McKinsey's agentic commerce survey, read for CPG brands: 48% of US adults now shop with AI

Most of the numbers about AI shopping this year have come from the companies selling it. Adobe counts traffic to retail sites, Forter counts orders on its network, Amazon tells investors how Alexa shoppers behave. Each is useful, and each sees the market through one window.

McKinsey's new report on agentic commerce, published October 28, comes from the other side of the counter. It is a survey of about 21,000 consumers in 11 markets, fielded in June 2026, asking what people actually did with AI when they shopped. Surveys have their own limits, which we cover below, but this is the broadest consumer read so far, and several of its findings land directly on packaged food, beauty and household brands.

Here is what it found, and what we think a CPG team should take from it.

Half of shoppers have tried it. Americans are a bit behind.

McKinsey asked whether people had used AI for at least one shopping task in the past three months. A shopping task can be anything from asking a chatbot to compare two moisturizers to letting an agent reorder paper towels.

Across the 11 markets, the average was 53%. In the US it was 48%, which McKinsey points out is below the global average even though the US leads in private AI investment.

Show the data

Group	Share of adults
Average, 11 markets	53%
United States	48%

Source: McKinsey, Who wins when AI does the shopping?, October 28, 2026. About 21,000 consumers aged 18 and older in 11 markets.

Younger shoppers are further along. McKinsey reports 62% of Gen Z and 63% of millennials across the markets have used AI to shop, and 57% of US Gen Z. Nearly a quarter of AI users use it daily.

For a CPG brand, the 48% figure is the one to put in front of a skeptical leadership team. This is no longer an early-adopter habit you can wait out. Roughly one in two American adults has already asked an AI tool something on the way to a purchase, and a meaningful share of those questions are about everyday products.

Using AI is common. Buying through it is still early, but growing fast.

The more important number for planning is how many people let AI carry the purchase all the way through. McKinsey found the share of US AI users who completed a purchase with AI grew from 10% to 17% year over year. Translated to the whole population, that is 5% to 10% of US adults. Across all markets, 14% of AI users had bought with AI, or about 9% of the population.

Show the data

Survey wave	Share of US AI users
Prior year	10%
2026	17%

Source: McKinsey, Who wins when AI does the shopping?, October 28, 2026. Average across the categories McKinsey measured. McKinsey equates 17% of US AI users to 5% to 10% of the US population.

By category, between 7% and 18% of AI users had completed a purchase with AI. Apparel leads. Grocery comes next.

That second place is worth reading twice. Grocery has long been one of the lowest e-commerce categories in the US, because people like to pick their own produce and because a weekly shop is a long list of cheap items. McKinsey's read is that grocery and beauty are catching up quickly through AI even with that low online share. A shopping list is exactly the kind of chore people are happy to hand to a tool.

There is also room for this to grow before anyone builds anything new. Between 21% and 30% of respondents said they would let AI complete a purchase, about 1.5 times as many as have actually done it. The willingness is ahead of the behavior.

Shoppers trust it more than most marketers assume

A common objection in brand meetings is that consumers do not trust AI with money. The survey says otherwise, at least for people who already use it. 74% of US AI users said they trust AI chat and search tools, close to the 73% global average and below India's 87%. Gen Z, often assumed to be the most comfortable group, came in lower at 67%.

The reasons people gave are practical, not technical.

Show the data

Benefit	Share agreeing
Narrows choices well	81%
Less scrolling	80%
Saves time	79%
Gets it right on the first try	67%

Source: McKinsey, Who wins when AI does the shopping?, October 28, 2026. Consumer survey fielded June 2026.

Read the top line from a brand's point of view. "Narrows choices well" means the assistant takes a shelf of 40 granola bars and shows three. The shopper is grateful for that. The 37 brands that were left out never find out it happened.

McKinsey also asked how influential different sources were. AI users were 3.2 times more likely to rank AI search as more influential than traditional search, and 2.6 times more likely to rank it above friends and family. That second comparison should worry any brand that has relied on word of mouth.

General AI tools get the questions. Retailers still get a lot of the checkouts.

The split between tools is where the survey gets most practical.

Tool type	Used for any shopping task	Used at the point of purchase
General AI chat or search tool	56%	14%
Retailer's own assistant	27%	15%

Source: McKinsey, October 28, 2026.

General tools like ChatGPT, Gemini and Perplexity win the research. Twice as many AI shoppers used them as used a retailer's assistant. But at checkout the two are almost level, with retailer assistants slightly ahead.

That matches what the retailers are doing. Amazon told investors on its second-quarter call that Alexa for Shopping users are 60% more likely to complete a purchase and spend 40% more, which McKinsey cites. Walmart and Target have built assistants of their own as well. And on October 28, the same day the report came out, Google said it is expanding ads in AI Mode to 10 more countries and reported that sponsored and organic product listings get similar click-through rates there. Every one of these companies wants the last step to happen inside its own walls.

For a brand, the practical meaning is that you need to be legible in two very different places. General assistants read your website, reviews, press coverage and feeds to decide what to recommend. Retailer assistants read the retailer's product listing. If those two sources disagree on pack size, price or ingredients, the shopper sees a different product depending on where they ask. We have written about keeping facts consistent before, and this survey is a good reason to recheck.

The structured data advantage is showing up in the numbers

McKinsey points to two figures that suggest clean product data is turning into an advantage you can measure. Shopify reports that AI searches on its structured catalog convert at about twice the rate of scraped data. And eMarketer estimates agentic commerce could add 11.5% to US e-commerce by 2030.

Shopify made the same point in its own way the same week. Its Fall '26 Edition says ChatGPT, Copilot, Meta and Instinct get product information from Shopify Catalog, and it added a view in the admin that tells merchants whether their store is eligible. When the agent reads a clean catalog instead of guessing from a web page, it makes fewer mistakes.

A separate sandbox study released the same day shows what those mistakes look like. General Machines ran more than 100,000 simulated shopping trips with one open-weight agent and found that reading product facts failed far more often than paying did. It is one agent in test stores, not a market survey, but it points the same way.

What the survey cannot tell you

A few cautions before anyone builds a plan on these numbers.

• It is self-reported. People describing what they did in the past three months will round and forget. "Used AI for a shopping task" is broad and could mean a single question.
• It is an average across categories. The 7% to 18% purchase range hides large differences. Your category could sit at either end.
• It was fielded in June. Several big launches came after, including Meta's Muse in September and Alexa for Shopping. Behavior has probably moved since.
• The money figures come from companies. Amazon's 60% and 40% and Shopify's 2x are company-reported numbers that McKinsey cites, not survey results.

None of that weakens the main point. It just means you should treat the survey as a direction, not a forecast for your brand.

What a CPG team should do with it

1. Use the 48% figure to reset the internal conversation. Many brand teams still file AI shopping under experiments. Half of US adults have already used it. The question is no longer whether to pay attention but who owns it.

2. Plan for grocery and personal care, not just apparel. McKinsey puts grocery second in purchase rates and names grocery and beauty as the categories catching up. If you sell snacks, beverages, supplements or skincare, assume your shoppers are among those asking.

3. Fix the facts an assistant has to read. Narrowing choices means comparing facts: size, count, flavor, ingredients, price per unit, certifications. If those are only in images or buried in a PDF, an assistant cannot compare them, and you get narrowed out. Put them in plain text on the product page and in your feeds.

4. Check both kinds of assistant. Ask ChatGPT and Gemini the questions your shoppers ask, then ask Rufus or Alexa, Walmart's assistant and any retailer assistant where you sell. Note where the answers differ and trace each difference to its source.

5. Watch whether you are recommended, not only whether you are clicked. Most of the value in the survey sits before any click: narrowing, comparing, deciding. You can check this by hand with a fixed list of shopper questions each month. Some teams use a tool for it; XENA Intelligence runs shopper-style questions across assistants such as ChatGPT, Gemini and Perplexity and shows which products come up in SearchPanel.

If you only do one thing this week, do the Shop Direct check in tip 39. It takes an afternoon, and it tells you whether Amazon's assistant already shows your products, and with which facts, even if you have never listed there.