Shoppers buying food and groceries online across the Americas spent 17% more per checkout between July 1 and August 18, 2026 than in the same weeks a year earlier, even as visits to retailer sites fell 8%, according to CPG Pulse 2026, a 35-page study that Criteo listed on its resources page on September 10. The same document places ChatGPT inside the consumer packaged goods funnel: in July, 86% of Criteo's CPG clients recorded at least one transaction referred by the chatbot.

In Short

Criteo, a company that sells online advertising for shops and brands, looked at what people bought in everyday categories like food, shampoo and cleaning products, and asked more than 6,300 shoppers in six countries how they buy them. It found that shoppers in the Americas are visiting grocery websites less often but spending more each time, and that ChatGPT is now sending buyers to most of the packaged-goods brands that advertise steadily with Criteo. If you sell or advertise everyday products, the report points to repeat customers and AI assistants as places where money is moving, though almost all of the numbers come from Criteo's own network and have not been checked by anyone else.

What the report measures

CPG Pulse 2026 follows Pulse reports on travel (May 4), health and beauty (June 30) and consumer electronics (August 6) this year, according to Criteo's resources page. The study combines two kinds of evidence. Criteo Commerce Datacovers transactions, browsing and baskets from what the report calls hundreds of brands and retailers. The Criteo Shopper Survey was fielded in June 2026 in the United States, the United Kingdom, France, Germany, Japan and South Korea.

The cover cites more than 6,300 consumers, but slide bases vary: 3,496 for most questions, 6,042 for shopping channels, 2,696 for discovery preferences and 6,307 for AI reordering. Two slides describe their base in identical words - category buyers in the past 30 days - yet carry samples of 3,496 and 6,042, a difference the report does not explain. No margins of error, weighting or fieldwork vendor are published. Observation windows shift too, from the fourth quarter of 2025 to single weeks in August 2026, and the regional figures classify CPG by retailer vertical while the holiday figures use product catalogues supplied by CPG partners. The document is a set of snapshots rather than one dataset.

Fewer visits, heavier baskets in the Americas

Year-over-year for July 1 to August 18, the Food & Groceries segment in the Americas lost 8% of its traffic and 4% of its sales. Yet conversion rate rose 3%, spend per order climbed 17% and revenue grew 12%. Non-Food fared worse: sales fell 6%, traffic and conversion each dropped 3%, and an 8% rise in average order value lifted revenue by just 2%.

The numbers broadly reconcile. Traffic down 8% and conversion up 3% yield roughly 5% fewer orders, and 17% more per order on that smaller base gives growth of about 11%, within rounding of the reported 12%. The same check works within a point in every region, which indicates that "sales" in the report counts transactions; the APAC slide calls it sales volume. What the report does not do is separate price from quantity. A slide heading says shoppers are "going bigger at checkout", but a 17% rise could equally reflect higher shelf prices or larger pack sizes.

EMEA and APAC run the other way

In EMEA, Food & Groceries sales rose 11% on 4% more traffic and a 7% higher conversion rate, but average order value fell 5%, so revenue grew 6%. Non-Food sales fell 5% and revenue 3% despite orders 3% larger - "a sharp divergence between the two segments," in the report's words. APAC grocery volume grew fastest of the three regions, up 13%, with traffic up 10%, conversion 3% and revenue 11%, while order value eased 2%. Non-Food volume grew 2%, but a 10% drop in order value cut revenue by 8%.

In Food & Groceries, the Americas bought less often and spent more each time; EMEA and APAC did the reverse. For media sold on sessions and clicks, the Americas grocery segment offered fewer of both over the summer.

Holiday peaks belonged to beauty

Against average October 2025 sales, Health & Beauty in the Americas ran 26% higher in November and 23% higher in December, while Food & Beverages rose 5% and then 14%. In EMEA, Health & Beauty jumped 37% in November and fell back to 2% above baseline in December; Food & Beverages rose 8% and 10%. APAC barely moved, with neither category more than 4% above October in either month. EMEA's uplift was concentrated in the month containing Black Friday, though the indices do not isolate that week.

The grocery run still ends at a till

Restocking rhythms differ sharply by country. In Japan, 69% of recent buyers shop for groceries or household items daily or several times a week, and in Germany 61%. In France only 24% shop that often, and 40% shop every two weeks or less.

Stores remain the default. Some 60% of recent grocery buyers and 64% of household product buyers purchased only in a store, against 11% and 20% who bought only online. Cosmetics and perfumes lean the other way: 36% of buyers purchased online only and 21% through both channels, so 57% involved an online purchase. Asked how their grocery, hygiene and pet food spending divided over 30 days, 29% of respondents described a roughly even split, 31% spent mostly online and 41% mostly in stores. South Korea leans hardest online, with 23% spending more than 90% there. Japan leans hardest the other way: 29% spend more than 90% in stores.

Online is still gaining ground. Globally, 37% spend more online on these categories than a year ago and 18% spend less. The balance is widest in the UK (43% against 17%) and the US (43% against 18%), and narrowest in France (37% against 29%).

For a retail media network selling placements on a retailer's own website and app, those shares mark how much of the grocery trip is visible at all. Stores are not beyond advertising's reach. Grocery TV's own survey of 1,018 US grocery shoppers in March 2026 found 62% reporting a purchase directly after seeing a product advertised on an in-store screen.

More retailers, more brands in view

Shoppers who browse several retailers encounter more CPG brands than those who stick to one, in every market measured, according to Criteo's data on grocery, household and Health & Beauty purchases made between July 25 and August 7, 2026. South Korea shows the widest gap, about 5.7 brands against 2.1. Reading from the chart, the UK follows at close to five against just over two, while the US, France and Germany sit near 3.7 to 3.9 against 1.8 to 2.1. The comparison is descriptive. Multi-retailer shoppers are, almost by definition, more engaged, and the report does not control for that.

Asked about preferences, 41% of online grocery and household buyers wanted to see every option and filter for themselves, against 33% preferring personalised recommendations. South Korea is the exception, with 44% choosing personalisation.

Promotions move almost everyone. Asked about their online grocery, hygiene and pet care purchases over the previous 30 days, 78% said they had bought at least one item because it was on sale, with Gen X, Millennials and Gen Z within a point of each other. Trial and ad influence split by age. Six in ten bought at least one new brand, rising from 51% of Gen X to 68% of Gen Z, and 52% said ads motivated some of their purchases, from 44% of Gen X to 60% of Gen Z.

ChatGPT enters the aisle

According to Criteo, from January through July 2026 "ChatGPT's share of product-page landings was 1.6x that of search engines, on average." The slide offers no definition. The plausible reading is a mix measure: a larger proportion of the visits ChatGPT refers arrive directly on a product page. A volume reading would clash with Ahrefs research published in February 2026 finding ChatGPT sending roughly 190 times less traffic than Google. Read as mix, the figure says ChatGPT visitors land deeper in the funnel, not that there are more of them.

The second chart counts CPG clients with consistent ad spend from January to July 2026 that recorded at least one ChatGPT-referred transaction: 79% in January, 75% in February, 78% in March, 81% in April, 85% in May and 86% in June and July. A single transaction in a month qualifies. The slide text says "Criteo CPG clients" while the chart says "Criteo CPG Partners"; neither gives a count, and paid placements are not separated from organic links inside answers.

January's 79% predates paid ChatGPT advertising altogether. OpenAI began formally testing ads on February 9, and Criteo became the first advertising technology partner in the pilot on March 2, so the early months capture organic referral. The climb from April tracks the growth of Criteo's integration to more than 1,000 brands by May 5 and more than 2,000 by June.

Criteo is not a neutral observer. It sells ChatGPT inventory through a Model Context Protocol layer that feeds data on more than $1 trillion in annual ecommerce transactions and about 5 billion product SKUs into AI agent environments. Its earlier claims ran the same way: conversion about 1.5 times that of other referral channels across 500 US retailers in February, approaching twice traditional search in some categories by May, and roughly 80% of paid traffic through the integration arriving new to the brand. None of the five leading ChatGPT categories it listed in June - apparel, home furnishings, consumer electronics, automotive and beauty - was food or household goods. The CPG Pulse measures presence, not performance. It contains no conversion rate, revenue share or order value for ChatGPT-referred CPG sales.

Independent evidence is mixed. An academic study of 973 ecommerce sites found ChatGPT referrals converting below every major channel except paid social. And on September 2, NIQ and Similarweb said they would start measuring shopping inside ChatGPT and Gemini from the fourth quarter of 2026, a sign that the measurement gap remains open.

Letting a machine restock

Asked which categories an AI assistant could reorder automatically when supplies run low, 40% of 6,307 respondents chose personal care such as shampoo and toothpaste, 34% household cleaning, 32% groceries and staples, and 28% vitamins or supplements. A third would not trust AI to reorder products automatically. Stated comfort with a hypothetical feature is not behaviour, and the question frames the easiest case - replacing a known product - where agentic AI has least to decide. Analyst Andrew Lipsman argued in October 2025 that agents may not solve enough of a problem in commodity household essentials to justify adoption, while Zeta Global found parents more willing than non-parents to let agents reorder household goods.

Ads and AI travel together

In all six markets, shoppers who said ads motivated some of their recent online CPG purchases were more likely to use AI assistants for grocery and household shopping: 55% against 21% in the UK, 57% against 26% in the US, and 35% against 17% in Japan, the narrowest gap. The report counts anyone using AI for "some" purchases as a frequent user, a generous threshold. It is also a correlation between two self-reported behaviours. Heavier online shopping could drive both, and the report does not test for it.

Segments built from baskets

Two US slides show how Criteo turns purchase histories into audiences. Shoppers who bought from at least two home-cooking categories in at least two separate months between April 1 and June 30, 2026 were labelled home-cooking enthusiasts. From July 1 to August 15, they were 4.1 to 5.4 times as likely as other active shoppers to buy fresh food and dairy, and category buyers among them spent 41% to 73% more. Part of that is built into the definition, since a group selected for buying fresh food in spring will buy more fresh food in summer. The result shows a habit persisting, not that targeting changes anything - the distinction incrementality testing exists to draw.

The young-family segment was 2.9 to 4.4 times as likely to buy selected products, led by yogurt and vitamins. Its spend data says more than the slide's summary. Young-family vitamin buyers bought 47% more units but spent 6% more; for household cleaning, 19% more units meant 3% more spend. That is consistent with cheaper items or multipacks, though the report does not say which.

The Black Friday cohort

Criteo tracked first-time buyers at major retailers whose first purchase, made during Black Friday weekend from November 28 to December 1, 2025, included CPG items, then followed them at the same retailer for six months. Those who bought again spent roughly six times as much as those who did not, with orders only about 6% larger. They returned to buy four more times, and nearly six in ten made their next purchase within four weeks.

By country, the multiple ranges from 9.4 in South Korea and 6.4 in Japan to 4.7 in the US and UK and 2.8 in Germany. A simple average of the six multiples is 5.7, consistent with the rounded global figure, while the six order-value gaps average 5.2%, a little below the report's "about 6%". The report does not say how its global figures are calculated. There is a structural caveat as well. Repeat buyers are defined by buying more than once, so a large gap over one-time buyers is close to guaranteed; the informative parts are the size by country and the speed of return.

CPG against everything else

Within the same US cohort, CPG items cost approximately 50% less than non-CPG items, according to the report, yet buyers purchased 3.9 times as many units and bought 18% more often, producing 29% greater spend. The figures do not reconcile. Half the price multiplied by 3.9 times the units implies spend close to double, whereas the published 29% implies CPG items costing about a third as much. The report does not explain the gap.

Carts that reach checkout

Baskets containing CPG items were abandoned 9% less often globally than baskets with only non-CPG products, and 13% less often in the US, among new basket users between August 1 and 7, 2026. These are relative differences on undisclosed base rates. Checkout is the leakiest stage of the conversion funnel: Baymard Institute, aggregating 50 studies, puts average cart abandonment at 70.22%, and DHL found 67% of shoppers had abandoned a purchase over the delivery offering.

Loyal buyers keep switching

Across eight markets, the six survey countries plus Canada and Brazil, Criteo measured how often a shopper's latest category purchase between July 25 and August 7, 2026 included a brand absent from the preceding purchase. In Candy & Chocolate, 51% of shoppers with four or more prior purchases did so, against 52% of those with one. Skin Care stood at 50%, Household Cleaning Products at 47% and Shampoo & Conditioner at 39%, all broadly flat across frequency bands. Coffee showed the clearest decline, from 32% to 27%, the one category where repetition visibly built habit.

The metric counts adding a second brand or rotating between familiar ones, so it measures fluidity rather than defection. Other data points the same way. Private label sales in Europe reached €352 billion, with retailer brands contributing 54% of grocery market growth, and Ibotta's 2026 State of Spend survey of more than 5,000 US grocery shoppers found 62% choosing price over brand, according to Ibotta.

Smaller retailers, bigger orders

Between May 1 and July 31, 2026, shoppers at small and mid-size retailers bought from 7% fewer categories and 12% fewer brands and made 25% fewer transactions than large-retailer shoppers, but their orders were 29% larger, leaving food and beverage spend per shopper broadly level. Large retailers are the top 10% by category sales in each country. Shoppers using both kinds are counted in both groups, which narrows the measured gap, and South Korea weighs as much as the United States in the average.

What Criteo is selling

The closing slides turn findings into prescriptions: follow each market's rhythm, coordinate across retailers and the open internet, activate across multiple retailers while consideration sets form, and use commerce signals to anticipate replenishment. Those map closely onto what Criteo sells, from the technology serving as commerce plumbing for roughly 225 retail media networks to open-web media and, since March, ChatGPT access.

The timing is awkward for the author. Second-quarter revenue fell 11% to $428 million and full-year guidance was cut on August 5, with Criteo pointing to a DoorDash pipeline concentrated in beauty, personal care, and food and beverage - the categories this report covers - under a partnership agreed on October 6, 2025. Vista Equity Partners and Quinti Capital offered a premium of more than 50% for the company, Reuters reported on July 6, citing sources familiar with the matter.

None of that invalidates the data, but it frames it. For CPG advertisers, the weightiest findings are the least glamorous: baskets in the Americas growing as traffic shrinks, loyalty that barely strengthens outside coffee, and a grocery basket that 60% of recent buyers purchased only in a physical store. The ChatGPT figures show the chatbot producing at least one sale a month for most of Criteo's CPG clients. How much it sells for them is a number the report does not contain.

Timeline

Summary

Who: Criteo, the commerce media company, published the study. It draws on Criteo Commerce Data from hundreds of brands and retailers and on a June 2026 survey of more than 6,300 consumers in the United States, the United Kingdom, France, Germany, Japan and South Korea. CPG brands, retailers, retail media networks and agencies buying grocery, household and personal care media are the affected parties.

What: CPG Pulse 2026, a 35-page report covering regional CPG performance, holiday seasonality, channel habits, cross-retailer browsing, ChatGPT referrals, AI reordering, audience segments, repeat buying, cart abandonment and brand switching. Headline figures include a 17% rise in spend per food and grocery order in the Americas against an 8% fall in traffic, 86% of Criteo CPG clients recording at least one ChatGPT-referred transaction in July, and repeat Black Friday buyers spending roughly six times as much as one-time buyers. Several figures are undefined or do not reconcile internally, including the 1.6x ChatGPT landing metric and the CPG versus non-CPG spend comparison.

When: Listed on Criteo's resources page on September 10, 2026, with a summary blog post on September 15. The underlying data runs from a year-ago comparison base starting July 1, 2025 to August 18, 2026.

Where: Commerce data covers the Americas, EMEA and APAC, with country detail for the US, UK, France, Germany, Japan and South Korea, plus Canada and Brazil for brand switching.

Why: Criteo is using the report to argue that CPG demand is fragmenting across retailers, channels and AI assistants, and that repeat buyers carry most of the long-term value. The study arrives as the company absorbs falling revenue, sells ChatGPT access to advertisers and faces an unresolved takeover approach, which gives it a direct commercial interest in the conclusions.