AI Shopping Assistants: 45% Use AI for Discovery Guide

The 45% Number in Context

The NRF and Salesforce joint survey, conducted across 3,200 online shoppers in January 2026, found that 45% used an AI assistant at some point during their most recent purchase journey. This is a significant jump from 18% in the same survey two years earlier — a 2.5× increase in adoption in 24 months. The rate of change matters as much as the absolute number.

The survey distinguished between different types of AI interactions. The most common use cases were product comparison queries (used by 62% of AI shoppers), sizing and fit questions (48%), compatibility checks between products (41%), and price-to-value assessments (38%). Notably, fewer than 12% reported completing a purchase through the AI assistant interface itself. The dominant pattern is AI as research layer, not AI as transaction layer.

The implication for retailers is that AI assistants are most active at the top and middle of the funnel — generating awareness, shortlisting options, and resolving pre-purchase uncertainty. When AI surfaces your product as a strong match for a comparison query, the shopper typically arrives at your product detail page already pre-qualified. Conversion rates for AI-referred sessions are consistently higher than for organic search visits in early 2026 data, averaging around 3.2× in categories like electronics, apparel, and home goods.

How AI Assistants Discover Products

Understanding the data pipelines AI shopping assistants rely on is essential before optimizing for them. Different assistants use different primary sources, and optimizations that work for one may be less effective for another. The major platforms in early 2026 fall into two broad groups: web-crawl-based assistants and marketplace-integrated assistants.

For retailers with direct-to-consumer sites, the web-crawl-based assistants represent the largest opportunity because product detail page optimization is fully within your control. ChatGPT's shopping mode, which launched its product carousel feature in late 2025, now surfaces structured product cards with images, prices, and direct links for purchase-intent queries. The data feeding those cards comes from crawled product pages and merchant data feeds. For a closer look at how ChatGPT's visual shopping capabilities work alongside AI discovery, see our comparison of ChatGPT visual shopping and image search.

Structured Product Data Optimization

Structured product data is the highest-leverage change most retailers can make. AI assistants parse schema.org Product markup directly without needing to infer attributes from prose descriptions. A product page with complete structured data will consistently outperform an equally good page without it in AI citation tests.

The additionalProperty field is particularly important for AI assistant accuracy on compatibility and specification queries. Each technical attribute — weight, dimensions, material, voltage, compatibility, color, size — should be a separate PropertyValue entry rather than being buried in prose descriptions. When a shopper asks an AI assistant whether a laptop stand is compatible with a 16- inch MacBook Pro, the assistant needs to find a numeric dimension value in structured data, not parse a sentence like “accommodates most laptops.”

Beyond schema markup, product feed quality matters for ChatGPT's shopping integration and Google Shopping feeds that feed into Gemini recommendations. Feed fields like product category, color, size, material, and condition directly influence which queries trigger your product as a candidate. Treat your product feed as a first-class data asset, not an afterthought to your product catalog.

Conversational Content Strategy

Structured data handles machine-readable attributes, but AI assistants also evaluate natural language content on product pages. The content pattern that drives the highest AI citation rates is question-and-answer formatted copy that mirrors the queries shoppers actually ask, rather than traditional marketing-style descriptions that lead with brand benefits.

The practical implementation is a dedicated FAQ section on every product detail page, structured with actual question headings and factual answers. Pull questions from your customer support tickets, live chat transcripts, product reviews, and any “people also ask” queries you see in Google Search Console. These are the queries real shoppers ask — and therefore the queries AI assistants receive about your products.

For category pages and collection pages, conversational content strategy means adding buyer's guide sections that explain how to choose between products in clear, factual language. AI assistants frequently surface these comparison-format pages when shoppers ask general category queries like “what is the best standing desk under $500.” The agentic commerce protocols emerging in 2026 are further structuring how AI agents navigate these decisions — see our guide on agentic commerce protocol for AI shopping agents for the technical underpinnings.

Checkout Flow and AI Handoff

When an AI assistant refers a shopper to your product detail page, that shopper arrives pre-qualified — they have already done their comparison research and resolved their main objections through the AI conversation. The checkout experience needs to be built for this new entry pattern, not just for cold traffic arriving from traditional search.

Inventory accuracy and real-time stock status on product pages also significantly affect AI assistant quality. When an AI assistant confidently recommends a product that turns out to be out of stock or discontinued, the resulting customer friction damages trust in both the AI assistant and your brand. Retailers with real-time inventory sync to their product pages — and accurate availability data in their product feeds — see fewer abandoned sessions from AI-referred shoppers.

Measuring AI-Driven Traffic

Standard UTM attribution and referral traffic reports significantly undercount AI assistant-driven visits. AI assistants typically relay URLs to users without appending UTM parameters, and many AI interfaces open URLs in new browser tabs that appear as direct traffic rather than referrals. Building accurate AI traffic measurement requires a multi-signal approach.

Retailer Readiness Checklist

The following checklist covers the concrete changes that most directly improve AI shopping assistant citation rates and AI-referred conversion performance. Items are ordered roughly by impact and implementation effort.

What Changes for Paid Acquisition

As AI assistants capture a growing share of upper-funnel product discovery, the paid media landscape shifts alongside them. Retailers who rely heavily on top-of-funnel paid search for product discovery will see diminishing returns as organic AI discovery fills the same role more effectively for a growing segment of shoppers.

The net budget impact for most retailers will be a shift toward AI-native ad formats and away from broad text search. The product feed quality work that drives organic AI citations also improves performance in Shopping ad campaigns, making structured data investment doubly valuable. For retailers building a full paid strategy that accounts for AI-native channels, our eCommerce solutions team can help map current budget allocation against emerging AI traffic patterns.

The Future of AI-First Commerce

The 45% adoption figure from January 2026 represents an inflection point, not a plateau. Industry projections from Forrester and IDC place AI-influenced commerce at $890 billion by 2028, accounting for a majority of online product discovery in high-consideration purchase categories. The trajectory is clear: AI assistants are becoming the primary interface between shoppers and product information.

The retailers who will win in this environment are not necessarily those with the largest advertising budgets but those with the most complete, accurate, and machine-readable product data. A $200 product with complete schema markup, a detailed FAQ section, and accurate specifications will be cited by AI assistants more reliably than a $2,000 product with vague marketing copy and incomplete attributes. Quality of information, not volume of advertising, determines AI discovery outcomes.

Longer term, the emergence of agentic shopping — where AI agents complete purchases autonomously on behalf of consumers — will further shift requirements. Agents that can browse, evaluate, and transact will need not just parseable product data but also standardized pricing APIs, real-time inventory endpoints, and structured checkout flows. The retailers building clean product data infrastructure today are also building the foundation for agentic commerce compatibility tomorrow.