How AI Agents Are Changing Online Shopping and Business Sales

AI-powered assistants are beginning to shop, compare products and make purchases on behalf of consumers, creating new demands for businesses. Research highlighted by Whitepage shows how automated traffic, AI visibility, cybersecurity and payment authentication are reshaping digital commerce.

5 November 2026 – Businesses are facing a new challenge as artificial intelligence agents begin acting on behalf of customers, searching for products, comparing prices and carrying out online transactions without requiring people to complete every step themselves.

An analysis by strategic presentation and narrative agency Whitepage identifies five problems companies must address as AI agents become more involved in purchasing decisions and customer interactions.

According to Cloudflare's September 2026 report, more than half of web traffic now comes from non-human sources, while daily requests from AI agents have increased by more than 1,700% over the previous year.

The figures suggest that businesses can no longer assume everyone interacting with their websites is a human visitor. Some automated activity represents legitimate customers using AI assistants to complete tasks, while other traffic may come from malicious bots.

Whitepage argues that companies need to reconsider how their websites communicate information, verify transactions and distinguish authorised AI activity from security threats.

AI Agents Are Becoming a New Audience for Businesses

Most business websites are designed around human visitors who read product descriptions, compare photographs and make purchasing decisions.

AI agents interact with the same information differently, often interpreting structured content, product labels and page layouts before making recommendations.

Research cited by Whitepage from Columbia Business School and Yale examined how changes to a simulated shopping page affected AI product selection.

Moving a product to the top row made one AI model approximately five times more likely to select it. Adding a "Sponsored" label reduced its selection rate from 10% to between 7.9% and 8.9%, while an "Overall Pick" badge increased selection rates to between 19.9% and 42.6%.

These findings suggest that product presentation can influence automated recommendations, although the results come from controlled experiments rather than measurements across all commercial AI assistants.

Adobe's analysis of more than one trillion visits to US retail websites also found that AI-referred traffic generated 83% more revenue per visit than other traffic, according to the figures cited in Whitepage's report.

The agency also reports that the average retail homepage achieves an AI readability score of 63%, indicating that some website information may be difficult for AI systems to interpret.

Tanya Slyvkin, CEO and Co-Founder of Whitepage Studio, said:

Brands now have two audiences for the same page: the person and the AI agent shopping for them.

Slyvkin recommends regularly testing how major AI assistants describe and compare products, since their responses can change following model updates.

Businesses Must Distinguish Legitimate AI Agents From Malicious Bots

For years, companies have relied on automated traffic detection systems to identify suspicious website activity and restrict potentially harmful access.

The arrival of consumer-facing AI agents complicates that approach because legitimate customers may now use automated tools to browse stores, compare products or make bookings.

Research from HUMAN Security, cited by Whitepage, found a difference of just half a percentage point between the rates of harmless and malicious automation.

The agency argues that blocking all automated traffic could prevent genuine customers from using AI assistants to interact with businesses.

The issue has also reached the courts. Whitepage references an August 2026 US federal appeals court decision concerning Perplexity's AI shopping agent and Amazon, which it describes as recognising customer-directed agent activity in the context of website access.

The legal and technical questions surrounding automated access remain complex, and individual court decisions should not be interpreted as granting unrestricted access to every website.

Whitepage recommends defining which activities AI agents may perform without additional verification. Product browsing and price comparisons could remain accessible, while account access and payments would require stronger authentication.

AI Agents Can Be Manipulated Through Website Content

Allowing autonomous systems to read information from websites introduces another security concern.

AI agents can encounter misleading content or instructions deliberately inserted into material they are processing, a technique commonly described as prompt injection.

Whitepage cites research from Nanyang Technological University, IBM Research and the University of Illinois examining attacks against AI-powered web agents.

In the experiments, hidden instructions placed within online content reportedly succeeded in redirecting agent behaviour in 41.7% to 68.2% of tested attacks.

Separate experiments involving manipulated product images increased the selection rate of a targeted product from 10% to 76.7%.

These results concern specific experimental conditions and do not establish that all commercial AI agents experience the same vulnerability rates.

For businesses, the concern is that manipulated product listings, reviews or other content could influence an agent's recommendations or actions.

Whitepage advises companies to monitor information presented through their own websites and investigate unusual automated activity, including unexpected changes to orders or payment details.

AI Payments Raise Questions About Customer Consent

Traditional online checkout systems are designed around a customer actively reviewing and confirming a purchase.

AI agents introduce additional complexity because they may be authorised to perform certain tasks without requiring approval at every stage.

A customer could instruct an assistant to compare products within a budget, but that instruction does not automatically establish permission to make unlimited purchases or change account information.

Whitepage references concerns raised by Federal Reserve Governor Christopher Waller about fraud detection systems originally designed around human behaviour.

The agency also cites the International Monetary Fund's observations that established identity verification and multifactor authentication methods are generally designed for people rather than autonomous systems.

Businesses therefore need reliable ways to determine whether a transaction was genuinely authorised by the customer.

Whitepage recommends direct confirmation for higher-risk actions, including large purchases, cancellations and changes to payment details.

Such controls could help prevent automated systems from completing transactions that exceed the customer's intended instructions.

Competing AI Payment Standards Create Integration Challenges

The fifth issue identified by Whitepage concerns the number of competing approaches to identifying AI agents and enabling automated payments.

The agency points to initiatives from Mastercard, Cloudflare, Google, OpenAI and Stripe, Visa, and Sierra and Meta, developed as technology companies respond to the growth of agent-mediated commerce.

These initiatives are not identical, and they address different aspects of agent identification, trusted interactions and payment processing.

For businesses planning to accept transactions initiated by AI assistants, supporting several approaches could increase development and integration costs.

Yet adopting one system too early may create problems if customers and payment providers move towards different technical standards.

Whitepage cites research referenced by Mastercard suggesting that agent-mediated purchases currently account for approximately 0.4% of global e-commerce transactions by volume and 0.2% by value.

The figures indicate that agent-mediated commerce remains a small share of overall transactions, despite growing interest in the technology.

Whitepage recommends that businesses begin with the standards already supported by their existing payment providers and technology platforms, while keeping their systems adaptable.

What Businesses Should Prepare for as AI Shopping Expands

The growth of AI agents creates new considerations for companies selling products and services online.

Businesses will need to examine whether automated assistants can accurately interpret their product information and whether existing security systems can distinguish authorised activity from malicious automation.

Payment authentication is another area requiring attention, since agents may increasingly perform tasks that involve customer accounts and financial transactions.

Whitepage's analysis suggests that companies should review their websites from the perspective of both human visitors and automated assistants, while maintaining appropriate controls over sensitive actions.

Although AI agents are attracting attention from retailers and payment companies, their commercial impact will depend on customer adoption, technical reliability and the development of accepted standards.

For businesses, the immediate priority is understanding how these systems interact with existing digital services and identifying where stronger verification may be needed before automated transactions become more common.

About Whitepage Studio

Whitepage Studio is a strategic presentation and narrative agency that helps organisations communicate business information through presentations and visual storytelling.

The agency's November 2026 analysis examines how AI agents are changing digital customer interactions and identifies issues involving product visibility, automated traffic, cybersecurity, payment consent and emerging transaction standards.

Official website: https://whitepage.studio/