Adobe: AI shoppers now convert 90% better than everyone else. What it means for CPG brands this holiday.

For most of the last two years, the standard caution about AI shopping traffic was that it was small and did not buy much. People arrived from ChatGPT or Perplexity, looked around and left. Adobe's newest data, published October 29, says the second half of that caution no longer holds.

In September 2026, according to Adobe, visits to US retail sites that came from generative AI tools converted 90% better than visits from every other source. In August 2025, the same kind of visit converted 9% worse. That is a full reversal in a little over a year.

Below we go through what Adobe measured, what it forecasts for Thanksgiving week, and what we would do about it if we ran a packaged goods brand.

What Adobe measured

Adobe's report rests on more than 1 trillion visits to US retail sites, covering 100 million SKUs in 18 product categories. It adds a survey of more than 5,000 US consumers fielded in September 2026. That is a large base. It is also a base of retail sites, so it tells us how AI-referred visitors behave once they land on a store, not how often an assistant mentions a particular brand.

Adobe compares visitors who arrived from a generative AI source with visitors who arrived from other sources, which Adobe describes as channels such as email, paid search and social. Adobe counts AI traffic when a user clicks a link. Every figure below is a difference between those two groups in September 2026.

Show the data

Measure	Value
Conversion rate	90%
Revenue per visit	83%
Time on site	73%
Add-to-cart rate	53%
Pages per visit	25%

Source: Adobe, AI traffic and holiday report, October 29, 2026. Each bar is how much higher the AI-referred figure was than the non-AI figure.

The picture is consistent. People who come from an assistant stay longer, look at more pages, put more in the cart and buy more often. The revenue gap, 83% per visit, is close to the conversion gap, which suggests the effect is mostly about more people buying rather than bigger orders.

Why the gap flipped

Adobe frames it as shoppers letting AI do the research and arriving ready to buy. It reports that in 12 months AI traffic went from being worth less than half of a typical visit to nearly twice as much. Beyond that framing we would be careful about inventing causes, but a few things in the public record line up with the change, though.

First, assistants now do more of the shopping before the click. A shopper who asks for a gentle sunscreen for a toddler, reads a comparison and then clicks one link has already done the browsing that a search visitor does on the site. By the time they arrive, they are closer to a decision.

Second, there are more such shoppers. AI traffic to retail sites grew 71% year over year in September, Adobe says. When a channel grows from early adopters into ordinary use, its visitors usually start to look more like normal buyers. Here they went further and started to look like better than normal buyers.

Third, the assistants themselves added shopping features through the year. Amazon said on October 29 that more than 400 million customers had used Alexa for Shopping in the past 12 months, and Fortune reported that Alexa shoppers spend 50% more per order. That is a different channel from the one Adobe measures, but it points the same way: shoppers who use an assistant are often already in a buying frame of mind.

Adobe's data says the AI visitor is no longer a browser. In September, it was the most likely visitor on a retail site to buy.

What Adobe expects for Thanksgiving week

Adobe also published a forecast for the biggest shopping days of the year. It expects AI traffic to US retail sites to grow 159% year over year on Thanksgiving Day, 95% on Black Friday and 88% on Cyber Monday. Adobe also forecasts that the revenue-per-visit advantage for AI traffic will double over the holiday season.

Show the data

Period	Value
September (actual)	71%
Thanksgiving (forecast)	159%
Black Friday (forecast)	95%
Cyber Monday (forecast)	88%

Source: Adobe, AI traffic and holiday report, October 29, 2026. Thanksgiving, Black Friday and Cyber Monday are forecasts, not results.

Thanksgiving stands out. It is a day when many people are at home, cooking, and planning the next day's shopping on a phone. That is a natural moment to ask an assistant to compare deals or build a list. For food brands it also overlaps with last-minute grocery runs, which is where the next number matters.

Where grocery sits

Adobe's report includes what it calls citation readability: how much of a page's content an AI system can actually read. The average US retail homepage scored 63%. Electronics scored highest at 79%. Grocery scored 55%. General merchandise homepages scored 40%.

Table: Adobe citation readability by page or category, September 2026

Page or category	Share of content AI systems can read
Electronics, overall	79%
Electronics category pages	79%
Electronics customer service pages	78%
Average US retail homepage	63%
Grocery, overall	55%
General merchandise homepages	40%

We covered an earlier Adobe readability analysis in (post:grocery-sites-ai-readability), which put grocery at 59% in July. The two reports may not use the same sample of pages, so we would not read 59% to 55% as a decline. The consistent message is that grocery sits below the retail average, and below electronics by more than 20 points.

Adobe ties the gap to the kind of content on the page: electronics pages are full of setup guides, troubleshooting articles and detailed specs, while navigation-heavy, promotional layouts give AI less to work with. Why does that matter if the AI visitor buys more? Because the visit starts with the assistant's answer, and the answer is built from what the assistant could read. If half of a grocery page is invisible to an AI system, the answer about that product draws on less of what the brand said and more of what other pages said. The 90% figure describes what happens after the click. Readability affects whether the click happens at all.

Why CPG brands should not over-read the number

A 90% conversion advantage sounds like a reason to move budget. We would slow down for three reasons.

It is retail-site data. Adobe measures visits to retailers. A packaged goods brand's own site usually does not sell much directly, and many CPG purchases from an AI answer will end at Amazon, Walmart, Target or a grocer, not on the brand's site. The behavior is still relevant, because it tells you the shoppers reaching those retailers through an assistant are serious. But your own analytics may not show the same pattern.

It is a share of a still-small channel. Adobe does not say in this report what share of retail visits came from AI in September, and we would not guess. Growth of 71% from a small base is still a small base. In (post:ai-traffic-is-small-wrong-question) we argued that the size of the channel matters less than who it influences. Adobe's conversion data supports that view.

It is an average across 18 categories. Electronics and other considered purchases probably shape the average more than a bag of chips. Adobe's summary does not break conversion out for grocery, so we cannot say whether a cereal shopper from ChatGPT converts 90% better too.

What we would do this month

None of this needs a new channel strategy. It needs the basics done before Thanksgiving week.

1. Measure it in your own analytics. Set up an AI referral channel in GA4 so you can see visits from ChatGPT, Perplexity, Gemini, Copilot and others. Tip (tip:13) covers how. Compare their engagement with your other traffic. If the pattern matches Adobe's, you have your own evidence.
2. Make the top pages readable. Put product names, sizes, ingredients and key claims in HTML text, not only in images or scripts. (tip:14) walks through why nutrition panels in photos are a problem.
3. Check what the answer says before the click. Ask the three or four questions a shopper might ask about your category in ChatGPT, Gemini and Perplexity, and read the answers. Note which retailer links appear for your products.
4. Fix the retailer listings those answers send people to. If an assistant links your product to a Walmart or Target page with an old pack size, the 90% advantage goes to a page that disappoints.
5. Watch the holiday days, not the month. Adobe's forecast puts the biggest growth on Thanksgiving Day itself. Look at your data for that day separately.

Teams that want to track assistant answers at scale use tools for it, such as SearchPanel from XENA Intelligence. By hand works fine for a short list of products.

The bigger picture

The NRF said in September that 47% of consumers were using AI tools to help with their Halloween shopping, as we cover in (post:nrf-halloween-2026-ai-shopping-data). Adobe's report shows what happens next: the people who take an assistant's advice and click through are now the most likely visitors to buy.

For packaged goods brands, the practical conclusion is narrow. The assistant's answer is becoming one of the last steps before a purchase. The work is to make sure the answer is right, the page it links to is right, and you can see the traffic when it arrives.

Adobe will publish actual holiday results after Cyber Monday. We will compare them with this forecast when they arrive.