Most catalogues imported from Shopify or a legacy platform have gaps: missing brand names, vague descriptions, wrong category, no apparel attributes. AI enrichment fills those gaps in bulk, across thousands of products in a background job. Every run shows a preflight quote of what it will cost before it starts, runs on a sample you approve first, and records every change in the activity log so you can revert any field.
Catalogues imported from a real store are never complete. Descriptions written for a product page are too vague for a shopping feed, categories are empty, and apparel attributes like colour, size and material — which Google and Meta both require — were never filled in because the storefront did not need them.
Fixing that by hand is thousands of small decisions. Handing it to AI without controls is how you end up with a catalogue full of confident nonsense, spent credits you did not budget for, and no way back.
How it works
FeedGraph finds the gaps first
Products are grouped by what is actually missing, and by category, so you are working on "412 products with no colour" rather than an undifferentiated pile.
You see the cost before it runs
Every batch shows a preflight quote — how many products, which fields, how many credits. Per-run and per-day ceilings mean a large job cannot quietly drain your balance.
It runs on a sample you approve
AI writes the fields for a handful of products in that group first. You approve, reject or edit those, and FeedGraph learns which fields and which style you accept before touching the rest.
Then it runs in the background
The rest of the group is processed as a background job, so you can close the tab. Every value written is recorded as a field-level change with the previous value kept.
Anything can be undone
AI-written fields are flagged as AI-written, against the value they carry. Any field can be reverted individually, and the revert is itself logged.
What you get
- ✓Titles, descriptions, product categories and structured attributes filled across thousands of products
- ✓A cost quote before every run, and hard per-run and per-day credit ceilings
- ✓Sample-then-apply per category, so you approve the output style before it scales
- ✓Field-level provenance: what AI wrote, when, and what was there before
- ✓Per-field revert, and a full record in the activity log
- —It will not invent facts it cannot see. It works from your existing product data and images — it cannot know a material or a measurement that appears nowhere.
- —It does not run automatically. Enrichment is something you start, quote and approve; there is no background job silently rewriting your catalogue.
- —It is not free. Every operation is metered in credits, and the rate depends on the field — filling a title is 1 credit, a description is 2, and reading the product photo to answer a field is 4.
Questions
What does it cost to enrich my catalogue?
It depends on which fields are missing. A title is 1 credit per product, a description 2, a category 2, attribute extraction 2, and a field answered by reading the product image 4. At $0.10 per credit, filling titles and descriptions across 1,000 products is about $300 — and you see the exact quote before the run starts.
Can I check the output before it touches my whole catalogue?
That is the default. Enrichment runs on a sample of each category first and waits for you to approve, reject or edit those values. Only then does it process the rest of that group.
What if the AI gets something wrong?
Every AI-written field is flagged as such and keeps its previous value. You can revert any individual field from the change history, and the revert is recorded too — nothing is quietly overwritten.
Rewind anything.
Every change to your catalog is snapshotted — manual edits, AI enrichment, syncs, pre-publish saves. Scrub back to any point and preview exactly what a rollback would restore, before you commit.
See AI enrichment & fix in use
More in Content and creative
All features →AI product images
Generate and edit product images at scale — white-background heroes for Google and Meta, and lifestyle shots.
View feature →AI product video
Short product videos generated from your own product imagery, in three quality tiers so you can choose cost per SKU.
View feature →Ad creatives
Platform-sized ad images built from templates — your product, your text, your brand colours. No AI credits used.
View feature →Ready to try it?
FeedGraph is in early access. Apply to join and we'll onboard you personally — no credit card required.
Join early access