	under 0.1%	0%
Oct 10	11%	0%	0%
Oct 11	20%	0%	0%
Oct 12	20%	0%	0%
Oct 13	16%	0%	0%

Source: WISLR, October 14, 2026. Same prompts on each platform.

The buttons appear only on shopping and buying questions. Ask how a product works and there is no button. Each recommended product shows up as a block with a photo, a badge such as "Best overall," a price, a short description, pros and cons, and then the single retailer button. WISLR ties the timing to OpenAI's GPT-6 rollout with what OpenAI calls Intelligent UI, which reached paid users on October 7 and free users on October 8, while noting that it cannot prove the link.

Perplexity and Google AI Overviews added nothing similar over the same days. For now this is a ChatGPT surface.

Where the buttons point

Across the four brands, 56% of the buttons pointed to a big-box retailer: Best Buy, Walmart, Target, Amazon and similar. Competitor brands and specialist retailers took 39%. The brand's own store got about 5%.

Show the data

Destination	Share of buttons
Big-box retailers	56%
Competitor brands and specialist retailers	39%
The brand's own store	5%

Source: WISLR, October 14, 2026. Four days of data and small per-brand counts; WISLR calls the direction solid and the exact percentages provisional.

WISLR also tried a category it does not track. Asked for a durable carry-on suitcase, ChatGPT recommended a Travelpro with a button to Nordstrom, a Briggs & Riley with a button to Macy's and a Samsonite with a button to Walmart. None of the three buttons went to the maker.

For a CPG brand the picture is probably sharper, not softer. Most of our readers sell the same item through Amazon, Walmart, Target, Kroger and a DTC site. When ChatGPT has five places to send a shopper, the brand's own store is the one with the least stock depth, the highest shipping threshold and the fewest reviews on the page. OpenAI's own help page on ChatGPT shopping says merchants are ranked on "availability, price, quality, and whether they are the maker or primary seller of that item." Being the maker is one factor of four. A shampoo two-pack that costs more after shipping on the brand's site, and sits behind a free-shipping minimum, loses on price and availability to the same two-pack at a big-box store that ships it with the rest of the basket.

The brand still gets named

The part of the study we found most useful is the before and after on the same prompts. The buttons did not make ChatGPT less likely to name the brand. It named the brand in 40% of shopping answers before October 10 and 43% after. It cited a big-box site as a source about 15% of the time on both sides. What moved was the links.

Before Oct 10After Oct 10

Show the data

Measure	Before Oct 10	After Oct 10
Names the brand	40%	43%
Cites a big-box site as a source	15.2%	15.3%
Links to a big-box store	3.7%	19.2%
Links to the brand's own store	3.2%	12.8%

Source: WISLR, October 14, 2026

Two readings follow from that chart. First, answers that link to the brand's own store went from 3.2% to 12.8%, so the change sent brands more clicks than before, not fewer. Second, links to big-box stores grew much faster, from 3.7% to 19.2%. In WISLR's words, the retailer arrives with the button, after ChatGPT has done its reading.

That has a practical consequence for anyone who reports AI visibility to a founder or a board. A prompt tracker will keep showing the brand holding steady. Meanwhile more of the purchases those answers lead to will go through a retailer's checkout, where the brand sees a sale in its retail sales data weeks later, if it can see it at all. Visibility and traffic are now two separate lines, and they can move in opposite directions.

The one lever WISLR found

WISLR checked whether the brand's own product page was among the sources ChatGPT read for the answer. That split the results sharply.

Show the data

Brand's product page was a source	Share of buttons to the brand's store
Yes	34%
No	6%

Source: WISLR, October 14, 2026. A stricter brand-matching analysis gave 39% against 7%. Correlation, not proof that adding a source causes the button.

When the page was a source, the brand's store got the button about a third of the time. When it was not, about one time in sixteen. A stricter matching method gave 39% against 7%. One of the four brands got no buttons to its own store at all over the four days.

This is a correlation. It does not show that getting your page cited causes the button to come to you, and WISLR does not claim it does. But it is the only lever anyone has measured, and the work it points to is work most CPG sites need anyway. WISLR's own recommendation is to put the price, availability and Offer structured data in the HTML the server sends, so an assistant reading the page gets them without running JavaScript. We covered how to check that in what crawlers see before you enrich product copy, and the basics are in tip 15.

The week ChatGPT clicks went missing

The study also explains a dip many teams saw in October and could not account for. From October 8 to 15, every outbound ChatGPT link WISLR tested opened with no referrer and no utm_source tag: buy buttons, source pills and ordinary citations alike. All 47 links across nine answers on October 14 came through untagged. With neither signal, GA4 and every other analytics tool files the visit as Direct.

On one store WISLR measures, arrivals from ChatGPT from October 9 to 13 averaged 66% less than over October 1 to 7, and a second store fell 57%. Gemini arrivals on the first store rose 16% over the same days. OpenAI confirmed the missing tag was a bug in the GPT-6 rollout to Adobe's Cedric Huesler, and a fix began reaching users on October 15. When WISLR tested again on October 16, every link carried utm_source=chatgpt.com. None passed a referrer.

That makes the tag the only thing standing between a ChatGPT visit and the Direct bucket. It is a fragile arrangement. Gemini went through a version of it a few weeks earlier, when it stopped adding utm_source=gemini to cited links around September 28 after adding them for three weeks. If your AI traffic report relies on either tag, you are relying on a product decision you do not control.

What the button does not change

Before anyone rewrites a channel plan, keep the scale in view. ChatGPT is not Google, and Google behaves differently again. A Findrix study of 55,198 answers found that Google AI Mode linked only to Google's own pages in 46.5% of US product answers, and that ChatGPT linked to Walmart in 45.1% of product answers and Amazon in 6.1%. Each assistant has its own habit for where it sends a shopper. ChatGPT's button is a new and visible version of a pattern that already existed: assistants recommend brands and hand the sale to retailers.

What we would do this month

1. Annotate October 8 and October 15, 2026, on every chart that shows ChatGPT, AI referral or referral-channel traffic. Compare only windows that sit fully after October 15. Our guide to finding AI traffic in GA4 shows where to put the segment.
2. Run your top 20 buying questions in ChatGPT and write down, for each answer that names you, which retailer the button points to. That tells you which listing is now your landing page. Tip 12 has a simple logging method.
3. Open that listing as a shopper would. Check the title, size, price, stock and the first image against your own page. If ChatGPT is sending people to Target for your product, the Target page is where the sale is won or lost. Tip 30 covers why retailer listings beyond Amazon matter.
4. Make your own product page readable without JavaScript, with price, availability and Offer markup in the server HTML. That is the one factor the study connects to the button pointing at you.
5. Track mentions and visits separately. If mentions hold and visits fall, the answer changed how it sends people, not whether it names you. If both fall, you have a visibility problem.

Teams that already log assistant answers can add one column, "button retailer," and get a usable picture in a week. Some brands do this by hand; XENA Intelligence runs shopper-style questions across assistants and keeps the answers per product in SearchPanel, which makes the button easy to log over time.

The bottom line

ChatGPT's Shop at button is not a penalty on brands. It sent brand stores more clicks than before, just far fewer than it sent to big-box retailers. The study says two things a CPG founder can act on. Your retailer listings are now part of your AI visibility work, because that is where a recommended shopper lands. And your own product page matters less as a destination than as a source: when ChatGPT reads it, your store gets the button more often. Both jobs are unglamorous, and both are cheaper than buying the click back with ads.