Getting Started with AI Search Analysis
With the rise of AI-powered shopping assistants like ChatGPT, Gemini, Claude, how your products are found and presented is evolving. PX Insights is your essential tool to understand and optimize your product performance in these new digital landscapes.
It empowers you to gain unprecedented insights into how AI platforms interpret, describe, and recommend your offerings. Whether you're a retailer focused on maximizing product visibility and driving AI-driven recommendations, or a manufacturer ensuring consistent brand messaging and optimal product positioning, PX Insights delivers tailored value to meet your strategic goals.
Key Capabilities include:
- Simulate Real-World AI Shopping Queries: Run accurate simulations of how customers search for products on leading AI platforms: ChatGPT, Gemini and Claude.
- Evaluate AI Product Visibility: Understand precisely whether your product appears in AI search results, and get a clear explanation of why it does or does not.
- Benchmark Against Competitors: Compare your product's AI visibility directly against similar or competing products to identify market opportunities and gaps.
- Receive Actionable Optimization Suggestions: Get concrete, data-driven recommendations to improve your product descriptions, content, and overall strategy for enhanced AI discoverability.
The process is straightforward and consists of three main steps:
- Generate AI Search Queries: Define the scope of your analysis and let AI generate relevant search queries for your selected products.
- Analyze the Results: Review a detailed report showing your product's rank, competitive landscape, and suggestions for improvement.
- Act on prioritized insights: fix issues across many products at once from the Home dashboard.
Tip: You don't have to launch an analysis yourself. PX Insights can analyze a selection of your products automatically every week.
Step 1: Generate search queries
Step 1 explains how to analyze specific products on demand. Products analyzed automatically by Autopilot skip this step.
Product selection
To begin your analysis, you must first select the products you want to analyze.
- Start from the Product Grid
- Select the products you want to analyse (maximum 20 products or 20 variants or 20 product models)
- Then select the PX Insights analysis button at the bottom of the page
You will automatically be redirected to the PX Insights use cases where you can select search queries configuration.
Parameters definition
To ensure an accurate analysis that matches your analytical goals, you'll need to define two main parameters: Business Type and Locale.
Business Type
This parameter determines the scope of the sales channels included in the analysis.
- Manufacturer: Select this if you want to analyze how your product ranks across all channels where it might appear. This provides a broad overview of your product's visibility regardless of the sales platform.
- Distributor: Choose this option if you want to focus the analysis on how your product ranks within a specific set of channels that you manage or own. You can specify up to five channel names (e.g., "Amazon," "Best Buy," "Walmart.com"). Do not include full URLs, just the channel names.
Locale
The Locale parameter is crucial because search results can vary significantly based on the consumer's geographical location and the language used. Defining the locale allows you to simulate a mock consumer's environment, ensuring the analysis is relevant to specific regional or linguistic contexts. For example, a search query from France might yield different results than the same query from the US, even if the language is the same. Similarly, the language used for the query (e.g., French vs. English) will influence the results.
AI model
Choose which AI assistant to simulate: ChatGPT, Gemini or Claude. Each assistant can find, describe and rank products differently, so pick the one your customers use most. To compare, run the same analysis with another model.
Once the checkbox is selected and the parameters defined, you can click the Launch AI Analyze button.
The selected model is running in the background to generate search queries.
Review and Refine generated Search Queries
Once the AI has generated the queries, you can review or edit them to ensure they align perfectly with your analysis goals.
To edit a query, click the pencil icon, make your changes, and click the save icon.
When you are satisfied with the list of queries, click the Confirm button to begin the full analysis.
The analysis is done in the background and can take some time. You will receive a notification once it is complete in your PIM notification panel.
Step 2: Analyze Your Results
Accessing the Analysis Report
This list also includes analyses run automatically by Autopilot.
Access analyses from the main dashboard
To access the AI-Generated results, go to Insights accessible from the main menu > AI Discovery tab
You will get the list of all the analysis you ran with the initial context provided: Business type & Locale selected.
To see more details about the analysis, select the Expand icon on the right.
Access analyses from a product
Open a product in the Product Edit Form, then select the PX Insights tab to access the relevant AI Discovery analyses for that product. This gives you a product-focused entry point without returning to the dashboard.
Detailed analysis
The detailed analysis report provides a comprehensive breakdown of your product's performance for each search query.
AI Search Rank & Overall Summary
The first insight you will get is the AI Search Rank, it means the ranking of your product for a given search query.
The product ranking is done with a maximum of 5 products.
You will also get an Overall summary on why you product has been ranked that way and some key points below to get more details.
Competitive Landscape
On the right hand side, you will get a list of competitive products and how they ranked for a similar search query. You can even see what are the sales channels that have been advertised by the semected AI model to purchase this product.
Suggested Actions
Our AI Discovery tool categorizes improvements into four distinct pillars, distinguishing between PIM-driven data enhancements and external authority signals.
1. Product Data Suggestions (PIM-Based)
This section focuses on the accuracy and completeness of your core product attributes by auditing your PIM product data against AI search standards. Because AI assistants rely on rich and consistent text-based data—such as detailed descriptions—to accurately understand and surface your items, Product Content Managers must ensure these fields are fully enriched. Optimizing these records directly within the PIM ensures AI assistants can confidently match your products to specific user queries.
2. New Attribute Suggestions (PIM-Based)
Beyond individual records, this section highlights how your product data is structured behind the scenes. AI engines require structured, machine-readable fields to compare specifications and filter results effectively; therefore, PIM Admins may need to adjust the PIM schema to include these new, suggested attributes. By refining the data model and standardizing these fields at the administrative level, you create a robust foundation for AI to navigate and recommend your entire catalog based on modern search intents.
3. Marketing Suggestions
This pillar evaluates how your product is positioned and differentiated across various digital touchpoints. AI language models analyze core messaging, competitive advantages, and trust signals (like warranties) to determine brand authority—tasks typically driven by Content and Marketing teams. By ensuring consistent and compelling positioning across all channels, these teams help AI assistants confidently recommend your products as the premier choice for specific buyer intents.
4. Technical Suggestions
This final section covers the technical foundation of your product pages across the web, where AI search engines look for unambiguous signals like canonical URLs and consistent identifiers. To prevent "entity confusion" between different versions of a product, Web Developers or Technical SEO/GEO Experts must maintain a clean technical structure. These updates resolve duplicate entities and consolidate ranking power, ultimately strengthening your product’s overall authority and discoverability.
Sources
For full transparency, the report includes the sources that the AI used to generate its analysis and ranking.
This content is based on AI outputs. Since this content may be incomplete or inaccurate, please validate these outputs to ensure they are accurate and complete before using it.
To strengthen this analysis, Akeneo utilizes a trusted third-party service to access and retrieve publicly available structured data (Schema.org) from the Product URLs belonging to the client own company. This process is strictly limited to "read-only" access of publicly indexed metadata and does not involve the collection of personally identifiable information (PII) or modifications to the User's domain. By using this feature, the User acknowledges and consents to the temporary processing of these public URLs by our designated sub-processor for the sole purpose of product data analysis.
Akeneo only audits URLs belonging to the client own company. Unauthorized scraping of third-party or competitor domains is prohibited.
Analysis is currently available in English only, regardless of the language used for the search query.
Step 3: Prioritize and act at scale from the Home dashboard
The PX Insights Home dashboard groups the suggestions from all your analyses into insight cards. Each card brings together one recommended change that applies to several products, so you can fix it once for all of them instead of product by product.
The cards that matter most come first
Cards are ordered so you see the most valuable ones first: those with the biggest impact and that you can act on directly in the PIM (Product Data and Data Model suggestions). Ranking is based on:
- How important the suggestion is: how strongly it affects your products' visibility in AI search
- How many products are affected
The dashboard shows up to 8 cards per page (2 rows of 4), across up to 3 pages, so you can focus on a short list of high-impact actions.
Filter your view by
- Country
- Product Family
- Suggestion Type (Product Data, Data Model, Marketing, Technical)
- Date Range
Open a card to see the details
Click a card to see:
- The affected families and attributes
- The Individual Suggestions behind the insight, with a counter showing how many there are
Take action
From a card, you can:
- Fix it in bulk with Ziggy: click Ask Ziggy to create the suggested attributes (Data Model) or enrich attribute values (Product Data) for all affected products. This requires the matching Ziggy permission in your role settings.
- Fix it yourself: open the affected products in Sequential edit to update them one by one, or add them to a workflow if the changes need a review step.
- Share it: click Share insight to copy a link to the card and send it to the teammate who owns the change (for example, marketing or IT for Marketing and Technical suggestions).
- Give feedback: use 👍 / 👎 to tell us whether the insight was useful.
