Skip to main content
AI Categorization step of the Enrich with AI modal with the Categories per product switch set to Auto and the Start Categorization button AI Categorization step of the Enrich with AI modal with the Categories per product switch set to Auto and the Start Categorization button AI Categorization reads each product’s content and places it in your existing category tree. It does not invent categories. You decide whether its answers are saved right away or wait for you in the Review inbox. What it enables: categorize a freshly imported catalog in minutes instead of days, keep categories consistent as you add products, and clear the uncategorized bucket that quietly hurts navigation and merchandising.
AI Categorization is promptless: it needs no prompt from your library. It costs 1 credit per product. Products no model could answer are refunded.

​
When to use it

  • After an import where products arrive with no category, or a flat supplier category that doesn’t match yours.
  • To clear the uncategorized bucket that builds up over time.
  • When category assignment is inconsistent, with similar products scattered across different branches.
  • As an automation step, so new products are categorized the moment they’re created.

​
How to run it

1

Select products

Check products individually, or filter (for example, products without a category) and select all matching results.
2

Open the Enrich modal

Click Enrich with AI in the toolbar.
3

Choose AI Categorization

Pick AI Categorization in the Organize product data shelf.
4

Set the options

Choose how many categories each product can get and what happens with the answers (see below).
5

Start Categorization

Click Start Categorization and track it in the Process Tracker. If a run stops early, use Resume there to finish only the products that were not processed yet, at no extra credit cost.

​
Configuration options

​
Categories per product

  • Auto (default): “AI will determine the optimal number”. Some products belong in one category, some in several.
  • Max categories: switch Auto off and set a cap of 1 to 10 categories per product. Use this when your merchandising rules limit how many categories a product may appear in.

​
What happens with the answers

Answers the AI is less sure about are not lost or marked as failed. They wait in the Review inbox, where you accept or reject them one by one or in bulk.
If your AI review setting asks you to approve category changes, Save sure answers in this run is not available: “Your AI review setting asks you to approve category changes, so this run can’t save them right away.” Change the rule with Change in AI review.

​
What the AI uses

The AI reads the product’s name, description and attributes. When the product has web research with matching identifiers, it reads that too. Your category rules steer every answer: Rules also apply to subcategories. Set them up and try them with Test on 5 products in Managing Product Categories. In an Enrich product content or Fill empty fields workflow, categorization only places products that have no category yet. If its answers need review, the later stages wait until you have reviewed them.

​
What good looks like

  • The uncategorized count drops sharply after the first run.
  • Assignments match your merchandising logic, so products appear where shoppers would look for them.
  • A small, reviewable remainder waits in the Review inbox: the genuinely ambiguous products.

​
Act on what you find

Your threshold or category rules may be strict for this catalog, or your category names are ambiguous. Review a sample: if most answers are right, lower Save answers at or above (%) for the next run, and clarify vague category names. Outcome: fewer answers to review without losing accuracy.
A misplacement usually traces back to thin product content. Run web research and enrich the title, description and attributes first, then re-categorize. Add a Never put products here rule if one category keeps attracting the wrong products. Outcome: accurate placement driven by richer data.
Switch Auto off and set Max categories to 2 or 3. Outcome: category counts that match your merchandising rules.
The answers need review before enrichment can continue. Review them in the Review inbox, then resume the workflow in the Process Tracker. Outcome: the rest of the workflow runs with the right categories.
Add AI Categorization as a step in an automation triggered on product creation or import. Outcome: new products are categorized the moment they enter the catalog.

​
Measure the impact

Better categorization improves navigation and category-level merchandising. Review structure and category health in Category Insights.

Managing Product Categories

Build your category tree and set category rules.

Category Insights

Analyze category structure, health, and revenue.

AI Attribute Extraction

Fill the specs that help the AI categorize accurately.

Enriching Products with AI

Every enrichment action and how to run it.