Walmart plans controls for AI shopping assistants

Separate the assistant’s purchases. Agent Profiles, due next month, would distinguish software-made orders. Shoppers would connect each assistant through Walmart and choose purchase approvals or spending limits.

Connect directly to the retailer. Walmart’s Agent Gateway would let approved assistants shop without clicking through human-facing webpages. It begins with a beta for companies and developers.

Keep a hand on checkout. Walmart can demand confirmation even below a spending limit. Agents cannot receive card details or change payment methods. The usual returns process would cover agent mistakes.

The merchant still sets the rules. Retailers elsewhere can study the same split: an assistant handles the conversation while the store keeps account authorization and returns. That is an operating lesson, not an overseas rollout.

Access needs an outcome. Walmart has disclosed no agent order or error baseline here. More completed purchases with fewer mistakes would establish value; allowing bots inside alone does not.

Businesses can buy directly. Advertisers can use technology partners or Ads Manager, OpenAI’s self-service campaign tool. Ads appear on Free and Go plans; Plus, Pro and Enterprise remain ad-free.

Paid space stays separate. OpenAI says ads do not influence ChatGPT’s answers and advertisers cannot read private conversations. More than 100,000 advertisers have tried campaigns since launch; that is a cumulative count, not today’s active customers.

Sales proof is still developing. OpenAI is expanding measurement, including early experiments to identify sales caused by advertising. A purchase recorded after an ad does not necessarily mean the ad created an extra sale.

Compare outcomes across markets. The portable question is whether conversational ads produce additional customers at an acceptable cost. More countries increase access; they do not answer that question.

Trials must become repeat budgets. Lifetime advertiser participation cannot show retention. Measurement improvements could build confidence, but evidence of recurring spending and representative sales results remains missing.

Displays and prices rose. Meta delivered 10% more ads worldwide in July–September. Its average price per ad rose 14% against a year earlier. That is total ad revenue divided by ads delivered, not a 14% increase in every advertiser's bill.

Markets diverged. The reported regional breakdown puts price growth at 23% in the US and Canada, 10% in Europe and 1% in Asia-Pacific. These are estimated locations of users receiving ads, not advertisers' home countries.

The cause remains partly unmeasured. Meta credits artificial intelligence with stronger advertising revenue. Its release does not quantify that contribution or establish whether an individual merchant earned more profit per customer order.

Measure the outcome that pays. Compare actual campaign spending with profit from the orders it produces. An ad display is not a sale.

Keep geography in the calculation. The global average conceals wide regional differences. It cannot replace evidence from the market a campaign actually reaches.

Fresh sales pull ahead. From August 2025 to August 2026, Amazon’s own perishables dollar sales rose 120.4%, versus 26.8% for shelf-stable goods and 21.9% for nonfood. These sales indices measure growth from each group’s starting level, not market share.

Several departments are accelerating. In the 12 weeks ending 29 August 2026, seafood sales grew 888.2% and produce 622.4% versus a year earlier, against 25.4% for traditional grocery. Fast growth can still start from small sales.

Fresh buyers return more often. In the 52 weeks ending 8 August 2026, 25.3% of US Amazon shoppers bought perishables. Their Amazon spending averaged $5,121 and 105 shopping occasions, versus $3,771 and 76 for the average buyer, who includes fresh buyers.

In Profitero+ findings, 31% of leading brands put social commerce, buying through social platforms, in their top two channels for understanding and growth over 12–36 months; only 6% chose agentic commerce, shopping with AI assistants.

Established online-focused retailers lead at 71% among leaders. Rapid local delivery reaches 47%, versus 36% among other brands: an 11 percentage-point difference in priority selections, not sales growth.

The 2026 benchmark surveyed over 200 ecommerce executives. Public materials omit geography and subgroup sizes, so leaders' answers cannot prove which channel choices cause success.

Comparison records 63%, learning 55% and discovery 46% for AI use across shopping tasks in the previous three months. Clear product information helps an assistant explain the options.

Basket building records 33%, support 39% and replenishment 21%. These independent task responses measure neither a checkout funnel nor completed autonomous orders.

The December 2025 survey covers 749 people, but task-question routing is unclear. Its separate overall shopping-use figure is 38%; 63% must not become an all-shopper adoption claim.

🤖 AI prompts: 14 AI Prompts for Customer Support Replies & Service or 👉 All here

🛠 Tools: 10 Best AI Copywriting & Product Description Tools for Ecommerce (2026) or 👉 All here

📖 Glossary: Headless & Composable Commerce, Decoded: 15 Terms You Actually Need to Know or 👉 All here

📅 Events: Connected Commerce Summit London 2026 or 👉 All here

PHP released emergency server fixes. Updating WordPress does not patch PHP: ask your host to confirm the fixed build or both vendor fixes, even if the version family is supported.

Meta says over 10 million people set up its Muse AI assistant in the US and Canada. About 25% use it daily; that is not a buyer count. Transaction fees remain a future plan.

ChatGPT ad changes become default: advertisers can opt out before the help page’s October 29 start. Existing opt-outs survive, and campaign settings override account defaults. Versions can run without advertiser sign-off.

Arkose says 97.7% of AI-agent sessions it judged legitimate lacked a verified signature. That is a gap in proving the sending product, not evidence those sessions were fraudulent; the sample covers its protected traffic.

Southeast Asia ecommerce is forecast to grow 18% to US$226 billion in 2026 across ten countries. The 75% AI-shopping figure measures conditional willingness among AI users in six markets, not completed purchases.

Airbnb limits routine cleaning chores for reservations booked from November 20, 2026. Hosts cannot add fees for skipped checkout tasks, but guests still owe basic care and cleaning fees can remain.

eBay plans to cut about 9% of staff, or around 1,000 jobs, after cutting about 800 roles in February. Those percentages refer to different headcounts; affected employee notifications are due in coming days.