Get products ready for AI search with Codex
Getting products ready for AI search with Codex means running OpenAI's terminal agent against StoreSEO to read a product's GEO checklist, preview FAQs and an AI snippet, and apply only the parts you approve while Codex's own tool approval stays on.
Last updated: October 2026
Set up Codex for StoreSEO
Skip this if Codex is already connected. The full guide, with the alternatives, is on the Codex page.
- 1
Add StoreSEO to Codex
Run this in any terminal. It writes a [mcp_servers.storeseo] table to ~/.codex/config.toml, which the Codex CLI, the IDE extension and the ChatGPT desktop app all read.
Terminalcodex mcp add storeseo --url https://mcp.storeseo.com/mcp
- 2
Sign in to your store
Run the login command. Your browser opens: pick your Shopify store and approve access. Then run codex mcp list, or type /mcp inside Codex, to confirm storeseo is there.
Terminalcodex mcp login storeseo codex mcp list
How does this work in Codex?
You start Codex in a terminal, normally in the project that holds your theme, with the storeseo server registered in config.toml or through codex mcp add. A request such as: read the GEO checklist for Ceramic Pour Over and tell me what is missing for AI shopping agents, makes Codex call storeseo_start and then storeseo_get_resource_seo. The geo block lists every AI-readiness criterion that passes or fails. Codex prints what it is about to call, and with the server's approval mode set to writes it runs reads straight away and stops to ask before a tool that changes something.
The preview is where you shape the result. storeseo_preview_resource_optimization generates FAQs and an AI snippet and projects the new SEO and GEO scores without saving, and it spends AI credit, so Codex should name that before it runs one. Generation can take a while. Codex waits 60 seconds for a tool by default, so if a preview times out on a long product, raise tool_timeout_sec for the storeseo server and run it again; re-previewing spends credit again, which is a good reason to set the timeout before the first try.
Applying is selective. storeseo_apply_resource_optimization accepts a subset, so you can approve the FAQs and AI snippet and leave out, say, a rewritten URL handle, and the skipped fields stay applicable from the same draft for one hour. When the apply returns, Codex has the re-analysed checklist and can report what now passes. For a thin description Codex uses storeseo_scan_product_description, relays the questions to you in the terminal, and calls storeseo_write_product_description with your answers after a yes.
-
You: Read the GEO checklist for Ceramic Pour Over and prepare it for AI search.
- storeseo_start read-only
- storeseo_get_resource_seo read-only
-
Codex: GEO is 38. It has no FAQs or snippet and the images lack alt text. A preview spends AI credit and saves nothing. Run it?
-
You: Run it.
- storeseo_preview_resource_optimization preview
-
Codex: Draft has six FAQs and a snippet, projected GEO 63. Apply only the FAQs and leave the snippet for later?
-
You: Yes, FAQs only.
- storeseo_apply_resource_optimization needs approval
What to ask Codex
Using the storeseo server, read the GEO checklist for Ceramic Pour Over and list failing criteria. Do not call anything that writes.
Preview FAQs and an AI snippet for Ceramic Pour Over and show me the projected GEO score. Apply nothing until I answer.
Apply only the FAQs from the last preview and skip everything else. Then re-read the GEO checklist.
- Which of my products have the lowest GEO score, and what is missing for AI shopping agents?
- Make the product Merino Beanie ready for AI search: FAQs, AI snippet and product schema. Preview first.
- Scan the description of my best seller for AI agents and ask me the questions it needs.
- Rewrite the description of Ceramic Pour Over from my answers. Keep a copy of the old one for me.
- Is product JSON-LD schema enabled on my store? If not, explain what turning it on changes.
What StoreSEO does: ai search readiness
GEO is StoreSEO's AI readiness score for products. The agent reads which GEO criteria fail, fixes the ones a tool can fix (FAQs, AI snippet, alt text, product schema), and can rewrite a thin description from your own answers to StoreSEO's questions.
Requires: FAQs, AI snippets and descriptions spend AI Content Optimizer credit. The product needs a focus keyword, set in the StoreSEO app, before a description scan or rewrite.
- 1
Read the GEO checklist
The agent reads the product's GEO block: the score and every AI-readiness criterion that passes or fails.
- 2
Turn on product schema if it is off
Product JSON-LD is a shop-wide setting, so enabling it clears that criterion for every product at once. If you then decide not to apply the rest, the agent switches back off only the schema it turned on.
- 3
Preview FAQs and the AI snippet
The preview generates FAQs and an AI snippet (summary, benefits, best for) and projects the new SEO and GEO scores without saving.
- 4
Scan the description
For a thin or missing description, StoreSEO scores it on seven pillars (identity, materials, dimensions, features, use cases, what is included, care) and returns questions. The agent asks you each one, word for word.
- 5
Write and apply
With your answers, StoreSEO writes the description and saves it immediately; the response holds the only copy of the old text. The previewed FAQs and snippet are applied after your yes.
AI search readiness reference
| Scope | Products only; collections, pages and articles have no GEO score |
|---|---|
| Fixable by tools | FAQs, AI snippet, image alt text, product schema, description |
| Fixed only in the app or admin | Product title, category, product type, image count, inventory, variant barcode, SKU, price, image |
| Description pillars | Identity, materials, dimensions, features, use cases, what is included, care |
| Description write | Saved immediately to Shopify and StoreSEO; no undo and no revision history |
| Old description | Returned once in the write response, the only remaining copy |
| Product schema | Shop-wide toggle; credits the schema criterion on every product |
| Prerequisite | A focus keyword on the product, set in the StoreSEO app |
Rules Codex follows
- The scan's questions go to you word for word and in order; the agent does not answer them for you.
- A description is only written without questions when you decline to answer or authorize a bulk run without them.
- Writing a description replaces the current one at once with no undo, so the agent asks first and keeps the old text the response returns.
- If the agent enabled product schema for this flow and you stop before applying, it turns that schema back off; it never disables a schema you already had on.
- Criteria only you can fix get one link to the product's fix page, not a link per criterion.
In Codex
Set the storeseo server's default_tools_approval_mode to writes or prompt, not approve. Approve skips Codex's prompt, and then the confirmation StoreSEO asks for in the chat is the only check, which matters here because a description write has no undo. Codex waits 60 seconds for a tool by default and generation can take longer, so raise tool_timeout_sec for this server before you preview a long product, since each repeat of a preview spends AI credit again. A project-level .codex/config.toml only loads in trusted projects.
What runs without asking. Reads such as SEO scores, store listings, settings and Google reports run straight away. Generating a preview spends AI credit but writes nothing, and StoreSEO tells the agent to say what a generation will cost before it runs one. Applying content, optimizing images, changing settings, publishing llms.txt or agents.md, writing a product description and deleting anything wait for your yes in the conversation. That yes is an instruction the agent follows, not a lock on the server, so your client's own tool-approval setting is the hard stop.
Questions
Can Codex apply only some of the generated AI-readiness content?
Yes. The apply tool accepts a subset, so Codex can save the FAQs and skip the snippet or a URL handle. Skipped content stays applicable from the same draft until it expires one hour after the preview.
Why did my Codex preview time out for a product?
Codex waits 60 seconds for a tool by default and generation can take longer. Raise tool_timeout_sec for the storeseo server in config.toml, then run the preview again. A second preview spends AI credit again.
What is a GEO score in StoreSEO?
GEO is StoreSEO's AI readiness score for products. It measures how well a product can be understood by AI shopping agents and generative search, using criteria such as FAQs, an AI snippet, image alt text, product schema and the quality of the description.
Can I undo a rewritten product description?
Not inside StoreSEO. The new description is saved immediately and StoreSEO keeps no revision history. The response returns the previous description once, so ask the agent to keep it for you before it writes.
Why does the agent ask me questions before writing a description?
StoreSEO scans the description on seven pillars and returns questions for the facts it cannot infer, such as materials or what is in the box. Your answers become stated facts in the new copy instead of guesses.
Does enabling product schema affect every product?
Yes. Product JSON-LD is a shop-wide setting, so turning it on clears that criterion for every product at once, and turning it off removes it everywhere.
Keep going
AI search readiness with other agents
More Codex guides