We Gave AI Shopping Assistants the Same Budget. The Results Were Very Different

Online shopping usually turns into the same routine. You search for a product, open several retailer tabs, compare specifications, read reviews, check prices, and eventually try to remember which option looked promising in the first place.

AI shopping assistants are starting to take over parts of that process. Some research products across the web and build a shortlist around your requirements. Others have direct access to large commerce catalogs, price history, stock information, or checkout systems. Browser agents add another approach by working directly with the websites where the shopping happens.

To make those differences easier to compare, we will use the same reference brief throughout the article:

Find a laptop under $1,000 for work and travel. It must have at least 16 GB of RAM and 512 GB of storage, weigh less than 3.2 pounds, and be available from a reputable retailer. Compare three options and explain the main tradeoffs.

This is a comparison of the current capabilities and workflows behind five AI shopping systems, rather than a controlled benchmark with a declared winner. Prices, stock, and product configurations change too quickly for a static recommendation to remain reliable for long.

AI Shopping Is Moving Beyond Product Search

Traditional ecommerce search starts with keywords and filters. You might search for a lightweight laptop, set a maximum price, select a memory configuration, and manually compare what remains.

AI allows the user to start with the actual goal instead.

A shopping assistant can interpret a request that combines budget, intended use, portability, specifications, and preferences. It can then search for suitable products, explain tradeoffs, refine the shortlist when requirements change, and in some cases continue into later parts of the shopping process.

Several major platforms are now building around that workflow.

OpenAI's is designed for purchases that involve multiple requirements and tradeoffs. Google's connects shopping activity across Search and Gemini while monitoring signals such as deals, price drops, stock, and compatibility. combines conversational product discovery with retailer information and price history. Amazon's works closely with Amazon's own product catalog and shopping environment.

Sigma approaches the same problem from inside the browser. Its can navigate pages, click, type, read web content, and work through multi-step website tasks. Product research is one of the agent's listed use cases.

Sigma Browser: Product Research Inside the Browser

Sigma differs from dedicated shopping interfaces because its agent works alongside the webpages the user is already visiting.

The can navigate websites, read page content, click elements, type into fields, and move through multi-step browser workflows. Sigma specifically lists product research as a use case, including comparing options across websites, checking details, and turning the findings into next steps.

For the laptop brief, this creates a straightforward workflow. The user can research qualifying models, open retailer and manufacturer pages, compare the exact configurations available, and keep the relevant pages inside the same browsing environment.

That becomes useful when the information needed for a purchase is spread across several sources. A retailer may have the current price, while the manufacturer's page provides the exact dimensions or weight. Reviews can add context around battery life, display quality, or other details that are difficult to judge from a specification sheet.

Sigma can work with those pages directly rather than requiring the user to copy information into a separate AI interface. Its broader also includes page-aware AI and Deep Research for workflows that require information from multiple sources.

The final product page still deserves a manual check before any purchase. Prices, stock, shipping conditions, configurations, and seller details can change while the research is happening.

ChatGPT Shopping Research: Research Built Around Constraints

is designed around purchases where the user has several requirements to balance.

A shopper can describe the product, budget, use case, and preferences in normal language. Shopping Research can ask follow-up questions, search public retail information and merchant data, bring back potential products, and refine the research as the user reacts to the options.

That structure fits a request like our laptop brief because several conditions have to remain visible at the same time. Price, RAM, storage, weight, and intended use all influence the shortlist.

The experience also focuses on explaining tradeoffs between products. A lighter laptop may offer less battery capacity, while another model may provide a stronger display at the cost of extra weight. The user can refine the research without starting the entire process again.

OpenAI also warns that product information can still be wrong or outdated, particularly for price and availability, and recommends checking the retailer before buying. makes that limitation explicit.

Google: Shopping Across Search and Gemini

Google is connecting AI shopping to a much larger commerce infrastructure.

Its works across retailers and Google services including Search and Gemini. Google says the cart can monitor deals and price drops, provide price-history information, alert users when an item returns to stock, and flag compatibility problems.

This makes the system particularly relevant when current commerce data is central to the decision.

For our laptop example, the research question may eventually become more specific: which qualifying model is available at the best current price, and is that price actually unusual?

Google's shopping infrastructure is built to provide that kind of context. The company is also expanding agentic checkout through the , allowing participating retailers to connect product discovery, cart activity, and payment more closely.

A product recommendation and the process of buying that product are therefore becoming part of the same AI-assisted workflow.

Microsoft Copilot: Product Discovery With Price Context

lets users search for products conversationally, compare options, explore retailer information, view review summaries, and inspect price history when that data is available.

Price history can change how a recommendation is interpreted.

A laptop at $899 may fit the budget, but that does not automatically mean the current offer is attractive. Knowing that the same model regularly sells for less gives the shopper useful context before buying.

Copilot also allows follow-up questions after the first set of results. The shopper can tighten the budget, change a specification, remove a brand, or ask for alternatives without rebuilding the search manually.

Microsoft recommends verifying important product information and current pricing on the retailer's website because availability and merchant data can change.

Alexa for Shopping: Shopping Inside Amazon's Ecosystem

Amazon renamed Rufus to in May 2026 as the assistant expanded into a broader agentic shopping experience.

The service works closely with Amazon's catalog and shopping activity. It can answer product questions, compare options, provide personalized recommendations, surface deals, add products to a cart, and support shopping actions tied to the Amazon environment.

Amazon has also added price-history features that can show shoppers how the current offer compares with previous prices. On eligible products, can cover up to 365 days.

That integration is useful when Amazon is already the marketplace where the user expects to buy.

The scope is different from a browser workflow that moves freely between retailers, manufacturer pages, review sites, and other sources. Amazon's advantage comes from how deeply the assistant is connected to its own commerce environment.

How the Five AI Shopping Assistants Differ

The main difference is where each system gets its shopping context and how far it can continue after finding a product.

ChatGPT is built around multi-source product research and conversational refinement. Google connects AI to a large commerce system with pricing, inventory, and checkout infrastructure. Copilot combines product discovery with retailer and price information. Alexa for Shopping is closely integrated with Amazon. Sigma keeps the workflow inside the browser and can interact with the websites involved in the research.

Those approaches can produce very different experiences even when the user starts with the same budget and product requirements.

What Makes an AI Shopping Assistant Useful?

The quality of a shopping assistant becomes easier to judge when the task contains hard requirements.

Finding a laptop below $1,000 is relatively easy. Finding one that remains below $1,000 while also meeting the RAM, storage, weight, seller, and use-case requirements requires more careful research.

Constraint Handling

A hard requirement should remain hard throughout the research.

If the maximum budget is $1,000, repeatedly suggesting $1,100 products makes the result less useful. The same applies to storage, compatibility, delivery dates, dimensions, or any other requirement that the shopper cannot compromise on.

Good shopping workflows also make it easy to change those constraints without rebuilding the entire search.

Source Transparency

Product names are often shared across several configurations.

Two laptops from the same product family may have different processors, storage capacities, displays, or memory. An AI can identify the correct model family while pulling a specification from the wrong configuration.

Recommendations become easier to verify when the assistant provides access to the retailer, manufacturer, or other sources behind the information.

Price Freshness

Shopping information has a short shelf life.

Prices change, temporary discounts end, sellers run out of stock, and configurations disappear. Any system used for deal discovery should make current price and availability easy to verify before checkout.

Price history can also help distinguish a genuinely unusual discount from an ordinary selling price.

Workflow Continuation

Product discovery is only part of online shopping.

After choosing a candidate, a shopper may still need to compare sellers, inspect shipping terms, select the correct configuration, check a discount, or move through the retailer's website.

This is where the gap between a research assistant, a commerce platform, and a browser agent becomes much more visible.

Why Browser Agents Change the Shopping Workflow

A research assistant can tell you what products look promising. A browser agent can continue working with the pages where those products are sold.

Sigma's can navigate websites, click, type, read content, and work through multi-step page flows. Sigma also lists product comparison and website actions among the workflows supported by the agent.

For shopping, that can reduce repetitive browser work. Instead of manually moving between several product pages and rebuilding the same comparison each time, the agent can help collect and organize the information while the relevant pages remain open.

The user should still review important actions before submitting information or making a purchase. Sigma's own guidance on recommends user oversight for sensitive steps such as forms, payments, messages, and other consequential actions.

Which Approach Fits Different Shopping Tasks?

Different shopping systems make sense for different parts of the purchase process.

ChatGPT's Shopping Research fits purchases where several requirements need to be researched and compared across sources. Google becomes especially relevant when current product data, stock, pricing, and commerce actions are important.

Copilot can be useful when the shopper wants conversational discovery alongside price information. Alexa for Shopping is closely tied to Amazon and makes the most sense when the purchase is already happening inside that ecosystem.

Sigma fits a different workflow. Its agent stays inside the browser, so the research can continue across the actual websites involved in the purchase. That is useful when the task requires moving between retailer pages, manufacturer information, reviews, comparisons, and other web sources.

None of these approaches removes the need to verify the final offer.

The Bigger Shift Is From Search to Delegation

Online shopping has traditionally rewarded users who are good at searching. They learn which queries to use, which filters matter, which comparison sites to trust, and how to keep track of several tabs at once.

AI changes the starting point.

Instead of constructing the perfect search query, the shopper can describe the outcome:

I have $1,000. Find a lightweight laptop for work and travel that meets these requirements and show me the options worth considering.

The system then has to turn that goal into research.

The next stage is already appearing across several platforms. AI systems can monitor prices, work with carts, interact with retailer data, navigate websites, and support more of the steps between the original request and the final purchase.

The user still makes the important decision. What changes is how much manual research has to happen before that decision.

Final Takeaway

AI shopping assistants are developing along several different paths.

concentrates on detailed product discovery and comparison. Google's connects AI with shopping data and commerce infrastructure. combines conversational discovery with retailer and pricing context. is deeply connected to Amazon's retail environment.

brings the agent into the browser itself, allowing product research to continue across the websites involved in the decision.

The useful question is therefore broader than which assistant can produce the longest product list. A shopping tool has to understand the constraints, surface information that can be checked, keep up with changing prices, and support the steps that come after the first recommendation.

As shopping becomes more agentic, those differences will matter more than the chatbot interface alone.