How AI agents browse, evaluate and recommend products and services

To optimise your Wix site for AI shopping agents, you first need to understand exactly how they work under the hood. AI agents do not browse the web like humans. They use a combination of direct crawling, API calls, structured data extraction, and large language model reasoning to evaluate websites in seconds. This lesson provides a detailed technical breakdown of the AI agent browsing and evaluation process, specifically focused on what this means for your Wix product and service pages.

The AI Agent Browsing Process: Step by Step

When an AI agent receives a shopping query like "best organic baby clothes UK under £30 with free delivery", it executes a sophisticated multi-step process that happens in seconds.

How an AI shopping agent processes a product query

• Query understanding: the agent parses the user instruction to identify product category (organic baby clothes), location (UK), price constraint (under £30), and requirement (free delivery)
• Source identification: the agent identifies potential sources using its training data, real-time web search, product feed databases (Google Merchant Center, Bing Shopping), and known authority sites in the category
• Page crawling: the agent visits candidate product pages, downloading the HTML, parsing JSON-LD structured data, and extracting visible text content
• Data extraction: the agent extracts specific data points: product name, price, currency, availability, delivery options, features, images, reviews, ratings, brand, and category
• Comparison and ranking: the agent compares extracted data across all candidate products against the user criteria, ranking options by relevance, price, reviews, and availability
• Trust evaluation: the agent assesses source credibility using brand recognition, review volume and recency, domain authority signals, and whether the business has verified structured data
• Recommendation generation: the agent generates a natural language recommendation with specific product suggestions, price comparisons, pros and cons, and direct links to purchase

What AI Agents Can and Cannot Parse on Wix Sites

Understanding the technical capabilities and limitations of AI agents when visiting Wix sites is essential for effective optimisation.

What AI Agents Can Parse Successfully

• JSON-LD structured data in the page head or body: this is the primary data extraction method and the most reliable
• Visible text content in the HTML DOM: product descriptions, specifications, pricing text, and feature lists
• HTML table data: comparison tables, specification tables, and pricing tables in standard HTML format
• Meta tags: title, description, Open Graph, and Twitter Card meta tags are read and used for context
• Image alt text: used to understand product images when visual processing is not available
• Wix Stores product data: Wix automatically generates some product structured data for Wix Stores products
• Canonical URLs and navigation structure: used to understand site hierarchy and find related products

What AI Agents Struggle With on Wix Sites

• Prices displayed only in images or graphics: agents cannot reliably extract pricing from non-text elements
• Dynamic content loaded after user interaction: if information only appears after clicking a tab, accordion, or button, many agents will not trigger that interaction
• Information spread across multiple pages: if delivery info is on one page and returns policy on another, agents may not connect them
• Custom JavaScript widgets: third-party review widgets, price calculators, or dynamic pricing tools that render content client-side may not be visible to all agents
• Login-gated content: any information behind a login wall is invisible to AI agents
• PDF-only information: product specifications, price lists, or catalogues in PDF format are harder for agents to parse than on-page HTML content

The Data Points AI Shopping Agents Extract

AI agents attempt to extract a specific set of data points from every product or service page they visit. The more of these data points your Wix pages provide in machine-readable format, the higher your chances of being recommended.

• Product or service name (exact match from schema or H1)
• Price and currency (from Offer schema or visible text)
• Price validity period (sale prices, promotional pricing)
• Availability status (InStock, OutOfStock, PreOrder from schema)
• Delivery options and costs (shipping details from schema or page content)
• Return and refund policy (from schema or on-page text)
• Aggregate rating and review count (from AggregateRating schema)
• Individual review text and ratings (from Review schema)
• Product features and specifications (from page content and schema)
• Brand name and manufacturer (from Brand schema)
• Product category and subcategory (from BreadcrumbList schema and page structure)
• Images with descriptive alt text (from ImageObject schema and HTML)
• Business location and service area (from LocalBusiness schema)
• Business credentials, certifications and awards (from page content and schema)
• Comparison data with competitors (from comparison tables and content)

How Different AI Agents Prioritise Information Differently

Not all AI agents weigh the same signals equally. Understanding these differences helps you optimise for the platforms that matter most to your business.

• Google Shopping AI: heavily weights Merchant Center feed data, product schema markup, and Google Reviews. Prioritises price accuracy, availability, and delivery speed.
• ChatGPT Shopping: weights visible page content heavily alongside schema. Prioritises comprehensive product descriptions, clear feature lists, and brand authority signals.
• Perplexity Shopping: weights structured data and pricing transparency. Prioritises products with complete schema, visible pricing, and high review volumes.
• Bing Copilot Shopping: weights Microsoft Merchant Center data and Bing index data. Benefits from Bing Webmaster Tools verification and Microsoft advertising data.

Complete How-To Guide: Auditing Your Wix Pages for AI Agent Parseability

Follow these steps to check whether AI agents can extract the data they need from your Wix pages

• Select your five highest-traffic product or service pages on your Wix site
• For each page, view the page source (right-click > View Page Source in Chrome) and search for application/ld+json to find any existing structured data
• Copy the JSON-LD code and paste it into Google Rich Results Test at search.google.com/test/rich-results to check for errors and missing fields
• On each page, manually verify that the following are visible as text (not only in images): product name, price, availability, delivery information, and return policy
• Check whether aggregate review data (star rating and review count) is visible on the product page itself, not just on a separate reviews page
• Test whether any important product information is hidden behind tabs, accordions, or click-to-reveal elements that an AI agent might not trigger
• Verify that your Wix Stores product pages include Product schema with at minimum: name, description, image, offers (price, priceCurrency, availability), and aggregateRating if reviews exist
• For service pages, verify that Service or ProfessionalService schema includes: name, description, provider, areaServed, and priceRange
• Ask ChatGPT with browsing enabled to describe what your product page offers and note which data points it correctly extracts and which it misses
• Create an AI agent readiness scorecard for each page: score 1 point for each data point that is both visible on page and present in structured data, out of the 15 data points listed in this lesson
• Prioritise pages scoring below 10 out of 15 for immediate optimisation using the techniques in the following lessons