Look for patterns, not a random sample
Bulk AI copy fails in clusters: one template, one supplier, one product type. Checking ten random products can miss a whole cluster. Group products by supplier or template and check each group, and trace every number and spec to a source cell.
Fill missing fields from text, and leave unknowns blank
Colour, size, material and fit are often stated in a description and blank in the field filters use. Fill a field only where the text states it, keep the quoted words, and never invent a barcode. A blank you can see is better than a guess you cannot.
- Keep the quoted source text beside each value.
- Count the blanks left.
- Check a set you labelled yourself, with half held back.
Make every bulk change undoable
Export before the change and again after, and compare the two row by row. Apply in small batches and check a sample of each. Keep the first export until you are sure, because it is your rollback.
Which paid job fits
The copy check traces up to 150 products to your source data. The supplier-facts job writes descriptions with a fact trace. The attributes job fills missing fields from text.
The bulk-edit job gives you a rehearsed procedure and a rollback file. You import every file yourself.
Sources and limits
- Shopify Community: checking AI-generated content after a bulk update Checked 2026-10-12.
- A thread of 25 to 28 Sep 2026 lists what goes wrong after bulk AI updates: copy describing a different item, sentences cut off, made-up specs, fields changed that should stay untouched, garbled characters, and page text drifting from the feed and structured data.
- A poster says random spot-checking tends to hide the failures because the bad ones cluster by pattern, and to keep the pre-update export as your rollback.
- Google Merchant Center Help: disclose content created with generative AI Checked 2026-10-12.
- Google asks merchants to disclose text created with generative AI in product data, using separate attributes for AI-written titles and descriptions.
- Google Search Central: spam policies Checked 2026-10-12.
- Google names using generative AI tools to generate many pages without adding value for users as scaled content abuse.
- Shopify Community: managing product data for large catalogues Checked 2026-10-12.
- A poster on 29 May 2025 says Shopify's AI tools help with descriptions but are not yet reliable for structuring data or tagging, and lists manually fixing vendor-supplied data that comes in messy formats among the manual tasks.