AI shopping agents are not one technology but three, and the type you encounter determines who controls it, what it can do, and how much of your data it touches. There are general discovery agents such as Google's AI Mode that research and compare products across many retailers; on-site assistants that retailers like Best Buy and Home Depot build into their own stores; and largely invisible business-to-business platforms that wire the whole system together. Consumer trust, meanwhile, is sharply uneven: a Deloitte survey cited by commercetools found only 8% of European shoppers are willing to let AI complete a checkout on their behalf.

Understanding AI Shopping Agents: A Three-Part Framework

The clearest way to make sense of the "AI shopping" landscape is to sort agents by their relationship to you and the retailer. Is this a tool you chose, a feature on a store's own website, or a backend service you never see? A price-monitoring firm, tgndata, offers a practical framing for retailers: AI shopping agents are "a new comparison layer that sees your prices before your shoppers do." For shoppers, the same layer works in reverse — it sees prices before you do.

This three-part classification is a synthesis of industry descriptions rather than an official standard, and the examples below illustrate the categories rather than exhaust them. But the framework holds up across very different sources, from e-commerce platform blogs to startup funding coverage, which is what makes it useful for recognizing what you are actually interacting with.

AI Shopping Agent Classification Model

Use this model the next time an AI tool helps you shop. Identify who runs it, and its likely capabilities and limits follow.

Three types of AI shopping agents and what each means for the shopper
Type of agentWho runs itWhat it doesWhat it means for you
General discovery agents Big tech companies (e.g., Google AI Mode) Research, compare, and recommend products across multiple retailers; Google says AI Mode has passed one billion monthly users A cross-store guide you choose to use; reach is broad but features vary by country
Retailer on-site assistants Individual retailers (e.g., Best Buy, Kohl's, Home Depot assistants; Firmly's Ask Agent) Shoppers describe what they want, get product recommendations, and can check out without leaving the chat Convenient within one store, but recommendations serve that retailer's catalog and goals
B2B infrastructure platforms Behind-the-scenes vendors (e.g., Firmly, Algolia, Insider One) Connect agents to retailer systems, power search, personalization, and predictive segmentation for purchase likelihood and churn Invisible to you, but shapes the recommendations and search results you see

The classification logic is straightforward: a tool you deliberately open is a discovery agent; a chat window on a retailer's own site is an on-site assistant; everything powering those experiences behind the curtain is infrastructure. When you know who operates the tool, you can anticipate its scope — a Google agent can compare across stores, while a retailer's assistant is built to keep you in that store.

General Discovery Agents: Your Cross-Retailer Guides

General discovery agents are the most visible category. Google's AI Mode, which according to Verity Score reached more than 40 European countries on 8 October 2025 — with France joining on 22 July 2026 after a regulatory hold — answers shopping queries directly in search. OpenAI, meanwhile, has expanded ChatGPT Ads to 31 European markets, positioned around the moments when people research and compare products, according to commercetools.

Not every general agent browses politely. GeekWire reported that Meta's Muse assistant can browse retail sites and buy items with a user's approval — and that Amazon has blocked Muse from its site and is suing Perplexity over a shopping agent in the startup's Comet browser. That friction matters for consumers: an agent that clicks through web pages like a human shopper can be welcomed or shut out retailer by retailer, which affects how reliably it can comparison-shop for you.

Retailer On-Site Assistants: Personalized Store Experiences

The second category lives inside a single store's website or app. Retailer-owned assistants are, per the commercetools data, "the other trend to watch": Best Buy and Kohl's have launched their own assistants, and Home Depot has rolled its assistant out to all its US stores. These tools meet you where you already are, describe products conversationally, and recommend items from that retailer's catalog.

Seattle startup Firmly, backed by JPMorgan and Mastercard, launched one such tool called Ask Agent for retailers' own websites, where shoppers describe what they want, get recommendations, and can check out without leaving the chat. The trade-off is structural: an assistant built by a retailer is optimized for that retailer's engagement and sales, so its "personalization" is personalization within one storefront, not across the market.

B2B Infrastructure Platforms: The Power Behind the Scenes

The third category is the one you never see, yet it increasingly determines what you're shown. Firmly's approach, as GeekWire described it, is different from browsing agents: instead of having software click through a retailer's web pages, it connects directly to the retailer's systems with the retailer's permission. That permission-based plumbing is what lets agents check out, verify inventory, and transact reliably.

Other platforms supply the intelligence layer. Algolia provides scalable, API-led search and product discovery for large catalogs, though Kore.ai's market guide notes it is best suited to search and discovery rather than end-to-end journey orchestration. Insider One's Agent One suite includes a Shopping Agent, a Support Agent, and Sirius AI, which builds predictive segments for purchase likelihood, churn, and discount affinity. Insider One describes a feedback loop in which every conversation enriches a customer's profile, making the agent better at its job over time — a useful reminder that on-site personalization is built on accumulated data about you.

Consumer Trust and Delegation: What Shoppers Allow AI To Do

Shoppers draw a clear line between research and purchasing. According to a Deloitte survey of 13,500 consumers across 15 European countries in August 2026, cited by commercetools, only 8% of consumers are willing to let AI check out on their behalf. The industry has noticed: ChatGPT's Instant Checkout was retired in March 2026, per Fast Company and Eco, and Google's agentic checkout and Universal Cart remain US-first with no published timeline for continental Europe, according to Lengow.

BigCommerce's guidance to merchants explains the gap from the other side: current checkout flows assume human users clicking buttons, and future systems need secure payment methods that can handle automated transactions. In other words, both the technology and the trust are still being built. Discovery and comparison are where agents and shoppers currently meet comfortably; the final "buy" button remains mostly human.

Data Privacy in AI Shopping: User Controls and Concerns

Personalization runs on data, which makes privacy the natural second question. Google's Safety Center states that Google Assistant is "built to keep your information private, safe, and secure," with "simple privacy controls to help you choose what's right for you." The company says the Assistant waits in standby mode and won't send what you're saying to Google until it detects an activation like "Hey Google," and it directs users to settings covering audio recordings, Web & App Activity, and ads personalization.

These are general controls for Google services, and specific practices for any individual shopping feature may differ; Google's privacy policy also points to service-specific notices and archived policy versions for review. The honest summary for consumers: the infrastructure-level platforms like Insider One openly describe profiles enriched by every conversation, while the general assistants offer settings you can adjust. Whether those settings match your comfort level is worth checking before you lean on an agent for anything sensitive.

Navigating the AI Shopping Landscape

Consumers should evaluate AI shopping agents based on their operational scope — general discovery, retailer-specific, or backend-powered — and their comfort level with delegating tasks, particularly regarding automated purchasing. The evidence suggests a sensible posture: comparison and research are where agents already add value and where shoppers are most comfortable, while checkout remains a step most people, by a wide margin, still want to take themselves.

The measurable indicator of where you stand is simple: your decision to use an AI shopping agent for price comparison versus avoiding it for automated checkout. If you find yourself happily letting an agent sort options across stores but still clicking "buy" yourself, you are squarely in the mainstream of consumer behavior as measured in August 2026 — and the retailer assistants and backend platforms competing for your attention will keep working to move that line.

Sources