Advances in artificial intelligence are rapidly transforming how we shop and pay online. Today’s AI shopping assistants can converse with users, answer product queries, compare options, and even complete transactions. At the same time, payment networks are developing new tools that let AI agents handle routine tasks, such as invoicing and checkout, via simple chat prompts.

Together, these developments herald an era of agentic commerce, where intelligent assistants do more than fetch information – they act on behalf of customers and merchants. PayPal now offers its users early access to Perplexity’s new AI “Comet” browser, which features an integrated AI assistant, native answer-focused search, and product comparisons. Comet serves as a personal shopper within the PayPal app.

Likewise, Visa and Mastercard are rolling out developer toolkits to let both coders and non-technical teams create AI-driven payment agents by typing simple requests (e.g., “generate an invoice” or “create a payment link”). These tools – built on Visa’s Intelligent Commerce platform or Mastercard’s new agentic payments framework – handle everything from completing transactions to summarizing revenue reports, all via conversational commands. This article explores these innovations, from front-end shopping bots to back-end payment assistants, and how they are ushering in a new age of AI-driven commerce.

How AI Shopping Assistants Are Reshaping the Buying Experience

AI shopping assistants are intelligent virtual concierges that help customers find and buy products. Unlike static search, these assistants use natural language processing and large language models to interact in real time. In other words, an AI shopper works like a personalized digital sales agent. AI shopping assistants are no longer optional – they’re essential for meeting modern consumer expectations.

They promise shoppers faster results, personalized recommendations, and on-the-spot comparison shopping at scale. A consumer might ask an assistant, “What are the best running shoes under $100?” and receive a curated list of relevant products with pros/cons. These assistants can also handle follow-up questions (“How do they differ?”) or even complete the checkout, all within the same interface.

PayPal’s recent partnership with Perplexity is a prime example. In September 2025, PayPal announced that U.S. PayPal and Venmo users could skip the waitlist for the Perplexity Comet browser by signing up for a free 12-month trial. Comet is like a personal shopper and personal assistant all in one. Instead of bouncing between apps, users can now do AI-powered product research and comparisons from within PayPal’s interface. This tight coupling means a user could, for instance, discover a deal and immediately pay for it without leaving the PayPal app.

Beyond PayPal, other tech companies are embedding AI into shopping workflows. OpenAI’s ChatGPT has added new shopping features: it now offers a “Shopping Research” mode (introduced November 2025) that builds personalized buyer’s guides based on user queries. For complex questions (“Find a quiet stick vacuum under $150 for a small apartment,” etc.), ChatGPT will ask clarifying questions and present a researched comparison of options as a conversational guide.

In late 2025, OpenAI also launched Instant Checkout within ChatGPT (in partnership with Stripe), allowing users to purchase from supported merchants directly in chat. This means you can find an item in ChatGPT’s search results and tap “Buy” to pay immediately, without leaving the conversation.

These examples highlight how conversational AI is turning e-commerce into a dialogue: users ask; the AI responds with choices and even handles checkout. The result is a more natural, guided buying experience. Compared to older rule-based bots, modern AI assistants can interpret nuanced queries, learn about personal preferences, and proactively suggest relevant deals.

Because they can pull real-time data from thousands of sources, today’s shopping assistants are much more versatile. In retail, brands report that these tools can boost customer engagement and sales: studies show that intelligent shopping assistants help increase average order value and reduce service costs by handling routine queries. With 24/7 availability, they support customers at any time without added staff costs.

Agentic Payments and AI-Driven Checkout

Behind the scenes, payment networks and fintech companies are also racing to support this new AI-driven commerce. In the “agentic commerce” model, the AI assistants described above act as agents that carry out transactions on behalf of users, but this requires a new payment infrastructure to ensure safety and efficiency. Agentic commerce is where AI agents can browse, purchase, and manage transactions on behalf of users, securely and with minimal friction. This means connecting AI agents to the payment network and merchant systems with strong security controls and easy integration.

Visa has been particularly active with this vision. In April 2025, Visa launched the Visa Intelligent Commerce initiative, a set of APIs and protocols to enable agent-led shopping. By September 2025, it announced two primary tools. First, a Model Context Protocol (MCP) Server that acts as an integration layer: AI agents can now plug into Visa’s Intelligent Commerce APIs via MCP, avoiding weeks of custom coding. Second, a new Visa Acceptance Agent Toolkit (in pilot) that lets even non-coders use plain-language prompts to trigger payment actions.

The Toolkit provides prebuilt workflows (such as invoicing and “pay by link” generation) and enables users to type commands in a simple chat interface. A merchant support agent could chat, “Create an invoice for $100 for John Doe, due Friday,” and the AI would automatically call the Visa Invoice API and generate a payment link, without any manual development.

Similarly, a business analyst could ask, “Create a summary of today’s revenue across all invoices,” and the agent would securely query the transaction data and compile the report.

Mastercard is pursuing parallel strategies. In September 2025, the company announced a suite of agentic commerce tools and standards. A significant focus is Mastercard Agent Pay, a framework working with industry partners (Stripe, Google, Ant, etc.), so that by the holiday season, all U.S. Mastercard accounts can participate in AI-enabled checkout. To help developers, Mastercard unveiled:

• Agent Toolkit: Available via the Mastercard Developers portal, this toolkit lets AI assistants consume Mastercard’s API documentation via the MCP server (as with Visa’s). An AI agent can use natural language to discover and call Mastercard payment APIs, integrating with platforms like GitHub Copilot or browser extensions.
• Agent Sign-Up: A mechanism for companies to register and identify their AI agents, so that the payment network can authenticate the agent’s identity before enabling transactions.
• Insight Tokens: Permissioned data tokens that allow AI agents to retrieve personalized or contextual information (like loyalty balances) with user consent. This builds Mastercard’s existing token and loyalty technology to deliver tailored recommendations in an authorized way.
• Agentic Consulting: Mastercard is also offering expert consulting services to help banks, merchants, and developers design compliant AI shopping solutions.

These tools aim to lower the barrier for building intelligent payment experiences. Together, Visa’s and Mastercard’s initiatives provide APIs, standards, and safety rails so that AI agents can reliably handle orders, payments, and settlements. Crucially, they emphasize security and trust: for example, Visa’s Intelligent Commerce program ties together tokenization, authentication, spending controls, and privacy safeguards into a cohesive system.

And both companies are working on protocols to recognize legitimate agents: Visa published a Trusted Agent Protocol (in partnership with Microsoft and Cloudflare) to signal agent intent and verify shopper consent. Mastercard similarly stresses that it will register and authenticate any agent before a transaction.

Key Examples and Capabilities

• Instant Invoicing & Analytics: Using the Visa Acceptance Toolkit, a non-technical user can type a request like “Create an invoice for $100 for John Doe, due Friday.” The system then calls Visa’s invoice API and returns a payment link automatically. Similarly, asking “Summarize today’s revenue” triggers an analysis of transaction data and the generation of instant reports. This means businesses can automate billing and reporting tasks with simple language.
• Plain-Language Integration: Both Visa and Mastercard emphasize “no-code” access. Visa’s toolkit requires no coding and uses plain prompts. Mastercard’s Agent Toolkit via MCP makes API documentation machine-readable so that AI tools can understand and use it. Agents can interpret requests and apply Visa APIs “in new contexts,” considerably speeding up development.
• AI-Driven Checkout: On the consumer side, tools like OpenAI’s Agentic Commerce Protocol enable ChatGPT to complete checkout on sites (via Stripe) with one click. The Agentic Commerce Protocol, which Visa and others support, defines how order details flow from the AI to merchants’ systems. Shopify, Etsy, and many payment partners are already aligned with this emerging standard. Similarly, Google introduced its own Agent Payments Protocol (AP2) for secure AI transactions in late 2025. These efforts indicate that big tech players are converging on open standards for agentic commerce.

Collectively, these back-end innovations create a new layer of commerce where AI agents and financial networks speak the same language. Developers can now focus on AI behavior, while standardized protocols handle plumbing payments.

Soon, shoppers could tell their phone or chatbot to buy a recommended product and rest assured that the network will securely route the transaction through the correct channels.

AI Shopping Assistants – Benefits, Challenges, and Outlook

The rise of AI shopping assistants and agentic payments promises several benefits. For consumers, the experience becomes far more convenient: the AI assistant handles tedious tasks such as searching across dozens of sites, tracking orders, and entering payment details. Personalization can also be improved by using AI to automatically surface deals or reward points based on a shopper’s past behavior and loyalty data (via insight tokens).

Early adopters report higher engagement and conversion rates: retailers using AI shopping agents find that customers discover products faster and complete purchases more often. For businesses, these tools can reduce support costs by automating routine inquiries (e.g., refund status or order updates) and increase sales by embedding the checkout right into conversational interfaces.

On the industry side, agentic commerce opens new opportunities. Online marketplaces and payment platforms gain fresh touchpoints with customers. Payment networks can reinforce their central role by powering these AI assistants behind the scenes. For example, by serving as the default payment rails for agentic transactions, Visa and Mastercard hope to extend their reach into AI-driven ecosystems. Fintech innovators like Stripe and Square are also building APIs (Stripe co-developed OpenAI’s protocol) so that their merchant customers can join this trend seamlessly.

However, there are challenges to address. Security and fraud prevention are paramount: networks must ensure that an AI agent truly represents an authorized shopper. That is why Visa and Mastercard emphasize verified agent identities and consent mechanisms. Privacy is another concern: because AI shopping might draw on personal data, systems must minimize data sharing and use encryption (as stipulated by the Agentic Commerce Protocol).

Regulations may eventually require new disclosures or require opt-in for AI-driven purchases. Finally, there is the practical hurdle of adoption: merchants and developers will need time to integrate these new APIs and trust the system. Both Visa and Mastercard note that their pilots will gradually scale – for instance, Mastercard plans a broad roll-out of its Agent Pay system by the 2025 holiday season.

Looking ahead, AI in commerce seems poised to grow rapidly. Industry analysts predict that agentic AI could account for many billions of dollars in online sales. Already, holiday season data has shown surges in AI-driven shopping traffic (one analysis found generative-AI referrals to retail sites jumping 4,700% by mid-2025).

As more consumers experiment with voice assistants, chatbots, and AI browsers, the volume of agent-led purchases will rise. The ecosystem is also aligning: major platforms like Shopify, Salesforce, and even Google are integrating agentic standards.

Conclusion

We are at the cusp of a new phase of digital commerce. AI shopping assistants are moving beyond novelty to become everyday tools, and payment providers are racing to back them with secure infrastructure. This is a shift where one prompt, one payment, one breakthrough at a time will redefine the shopping experience.

By blending seamless search, personalized recommendations, and conversational checkout, AI agents promise to make buying faster, easier and more intelligent. Consumers and businesses who embrace these agentic tools stand to benefit from greater convenience and efficiency, even as the industry works to ensure trust and safety in this bold new era of AI commerce.

Frequently Asked Questions