AI Import
Import products from documents, spreadsheets, and PDFs with AI-powered extraction.
Overview
AI Import extracts structured product data from your files — PDFs, CSVs, Excel sheets, images, and text — and lets you review and selectively import them into your catalog. You can optionally auto-enrich products in the same workflow.
Supported formats: PDF, CSV, XLSX/XLS, JPG/PNG/WebP, TXT — up to 5 files, 20 MB each, 100 MB total.
Getting Started
- Go to Products → AI Import
- Upload your files (drag and drop or click to select)
- Choose an import mode:
- Review and import — review extracted products before anything is written to your catalog
- Import and enrich — finalize products and start enrichment automatically after extraction
- Select an AI model (see AI Models below)
- Optionally add a custom prompt to guide extraction — e.g.
"Extract product name, SKU, price, and available colors" - Click Start Import — extraction runs in the background and you'll be notified when it's ready to review
File Format Tips
| Format | Notes |
|---|---|
| CSV / Excel | First row must be column headers, one product per row. Column order doesn't matter. |
| Works best with structured layouts and tables. Scanned PDFs work but with lower accuracy. | |
| Images | Use clear, well-lit images. Good for product labels, packaging shots, or scanned datasheets. |
| Text | Both structured (key: value) and unstructured prose are supported. |
If you're importing a large catalog, test with a small sample file first to validate your prompt and model choice before scaling up.
AI Models
| Model | Best For | Cost |
|---|---|---|
| Budget | Large volumes of simple products, price lists, inventory updates | Lower |
| Pro | Technical products, mixed formats, complex documents, higher accuracy needs | Higher |
Review Interface
After extraction completes, open the Review Dashboard from Import History → Review. This is where you inspect extracted products and choose what to import.
Views
Tree View — shows each product family as a row with its variant cards displayed side-by-side. Each card shows name, SKU, variant options (e.g. Color, Size), and a status icon (✓ imported, ✕ failed, — pending). Good for reviewing variant structure at a glance.
List View — a spreadsheet-like table with inline expandable variants. Parent rows have a colored left accent and show aggregate import progress (e.g. 2/3). Expand a parent with the chevron to see its variant rows inline. Good for reviewing a large flat list quickly.
Toolbar
| Control | Description |
|---|---|
| Search | Filter by name or SKU — matches are highlighted across all visible rows |
| Sort | Alphabetical or families/standalones first |
| Export | Download all extracted products as a CSV |
| Import N | Import all selected products into the catalog |
Selecting and Importing
- Check a parent row to select it and all its variants at once
- In List View, expand a parent to select individual variant rows
- Already-imported products are disabled and cannot be re-imported
- You can finalize in multiple passes — import a subset, review results, then import the rest
Click the maximize icon on any row to view all extracted attributes and any error detail.
Auto-Enrichment
Enable Auto-Enrich during import setup to enrich products immediately after extraction completes — no separate step needed.
What you can configure:
- Enrichment AI model (Budget or Pro — can differ from the import model)
- Web search to pull in additional product information
- Image extraction to automatically fetch product images
- An attribute set to scope which attributes are enriched
Auto-enrichment uses additional credits (shown in the cost estimate before you start). You can always skip it and enrich products manually later via Product Enrichment.
Job Statuses
| Status | Meaning |
|---|---|
| Pending | Queued, starting soon |
| Processing | Extracting products from your files |
| Completed | All products extracted successfully |
| Completed with Errors | Finished but some products failed — review error details |
| Failed | Job could not complete |
| Stopped | Manually stopped — already-extracted products are retained |
To stop a running job, click Stop on the job row in Import History. You can restart at any time with the same files.
Troubleshooting
| Problem | Fix |
|---|---|
| Poor extraction quality | Switch to Pro or add a custom prompt specifying exactly what to extract |
| Missing fields | Name them explicitly in your prompt — e.g. "Always extract warranty period" |
| Duplicate SKUs | Duplicates are detected and skipped automatically |
| Auto-enrichment failed | Check credit balance; switch to Budget model or enrich manually later |
| Job appears stuck | Jobs can take up to 30 minutes for large files — if truly stuck, stop and retry |
Next Steps
- Product Enrichment — enrich imported products with additional attributes
- Attribute Sets — guide extraction with custom attribute definitions
- AI Import API — automate imports programmatically
- Credits & Billing — understand credit usage