Direct answer: Agentic shopping (also called agentic commerce) is online buying in which an AI agent does part or all of the work for you: it searches, compares, builds a cart and, if you allow it, pays. As of September 29, 2026, no single company leads. OpenAI, Google, Amazon and Meta compete to be the agent you talk to; Visa, Mastercard, Stripe and Google are building the payment rails that let agents pay safely. The main features announced so far launched in the United States first, and none of the announcements we reviewed names the Philippines.
Last materially reviewed: September 29, 2026
Evidence and action
- Verified: Launch details below come from the companies’ own announcements (OpenAI, Google, Amazon, Meta, Visa, Mastercard) and are dated in each section.
- Qualification: Market-size and job figures are forecasts or employer surveys, not measured results. Product availability changes quickly.
- Philippine law: No Philippine law is written specifically for AI shopping agents, but the Electronic Commerce Act, the Internet Transactions Act and the Data Privacy Act already apply to many of the transactions an agent would make.
- Action: If you try a shopping agent, keep purchase approval on, set a spending limit and use a card or wallet with a low limit or one-time card number.
Key takeaways
- What it is: an AI agent that acts on your instructions across the shopping journey, from discovery to checkout.
- Who leads: leadership is split by layer. Google has the broadest stack (search, Gemini, payments protocol, commerce protocol); Amazon has the largest captive shopping audience; OpenAI and Meta have huge chat audiences; Visa and Mastercard control card-payment trust.
- Jobs: the clearest pressure is on routine retail and clerical tasks. Global employer surveys still expect more roles created than lost overall, but the new roles need different skills.
- Main risks: wrong or unauthorized purchases, manipulation of the agent (prompt injection), heavy data collection and unclear responsibility when the agent errs.
- Here to stay? AI-assisted product discovery is already mainstream. Fully hands-off buying is advancing more slowly; OpenAI scaled back its in-chat checkout in March 2026.
Jump to: What it is · Who leads · Benefits · Jobs · Downsides · Philippine law · Here to stay? · FAQ
Decision snapshot
| Question | Practical answer |
|---|---|
| Can an AI agent legally buy something for me? | Generally yes, if you set it up to act for you. Philippine e-commerce law treats messages sent by a system programmed by you or on your behalf as yours. |
| Is it available in the Philippines now? | The flagship launches we reviewed (ChatGPT checkout, Google Universal Cart, Meta Muse) started in the US. Mastercard’s ASEAN rollout named Singapore and Malaysia, not the Philippines. |
| Will it replace retail and support jobs? | It will automate tasks within many jobs. Forecasts point to shifts in work, not a clean one-for-one replacement. |
| Should I let an agent pay without asking me? | Not yet for most people. Keep approval on for every purchase. |
| Who do I complain to if an agent buys the wrong item? | Start with the merchant and platform, then your bank or e-wallet for payment issues; DTI handles online consumer complaints. |
What is agentic shopping?
Agentic shopping is online commerce in which an AI agent takes actions on a shopper’s behalf instead of only answering questions. A regular chatbot recommends a pair of running shoes. A shopping agent can compare prices across stores, check your size and delivery address, put the item in a cart, wait for a price drop and complete the payment within limits you set.
McKinsey, in an October 17, 2025 report, describes agentic commerce as shopping powered by AI agents that can anticipate, personalize and automate steps of the process. It identifies three patterns: an agent dealing directly with a merchant’s site, agents negotiating with other agents, and brokered set-ups where a platform sits between many agents and merchants. Source: McKinsey
The levels of agent autonomy
| Level | What the agent does | Example |
|---|---|---|
| Assist | Searches, compares and summarizes; you buy | Asking an AI which air fryer has the best reviews |
| Prepare | Builds the cart and fills checkout; you approve | An agent that stops at the payment screen for your confirmation |
| Delegate | Buys automatically when your conditions are met | “Buy this when it drops below a set price” |
| Negotiate | Deals with another business’s agent | Agent-to-agent booking or price negotiation |
Who leads the agentic shopping market?
No company leads every part of agentic shopping. The market has two layers: the agents consumers talk to, and the rails that let those agents pay and prove the shopper approved the purchase. This is an editorial assessment based on the announcements below, not a market-share ranking; reliable market-share data does not yet exist.
| Company | What it launched | Status and reach |
|---|---|---|
| Agent Payments Protocol (AP2), Universal Commerce Protocol (UCP), Universal Cart in Search and Gemini | AP2 announced Sept 16, 2025 with 60+ partners; UCP announced Jan 11, 2026 with Shopify, Etsy, Wayfair, Target and Walmart; Universal Cart rolling out in the US | |
| Amazon | Rufus, renamed Alexa for Shopping (May 13, 2026); Buy for Me; automatic buy on price drop | Amazon says 250 million customers used Rufus in the reported year |
| OpenAI | Instant Checkout in ChatGPT and the Agentic Commerce Protocol, built with Stripe | Launched Sept 29, 2025 for US users; in March 2026 OpenAI shifted toward merchant-controlled checkout |
| Meta | Muse personal AI agent, which can shop and prepare a purchase for approval | Launched Sept 8, 2026 in the US; pays through Link by Stripe one-time card numbers |
| Visa | Visa Intelligent Commerce and Trusted Agent Protocol | Agent verification, spending controls and consumer consent; being deployed |
| Mastercard | Agent Pay with agentic tokens | ASEAN rollout announced April 6, 2026 in Singapore and Malaysia |
Google: the broadest stack
Google is the only company that controls a major search engine, an AI assistant, a payment wallet and two open protocols at once. Its Agent Payments Protocol uses cryptographically signed “mandates”: an intent mandate records what you asked for (for example, a price limit), and a cart mandate records the exact items and price you approved. Google says this creates an audit trail showing whether the shopper authorized a purchase. Its Universal Commerce Protocol standardizes how agents talk to merchants. At Google I/O 2026 it announced Universal Cart, a cart that spans multiple merchants, rolling out in the US first, with Canada and Australia next.
Amazon: the largest shopping audience
Amazon’s advantage is that its shoppers are already on Amazon. It says Rufus, now Alexa for Shopping, was used by more than 250 million customers, that Buy for Me lets its agent buy from brand websites outside Amazon (over half a million products), and that shoppers can tell it to buy an item automatically when the price drops. Source: Amazon See also our report on Amazon’s agentic seller assistant.
OpenAI: first mover, then a retreat on checkout
OpenAI launched Instant Checkout on September 29, 2025, letting US users buy from Etsy sellers inside ChatGPT, with Shopify merchants to follow. Shoppers had to confirm each step. In March 2026 OpenAI moved away from completing purchases inside the chat and toward checkout run by merchants on their own sites, apps or merchant-built ChatGPT experiences. Source: Checkout.com analysis The lesson: getting people to discover products through AI proved easier than getting them to pay there.
Meta: Muse brings agents to Facebook-scale audiences
Meta’s Muse, launched in the US on September 8, 2026, can browse, fill in forms and complete purchases. Meta says Muse asks before sensitive actions such as buying, keeps an audit trail, and pays through Link by Stripe, which generates a one-time card number so your real card is not shared with the merchant. Muse is free for basic use, with paid tiers reported at US$20 and US$100 a month. Because Muse draws on activity across Facebook and Instagram, it raises the sharpest privacy questions of any agent on this list. For Muse on glasses, the Muse Charm device and what Meta said about the Philippines, read our Meta Connect 2026 and Muse report.
Visa and Mastercard: the trust rails
Card networks do not need to win the chat window; they need every agent to pay through them. Visa Intelligent Commerce offers agent verification, spending controls and consent tools, and its Trusted Agent Protocol is designed to tell legitimate shopping agents apart from malicious bots. Mastercard Agent Pay combines tokenized card credentials with “verifiable intent” so each AI-initiated payment can be matched to the consumer’s authorization. Its April 2026 ASEAN announcement named Singapore and Malaysia, with additional markets to follow.
What will agentic shopping do for people?
The main benefit of agentic shopping is time: it can take over repetitive comparison, re-ordering and price-watching. Potential gains include:
- Less comparison work. An agent can check specifications, reviews and delivery fees across many stores in seconds.
- Price discipline. Delegated rules such as “buy only below this price” can reduce impulse buying, if the shopper sets them.
- Accessibility. Voice-driven agents may help older shoppers and people with disabilities who find checkout pages hard to use.
- Routine purchases on autopilot. Groceries, household refills and subscriptions are the most natural early use.
- Better protection, in theory. One-time card numbers and signed purchase records can make fraud harder and disputes easier to prove than today’s saved-card checkout.
For merchants and MSMEs, the upside is new sales channels. The downside is that products must be readable by machines, not just attractive to people: structured product data, accurate stock and clear return terms start to matter as much as photos.
McKinsey projects agentic commerce could orchestrate US$900 billion to US$1 trillion in US business-to-consumer retail revenue and US$3 trillion to US$5 trillion globally by 2030. These are forecasts, not measured sales. Source: McKinsey
Will agentic shopping take away jobs?
Agentic shopping is likely to remove some routine retail and support tasks, but the evidence so far points to job change more than simple job loss. The World Economic Forum’s Future of Jobs Report 2025, based on more than 1,000 employers, expects about 170 million jobs created and 92 million displaced worldwide by 2030, a net gain of 78 million. It lists cashiers and administrative assistants among the fastest-declining roles, and 41% of employers surveyed plan to reduce staff where AI automates tasks. That survey covers AI and other forces broadly, not agentic shopping alone.
| Work most exposed | Work likely to grow |
|---|---|
| Basic product-question support and order-status chat | Handling escalations, disputes and refunds that agents cannot resolve |
| Manual listing, price-checking and data entry | Product-data management and marketplace operations |
| Routine checkout and cashier tasks | Fraud, risk and payment-security roles |
| Simple comparison-shopping content | AI supervision, testing and quality assurance |
For the Philippines, the most exposed group is e-commerce customer support, much of it handled by contact centers and BPO teams. Our guides on the future of Philippine BPO jobs with AI and IBPAP’s role in AI readiness cover the skills that are moving up in value.
Will agentic shopping benefit humanity?
It can, but only if shoppers keep control over what the agent buys and what data it uses. The benefits (saved time, accessibility, stronger payment security) are real. So is the risk that a handful of platforms decide which products shoppers ever see. When an agent chooses for you, the question “why did it pick this brand?” matters more than it did with a search results page, and ranking rules or paid placement may not be visible. Consumer trust is still low: a Vogue Business survey reported by Fortune found 72% of respondents unwilling to share card details with AI and only 31% willing to let an agent shop for them. General-purpose agents now build that choice in: OpenAI’s always-on dots agents let users decide whether the agent acts without asking or asks first.
What are the downsides and risks of agentic shopping?
| Risk | What can go wrong | How to reduce it |
|---|---|---|
| Wrong or unwanted purchase | Agent misreads size, quantity, seller or price | Keep approval on for every purchase; review the final cart |
| Prompt injection | Hidden text on a website or product listing instructs the agent to do something you did not ask | Use agents with purchase confirmation and spending caps; limit the accounts it can access |
| Over-sharing personal data | Agent discloses your address, phone or history to the wrong party | Connect only the accounts it needs; check what it can send without asking |
| Fraudulent sellers and fake agents | Scam stores targeting agents, or apps posing as shopping agents | Use agents from known providers; watch for unexpected OTP or payment requests |
| Unclear responsibility | Merchant, platform and AI provider each blame the others | Keep screenshots, order records and the agent’s activity log |
| Hidden steering | Agent favors partners or sponsored products | Ask the agent why it chose an item; compare manually for large purchases |
Prompt injection is the security risk most specific to shopping agents. The Open Worldwide Application Security Project (OWASP) ranks prompt injection as the top risk for large-language-model applications and describes “indirect” injection, where instructions hidden in websites or files change the model’s behavior. OWASP recommends giving AI systems only the access they need and requiring human approval for privileged actions, which for shopping means payments. Source: OWASP Our guide to AI security risks explains this in more detail.
Real-world failures are already surfacing. In September 2026, a Muse user publicly reported that the agent accepted a low offer on a Facebook Marketplace listing and shared his home address with a buyer. A Meta executive said the agent had asked for permission and that the company was looking into the case. Whatever the final finding, it shows why permission settings and activity logs matter. (See our Muse privacy controls table.)
What Philippine laws apply to agentic shopping?
No Philippine statute specifically regulates AI shopping agents as of September 29, 2026, but existing e-commerce, consumer, privacy and payment laws already cover much of what they do. How they apply to specific agent disputes has not yet been tested in Philippine courts.
Can an AI agent make a binding purchase for me?
Very likely yes, if you set it up to act for you. Section 16 of the Electronic Commerce Act (Republic Act No. 8792) says a contract is not denied validity only because it was formed through electronic data messages. Section 18(2)(b) goes further: as between the sender and the recipient, a message is deemed to be the sender’s if it was sent “by an information system programmed by, or on behalf of the originator to operate automatically.” In plain terms, an order your agent places on your instructions will generally be treated as your order. Section 18(5) gives some room where the recipient knew or should have known the message contained an error. Read more in Are online contracts legally binding?
Who protects me if the item is wrong or defective?
The Internet Transactions Act of 2023 (Republic Act No. 11967) covers business-to-consumer online transactions where a party is in the Philippines or the platform has minimum contacts here. It created the Department of Trade and Industry’s E-Commerce Bureau, lets consumers seek repair, replacement or refund under the Consumer Act (Section 20), and can make e-marketplaces and digital platforms subsidiarily or solidarily liable in specific situations (Sections 26 and 27). Consumer-to-consumer sales, such as many Marketplace deals between individuals, are outside its scope. See our Internet Transactions Act guide.
What about my personal data?
A shopping agent handles names, addresses, purchase history and sometimes payment data. When a business processes that information, the Data Privacy Act of 2012 (Republic Act No. 10173) applies, and the National Privacy Commission (NPC), the regulator that administers the Act, has issued NPC Advisory No. 2024-04 on AI systems that process personal data.
What if the agent makes a payment I did not approve?
Report it immediately to your bank or e-wallet and keep evidence. Banks and e-wallets supervised by the Bangko Sentral ng Pilipinas (BSP) must handle consumer complaints, and you can escalate unresolved complaints to the BSP. Whether a payment made by an agent you set up counts as “unauthorized” is a new question; your approval settings and the agent’s activity log will likely be central evidence. If the purchase was part of a scam, see online estafa in the Philippines.
What should Filipino shoppers and businesses do now?
Shoppers:
- Keep purchase confirmation on for every transaction.
- Set a spending limit, or use a virtual or one-time card number or a low-balance wallet.
- Connect only the accounts the agent needs.
- Check the final cart: seller, item, quantity, total and delivery address.
- Save the agent’s activity log, order confirmation and chat history for any disputed purchase.
Merchants and MSMEs:
- Keep product data accurate and structured (price, stock, size, delivery fees, return terms).
- Comply with the seller disclosure rules under the Internet Transactions Act; see online seller disclosure requirements.
- Review how your checkout handles orders placed by bots and agents, including fraud checks.
- Update your privacy notice if you use AI agents to serve customers.
Is agentic shopping here to stay?
AI-assisted shopping is here to stay; fully autonomous buying is arriving more slowly. The signs of permanence are strong: Google, Amazon, Meta, OpenAI, Visa, Mastercard and Stripe have all shipped products or protocols, and the payment networks are building authentication specifically for agents. The signs of caution are also clear: OpenAI scaled back in-chat checkout, consumer trust in AI with money is low, and most flagship launches are limited to the US. The likely near-term pattern is AI that finds, compares and prepares, with a human tapping “approve.”
Frequently asked questions
What is agentic shopping in simple terms?
It is shopping where an AI agent acts for you, searching, comparing and sometimes buying, instead of only giving suggestions.
Is agentic commerce the same as agentic shopping?
Largely yes. “Agentic commerce” is the industry term and also covers merchant-side and business-to-business uses; “agentic shopping” usually refers to the consumer side.
Can I use agentic shopping in the Philippines?
AI shopping assistants that recommend products are widely available. The checkout-capable features announced by OpenAI, Google and Meta launched in the US first, and none of those announcements names the Philippines. Check each provider’s availability page before relying on it.
Is it safe to give an AI agent my card?
It is safer with one-time or tokenized card numbers, spending limits and purchase confirmation. It is not risk-free.
Who is liable if an AI agent buys the wrong thing?
It depends on your settings, the merchant’s terms, the platform’s role and the payment provider’s rules. Under RA 8792, an order sent by a system you set up is generally treated as yours, so keep approval on.
Will AI shopping agents replace online sellers?
No. Agents still buy from merchants. They change how shoppers find sellers, which makes accurate product data and good reviews more important.
Which company is winning agentic shopping?
No one yet. Google has the broadest set of tools, Amazon the largest shopping base, and Visa and Mastercard the payment trust layer.
Official and primary sources
- Republic Act No. 8792, Electronic Commerce Act (Official Gazette)
- Republic Act No. 11967, Internet Transactions Act of 2023 (Supreme Court E-Library)
- National Privacy Commission
- OpenAI: Buy it in ChatGPT (Sept 29, 2025)
- Google Cloud: Agent Payments Protocol (Sept 16, 2025)
- Google: Universal Commerce Protocol (Jan 11, 2026)
- Google: Universal Cart and agentic checkout
- Amazon: generative and agentic AI in shopping
- Meta: Introducing Muse (Sept 8, 2026)
- Visa Intelligent Commerce
- Mastercard: ASEAN agentic payments (April 6, 2026)
- OWASP: LLM01 Prompt Injection
- World Economic Forum: Future of Jobs Report 2025
- McKinsey: The agentic commerce opportunity (Oct 17, 2025)
Related: before an agent can buy from you, it has to name you. See what the Spiralytics study found about Philippine brands in ChatGPT, Gemini and Perplexity answers.
CyberCode takeaway
Agentic shopping moves AI from advising to acting with your money. Philippine law already treats many of those actions as yours, so the practical protection is in your settings: approval on, spending capped, accounts limited and records kept.
Cybercode.ph provides general educational information about technology, cybersecurity, privacy and related legal issues. It is not a substitute for legal, cybersecurity or professional advice for a specific situation.
Featured image: Photo by Vitaly Gariev via Unsplash.