What Is Agentic Shopping? How Close Are We to Just Saying “Add It to My Cart”?

In brief: Agentic shopping, or agentic commerce, is a shopping model in which AI agents search for products, compare them, and make purchases on someone's behalf within the authority they have been given. McKinsey projects that this model could facilitate $3–5 trillion in global sales by 2030. The payment infrastructure was put in place in 2025. But the first lesson of 2026 was this: customers discover products through AI, but still complete most purchases on the brand's own website. The winners will therefore be brands that can sell through conversation to both AI agents and the people visiting their sites.

---

What is agentic shopping?

Agentic shopping is a commerce model in which AI agents handle part or all of the shopping process on a consumer's behalf: understanding the need, searching through options, comparing them, preparing the cart, and making the payment if authorized.

McKinsey describes this as a model in which AI agents shop, negotiate, and transact on behalf of people (Digital Commerce 360, citing McKinsey, October 2025). An article in Harvard Business Review by authors from INSEAD and Artefact defines agents simply as “algorithms authorized to act on a user's behalf” (HBR, February 2025).

The difference from traditional e-commerce can be summed up in one sentence: Until now, customers browsed websites. In agentic shopping, customers tell an agent what they want, and the agent browses the websites.

The three levels of agentic shopping

In practice, it is more useful to consider agentic shopping at three levels rather than as a single concept:

1. Advisor level: The agent recommends and compares products; the customer makes the decision and pays. This is the most common use today.
2. Copilot level: The agent prepares the cart, and the customer approves with a single tap. This is the level of the “add this to my cart” experience.
3. Autonomous level: The customer sets rules in advance, such as “buy this product if it drops below TRY 2,000.” The agent makes the purchase itself when the time comes.

The infrastructure for the third level exists today, but usage is concentrated at the first and second levels. The data below shows why.

---

Why are we talking about it now? What do the numbers say?

Three sets of data suggest that agentic shopping is more than a fad.

1. Consumers have started using AI to research purchases.

• In McKinsey's February 2026 survey of around 4,000 US consumers, 68% said they had used AI tools in the previous three months, mostly to support decision-making (Practical Ecommerce).
• In Europe, 63% of respondents say they use AI-powered tools to compare options and learn about products (McKinsey).
• According to Morgan Stanley's AlphaWise survey, around 23% of Americans had shopped with AI assistance at least once in the previous month (Digital Commerce 360, December 2025).

2. Visitors arriving from AI are more valuable. According to Adobe Analytics' 2025 holiday season data, based on more than 1 trillion visits to US retail websites (Digital Commerce 360, January 2026):

• Traffic from generative AI tools to retail websites grew 693% year over year.
• These visitors were 31% more likely to make a purchase than those arriving from other sources. In 2024, that gap was half as large.
• Revenue per visit from AI increased 254%. These visitors were also 33% less likely to leave the site immediately.

3. Major institutions project a trillion-dollar market.

• McKinsey: Up to $1 trillion in US retail alone and $3–5 trillion globally in agent-mediated sales by 2030. It is important to note that this is a scenario-based calculation, not a definitive forecast.
• Morgan Stanley: Projects that 10–20% of US online spending, or up to $385 billion, could go through agents by 2030.

---

The infrastructure is in place: What have payment giants and tech companies done?

The idea that “one day AI will shop for us” is not new. What made 2025 different was the series of announcements of the secure payment infrastructure needed to make it happen:

In other words, saying “add this to my cart and pay with my card” is now technically possible. So why isn't everyone shopping this way yet?

---

The lesson of 2026: Discovery moved to AI, purchasing stayed on the website

Most articles about agentic shopping skip this part.

On 24 March 2026, OpenAI withdrew direct purchasing within ChatGPT (Instant Checkout) as a standalone feature. The company explained that the initial version had not offered the flexibility it wanted. Merchants now use their own checkout pages, while ChatGPT focuses on product discovery (TechCrunch, 24 March 2026). According to the report, users were not using ChatGPT much to actually make purchases.

Consumer data points in the same direction. According to McKinsey's European research, more than half of shoppers are happy for AI to suggest options, but want to make the final decision themselves (McKinsey). McKinsey sums up this finding as “AI to browse, but not to buy.”

This does not mean agentic shopping has failed. What is actually happening is this:

1. The beginning of the customer journey has moved to AI. Assistants such as ChatGPT, Gemini, and Copilot now handle research, comparisons, and shortlisting.
2. The end of the journey—trust, the final question, and payment—is still on the brand's website. When customers arrive on your site from AI, they still have questions: “Will these boots fit me?”, “When will my order arrive?”, “Will this fabric make me sweat?”
3. The infrastructure for autonomous purchasing is ready, but earning consumer trust will take time. An HBR article by PwC authors also emphasizes that this is still an “early-stage” period (HBR, February 2026).

In short, the “add this to my cart” experience is coming. But it seems more likely to become widespread first on the websites of stores customers already trust, rather than on external platforms.

---

What it means for brands: How do you persuade an AI agent?

An HBR article by Bain authors describes how, in some cases, agents complete the entire journey from research to payment without a human click. It cites Amazon's “Buy for Me” feature as an example (HBR, October 2025). When an agent takes over, some marketing designed for humans loses its impact.

Research published in HBR in May 2026 demonstrates this experimentally (HBR, May 2026):

• Persuasion tactics designed around human psychology do not reliably work on AI agents and sometimes even backfire. Examples include “only 3 left,” countdown timers, and crossed-out prices.
• The only signal that consistently increased selection by agents was the star rating. Higher prices, meanwhile, consistently reduced selection.
• The authors recommend treating each AI model as a separate customer segment and focusing on competitive pricing and genuine customer reviews.

This points to two rules for brands:

1. Your product information must be clear enough for a machine to read. Dimensions, materials, compatibility, and use cases should be stated explicitly. Agents make decisions based on data, not intuition.
2. Your trust signals must be genuine. Real reviews, clear return and shipping policies, consistent pricing. Artificial urgency messages do not influence agents.

---

The picture in Türkiye: A TRY 4.57 trillion market and its opportunity

Türkiye's e-commerce volume reached TRY 4.57 trillion in 2025, up 52.2% year over year. E-commerce accounts for around 19% of total commerce, with 634,000 businesses selling online (Republic of Türkiye Ministry of Trade, May 2026).

Much of the infrastructure described above is currently US-focused. But customer behavior knows no borders: Turkish consumers also ask ChatGPT, Gemini, and Copilot “which product should I buy?” This has two implications for Turkish e-commerce websites:

• Becoming one of the brands AI assistants recommend is a new competitive arena. Clear product content in Turkish is still scarce in this area.
• For Turkish brands selling abroad, AI agents offer a new shop window. But when international customers arrive on your website, they need to be able to ask questions and get answers in their own language.

---

5 steps to prepare your store for agentic shopping

1. Make your product pages readable by AI. Clearly state each product's dimensions, materials, compatibility, intended audience, and use cases. Use measurable information rather than empty phrases such as “great quality.”

2. Use structured data (schema). Product, Offer, AggregateRating, and FAQPage markup help both search engines and AI agents understand your product correctly.

3. Collect genuine reviews. Research shows that star ratings have the most consistent influence on agents' choices. Automate your post-purchase review request process.

4. Keep prices, stock levels, and policy information consistent everywhere. Shipping times, return conditions, and prices should match across your website, marketplaces, and product feeds. Agents interpret inconsistencies as a trust issue.

5. Have an assistant on your website that can talk to customers. Customers who research with AI and then arrive on your site are already used to shopping through conversation. On your site, too, they should be able to ask questions, get product recommendations, and add a recommended product to their cart from within the conversation. The place where you can offer the “add this to my cart” experience today is the AI assistant on your own website.

---

Where does RoMuAI fit in?

At RoMuAI, we develop AI sales assistants that handle the part of agentic shopping that takes place on the store's own website:

• RoMuAI – AI Sales Chatbot for Shopify stores: Talks to visitors in 9 languages, recommends products from the live catalog, displays product cards within the conversation, and lets customers add items to their cart without leaving the chat. It also tracks orders.
• AI Sales Chatbot for Ticimax and ikas stores: Answers product and order questions using the store's own information, in 9 languages, 24/7.
• Trendyol and Hepsiburada Q&A Bot: Answers marketplace product questions using your product information, in an average of 4 minutes.
• RoMuAI Lead for websites: Talks to visitors on B2B websites, learns about their needs, and hands them over to your sales team with a meeting booked.

Your customers have started shopping with AI. Your website needs to speak the same language.

👉 Request a free demo

---

Frequently Asked Questions

What is the difference between agentic shopping and an AI chatbot?

A traditional chatbot usually answers questions with predefined responses. An AI agent in agentic shopping, by contrast, takes action: it searches for products, compares them, prepares the cart, and makes purchases if authorized. Modern sales assistants sit between these two worlds: they answer questions, recommend products, and perform actions such as adding items to the cart on the customer's behalf. We explain the difference between rule-based and AI chatbots in detail in this article.

Can you order directly through ChatGPT?

OpenAI launched purchasing within ChatGPT (Instant Checkout) in September 2025, then withdrew it as a standalone feature in March 2026. Today, products are discovered in ChatGPT, while purchases are mostly completed on the brand's own website or in brand apps within ChatGPT.

Can AI agents pay on my behalf, and is it safe?

Visa, Mastercard, and Google have developed dedicated infrastructure so that agents can only make payments within limits set by the user. For example, Google's AP2 protocol cryptographically signs the user's instruction and approved cart. The infrastructure is ready, but consumer trust is still developing ahead of widespread adoption.

Is agentic commerce being used in Türkiye?

Most of the payment infrastructure is currently US-focused. However, Turkish consumers use assistants such as ChatGPT, Gemini, and Copilot to research purchases. The most urgent step for Turkish brands is therefore to make product information readable by AI and enable conversational selling on their own websites.

How can a small e-commerce website prepare for agentic shopping?

Write clear, measurable product descriptions, add structured data, collect genuine reviews, keep pricing and policy information consistent across channels, and add an AI assistant that can recommend products and add them to the cart. None of these steps requires a large budget.

How large will the agentic shopping market become?

Estimates vary by institution. McKinsey projects $3–5 trillion in agent-mediated sales globally by 2030. Morgan Stanley estimates that 10–20% of US online spending could go through agents in 2030. The difference stems from different definitions of “agentic.”

---

Sources

1. Gaarlandt, J., Korver, W., Furr, N., Shipilov, A. — “AI Agents Are Changing How People Shop. Here's What That Means for Brands.” Harvard Business Review, 26 February 2025. https://hbr.org/2025/02/ai-agents-are-changing-how-people-shop-heres-what-that-means-for-brands
2. Vu, M., Burns, M., Cheris, A. — “What Should Retailers Do About AI Shoppers?” Harvard Business Review, 27 October 2025. https://hbr.org/2025/10/what-should-retailers-do-about-ai-shoppers
3. Furman, A., Gürdeniz, E., Safari, R., Ural, R. — “How Brands Can Adapt When AI Agents Do the Shopping.” Harvard Business Review, 19 February 2026. https://hbr.org/2026/02/how-brands-can-adapt-when-ai-agents-do-the-shopping
4. Sabbah, J., Acar, O. A. — “Research: Traditional Marketing Doesn't Work on AI Shopping Agents.” Harvard Business Review, 12 May 2026. https://hbr.org/2026/05/research-traditional-marketing-doesnt-work-on-ai-shopping-agents
5. McKinsey & Company — “The agentic commerce opportunity” (October 2025), as reported by Digital Commerce 360, 20 October 2025. https://www.digitalcommerce360.com/2025/10/20/mckinsey-forecast-5-trillion-agentic-commerce-sales-2030/
6. McKinsey & Company — “AI to browse but not to buy.” https://www.mckinsey.com/featured-insights/charts/ai-to-browse-but-not-to-buy
7. McKinsey US consumer survey (February 2026), as reported by Practical Ecommerce. https://www.practicalecommerce.com/shoppers-want-ai-help-not-control
8. Adobe Analytics 2025 holiday season data, as reported by Digital Commerce 360, 13 January 2026. https://www.digitalcommerce360.com/2026/01/13/generative-ai-online-holiday-shopping-traffic-2025/
9. Morgan Stanley AlphaWise, as reported by Digital Commerce 360, 9 December 2025. https://www.digitalcommerce360.com/2025/12/09/morgan-stanley-ai-agentic-shoppers-385-billion-online-sales/
10. Mastercard — “Mastercard unveils Agent Pay”, 29 April 2025. https://www.mastercard.com/cz/cs/news-and-trends/press/2025/april/mastercard-unveils-agent-pay-pioneering-agentic-payments-technology-to-power-commerce-in-the-age-of-ai.html
11. Visa — “Find and Buy with AI: Visa Unveils New Era of Commerce”, 30 April 2025. https://usa.visa.com/about-visa/newsroom/press-releases.release.21361.html
12. Google Cloud — “Announcing Agent Payments Protocol (AP2)”, 16 September 2025. https://cloud.google.com/blog/products/ai-machine-learning/announcing-agents-to-payments-ap2-protocol
13. Stripe — “Stripe powers Instant Checkout in ChatGPT and releases Agentic Commerce Protocol codeveloped with OpenAI”, 29 September 2025. https://stripe.com/newsroom/news/stripe-openai-instant-checkout
14. Shopify — “Winter '26 Edition: Agentic Storefronts”, 10 December 2025. https://www.shopify.com/news/winter-26-edition-agentic-storefronts
15. TechCrunch — “OpenAI's plans to make ChatGPT more like Amazon aren't going so well”, 24 March 2026. https://techcrunch.com/2026/03/24/openais-plans-to-make-chatgpt-more-like-amazon-arent-going-so-well
16. Republic of Türkiye Ministry of Trade — “Türkiye's e-commerce volume reached TRY 4.6 trillion in 2025”, May 2026. https://ticaret.gov.tr/haberler/turkiyede-e-ticaret-hacmi-2025te-4-6-trilyon-liraya-ulasti

---

Mustafa Kahraman is the founder of RoMuAI, which develops AI sales assistants for e-commerce websites and B2B companies.