№09 · Workflow · E-commerce
Catalog to AI-Ready Product Pages
Your whole catalog, structured and reviewed.
Paid once. Checkout by Polar, our merchant of record, which works out any sales tax or VAT.
In these sets: AI-Ready Store, $19 · All-Access, $49
Or included in membership.
- Format
- Workflow
- Made for
- Solo founders, small brands
- Difficulty
- Advanced
- Time to value
- ~Half a day
Times are estimates until a platform test run confirms them.
CompatibleChatGPT · Claude · Gemini · Grok · Any model
Overview
- The problem
- “I have a lot of SKUs with inconsistent data, and doing them one page at a time on my own won’t scale.”
- What you get
- A cleaned attribute sheet for the catalog, rewritten PDP copy for each style, a markup and feed QA checklist completed, and a monitoring log, all through 3 named review gates that one person can run.
- What’s included
- Workflow guide (6 stages, 3 gates, solo owners)
- Bot profile (system prompt, knowledge and setup, guardrails)
- Attribute schema by category
- Extraction prompts
- PDP copy rules + prompts (Product Page Rewrite Bot optional)
- Markup/feed QA checklist (Google merchant-listing Product fields, feed completeness)
- Batch run sheet (.md + .docx)
- Working catalog workbook (.xlsx: attribute schema, catalog with completeness/missing/conflict checks, batch review G1–G3, markup & feed QA, monitoring log)
- Platform setup notes (ChatGPT, Claude, Gemini, Grok Bot, any chat tool)
- Ready-made skill (folder + zip)
- PDF guide
- Illustrative example
- What you’ll need
- Product export (CSV/feed, product data only)
- Existing PDP copy
- Spec sheets or tech packs
- Brand bible or voice notes
How it runs
6 steps- Export one batch
- Bot extracts and normalizes attributes → G1: you resolve conflicts and gapsYour call
- Bot drafts copy per style → G2: you read every page in batches 1–2, then a sample + every flagged pageYour call
- Markup/feed QA checklist → G3: you (or a developer you hire) validateYour call
- You publish by hand
- Log monthly checks
Your call marks a step where you review and decide. Nothing moves on without you.
Estimated time: first batch of 25 SKUs in ~half a day including your reviews. Full catalog depends on size.
Starter prompts
- Extract attributes from this catalog export into the schema
- Flag SKUs with missing fit/fabric/care
- Draft PDP copy for batch {n}
Fill in the parts in braces. The full prompt set is in the kit.
Sample output
Batch 1 (6 SKUs): 3 complete. Missing: image (1), care (2). Conflicts: fiber on 2 Charcoal Crew SKUs (export 80/20 vs spec 70/30). I haven’t picked one. After G1: copy drafted for 2 styles. Beanie held until its photo is added. JSON-LD drafted without gtin (none supplied).
Produced by a manual walk-through of the prompt, not a platform test. It will be replaced with real output after a test run.
Platforms
| AI tool | Label | Setup note |
|---|---|---|
| ChatGPT | Compatible | Project setup, or skill upload where Skills show |
| Claude | Compatible | Project or skill |
| Gemini | Compatible | Skill, workbook kept in your own sheet |
| GrokRuns as a Grok Bot (beta) | Compatible | Read-only files |
| Any modelAny capable chat model | Compatible | Paste one batch at a time |
We ran it ourselves on this tool. Shows the date and model of the run.
No piece carries this yet. Tested means we have run it ourselves. We update the label as each piece is tested.
Built to run on this tool. Not yet confirmed by our own test run.
Works with a limit we name.
Tools required
Chat model that can read files (or paste rows). Spreadsheet (Excel, Google Sheets or Numbers). Read-only store export. Google’s Rich Results Test. A developer or store admin only if your theme needs markup changes.
House standards for this piece
All our standards- Human review gate
- Three gates, all yours: you resolve conflicts and gaps (G1), review the copy (G2), and validate markup and feeds (G3) before anything is published.
- Data grounding
- Your catalog and spec sheets are the source of truth. Conflicts are shown as [CONFLICT], never settled by a guess.
- No invented results
- No ranking, traffic, AI-visibility or sales promises.
- AI disclosure
- Product copy is AI-assisted and reviewed by you.
- IP & bias
- No competitor copy. Inclusive fit language that describes the garment, not the wearer’s body.
- Privacy
- Product data only. No customer or order data.
- Approval boundary
- Never writes to your store, PIM, feed or theme. You publish by hand.
Questions people ask
I’m not technical. Can I do stage 4?
Often, yes: many store platforms add Product markup for you, and stage 4 is then mostly checking. If your theme needs code changes, hand the checklist to a freelance developer or store admin for that part only.
Will this get my products into AI shopping answers?
We can’t promise that, and neither can anyone else. Google says it doesn’t guarantee that features using structured data will appear. Clean, consistent product data is the part you control. The monitoring log helps you see what changes.
How big should a batch be?
Start with 10 SKUs, then up to 25. Small batches keep your reviews honest.
Do I have to read every page?
In your first two batches, yes. After that, read a random sample of at least 5 per batch plus every page with a [VERIFY] flag or a fact change.
Which AI tool should I use?
One that can read files, or paste rows into any chat tool. ChatGPT (with a Project), Claude (with a Project or skill), Gemini (with a skill), Grok Bot and any chat tool are marked compatible: they should work, but we haven’t completed verified test runs yet. ChatGPT, Claude, Gemini and the any-chat-tool route work on free plans. Grok Bot needs a SuperGrok or Cursor plan.
Can I use it commercially?
Yes, for your own brand. The license covers one business. Please don’t resell or republish it, or use it for client work.