The argument in one line.
Feeding Claude Code a store's own keyword data and product catalog, then forbidding it from using any word not already on the site, turns AI-generated SEO pages into a fast, repeatable, indexable pipeline instead of a hallucination risk.
Read if. Skip if.
- You run or manage a Shopify (or similar) e-commerce store and want a repeatable way to turn keyword research into shippable pages, not just a list of ideas.
- You already use Claude Code and want to see connectors (Shopify, a keyword-research tool) chained together in one real, unscripted session.
- You're experimenting with making a store's catalog legible to AI shopping agents (metafields, llms.txt) and want a concrete first pass at what that looks like.
- You're looking for a clean, replicable script — this is a live, occasionally messy walkthrough with a heavy mid-video sponsor break, not a polished tutorial.
- You need technical depth on the specific MCP tool calls or Shopify's metafield schema — the video shows results on screen, not the underlying configuration.
The full version, fast.
This is a live screen recording of an SEO YouTuber using Claude Code, connected to both a real Shopify store and a keyword-research MCP tool, to build and publish new pages in real time. The core mechanism is a two-part loop: pull real keyword and product data into the conversation first, then instruct Claude to write only using words that already exist on the site — a guardrail against AI hallucination in commerce copy. In one session it drafts a new SEO guide page, builds a missing product collection, requests Google indexing for both, and — in a final parallel-agent pass — adds metafields and renames product images so the catalog is more legible to AI shopping agents, not just search engines.
Chat with this breakdown — free.
Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.
Create a free account →Where the time goes.

01 · A live AEO/GEO/SEO session
States the premise: a live, mostly unedited SEO session on a real Shopify client's store.

02 · The SEMrush MCP (sponsor)
Thanks SEMrush for sponsoring and explains he's used their MCP connector for months with real results.

03 · The setup: Claude Code + Shopify connector
Shows connecting Claude Code's Mac app to the Shopify store via the built-in connector.

04 · Model choice: Fable 5 on low
Picks the Fable 5 model on a lower reasoning setting for this kind of task.

05 · Step 1: export GSC data, find unbranded terms
Exports a Google Search Console CSV and asks Claude to mine it for unbranded keyword opportunities.

06 · SEMrush MCP: seed and secondary keywords
Switches to a higher reasoning setting and asks the SEMrush MCP tool for seed and secondary keywords for pages that don't exist yet.

07 · The proof: guide pages that grew another site
Points to guide pages this same technique already built on another client site as evidence it works.

08 · SEMrush One (sponsor)
A dedicated SEMrush One ad covering its combined SEO plus AI-visibility platform and LLM-ranking tracking.

09 · The keyword opportunities it found
Reviews the specific unbranded keyword opportunities Claude surfaced from the SEMrush data.

10 · Step 2: building the guide page
Prompts Claude to build a full SEO guide page under one hard rule: only words already on the site, no embellishment, plus schema, images, and embedded collections.

11 · More opportunities: no rings collection exists
Notices the store sells individual rings but has no dedicated rings collection page.

12 · Step 3: the smart collection
Builds a rule-based Rings smart collection (title contains ring) packed with keywords and JSON schema.

13 · Previewing the AI-built page
Previews the finished guide page live and reacts to how thorough the generated SEO content is.

14 · Why he stopped making SEO videos
Explains SEO content had been underperforming on his channel until recently, and why he's bringing it back.

15 · Nothing beats Claude Code for pages like this
Declares Claude Code the best tool he's found for producing this specific kind of SEO page.

16 · "Five minutes' work could be 10,000 clicks a month"
States the headline ROI claim of the video, tied to the keyword's measured click potential.

17 · Step 4: request indexing
Submits the new guide page to Google Search Console for indexing and it's confirmed indexed within seconds.

18 · Real talk: Shopify is a seriously good CMS
A brief aside praising Shopify as a platform for a serious e-commerce build.

19 · The collection page and its on-page SEO
Builds the rings collection in Shopify admin and tries, then partly abandons, moving SEO description text below the product grid via the theme editor.

20 · Indexing page two
Requests indexing for the collection page and confirms the first guide page is already indexed.

21 · Step 5: metafields, llms.txt, Shopify UCP
Prompts Claude to add metafields to every product and set up llms.txt plus Shopify's UCP system so AI shopping agents can read the store.

22 · Step 6: renaming every image with keywords
Has Claude rename every product image file to include target keywords.

23 · Two Claude agents working in parallel
Runs two Claude Code agents simultaneously, one on metafields and one renaming images, since neither depends on the other's output.

24 · Want more of this? Tell me below
Wraps up, confirms the collection page got indexed, and asks viewers to comment for a ranking follow-up.
Lines worth screenshotting.
- Feeding an AI a store's exported keyword and product data before asking for content ideas anchors every suggestion to real search demand instead of guesswork.
- The single guardrail that keeps AI-generated commerce copy honest is simple: forbid the model from using any word that isn't already somewhere on the site.
- Asking for 'full SEO' on a page really means bundling four things into one prompt: keyword coverage, embedded collections, images, and JSON-LD schema.
- A store can have every individual product in a category yet still lack a collection page for that category — a rule-based smart collection fixes it in minutes.
- New pages can be requested for indexing straight from Google Search Console the moment they publish, rather than waiting for a natural crawl.
- A freshly indexed page showed up as fully indexed in Search Console roughly 30 seconds after the indexing request was submitted.
- Not every on-page text placement is editable from a theme's visual editor — moving collection description text below a product grid required writing custom theme code.
- Making a store legible to AI shopping agents now means structured product metafields plus a dedicated llms.txt file, not just traditional SEO copy.
- Running two Claude Code agents in parallel works cleanly when the tasks are independent — one on metafields, one renaming product images, neither waiting on the other.
- Descriptive, keyword-bearing image filenames are still a real lever precisely because almost no store owner bothers to set them.
- Checking whether an existing page already covers a topic before building a new one avoids the new page cannibalizing traffic from the old one.
The guardrail matters more than the prompt.
Wiring Claude Code to a store's real keyword and product data turns AI content generation from a hallucination risk into a repeatable pipeline, if you force it to use only the site's own words.
- Connecting Claude Code to a live Shopify store is a one-time setup: add the Shopify connector, then ask the model to confirm it's actually talking to your store before doing anything else.
- The workflow starts from an exported Search Console CSV of real keyword and click data, not a generic content prompt — the AI needs actual demand signals before it can find opportunities.
- The first ask isn't 'write content' — it's 'find unbranded terms with real click potential' — every later page idea gets anchored to measured search demand.
- A keyword-research tool exposed as an MCP connector lets Claude pull live seed and secondary keyword data mid-conversation instead of you copy-pasting spreadsheets back and forth.
- The single load-bearing instruction in the whole process is a guardrail: never use a word that isn't already on the site, and don't embellish — that's what keeps AI-generated commerce copy from becoming a hallucinated product claim.
- 'Fully SEO this page' is shorthand for bundling four things into one prompt: keyword coverage, embedded collections, images, and JSON-LD schema.
- A store can have every individual product in a category yet still lack a collection page for that category — that gap is a low-effort, high-upside page to build.
- A Shopify smart collection is just a rule, such as 'if product title contains ring' — no manual curation needed to group an entire product category.
- The 'five minutes of work could be 10,000 clicks a month' claim isn't a guess — it's tied to click-potential numbers already pulled from the keyword tool earlier in the session.
- A newly published page can be submitted for indexing directly from Google Search Console the moment it's live, and verified as indexed within roughly 30 seconds — check, don't assume.
- Not every on-page text placement is editable from a theme's visual editor — moving description text below a product grid required custom theme code, not a settings toggle.
- Making a store legible to AI shopping agents is now more than SEO copy: it means structured metafields on every product plus a dedicated llms.txt file for how agentic systems read the catalog.
- Two Claude Code agents ran in parallel — one filling in metafields, the other renaming every product image with target keywords — because neither task depended on the other's output.
Terms worth knowing.
- MCP (Model Context Protocol)
- A standard that lets an AI model like Claude call out to external tools and data sources — in this video, a Shopify store and a keyword-research database — directly inside a conversation.
- AEO / GEO
- Answer Engine Optimization and Generative Engine Optimization — optimizing content to be surfaced and cited by AI answer engines and chatbots, not just ranked in traditional search results.
- Smart collection (Shopify)
- A Shopify product grouping built from an automatic rule, such as 'title contains ring', instead of being manually curated product-by-product.
- Metafields
- Custom structured data fields attached to a Shopify product beyond the default title, price, and description, used to make product attributes machine-readable.
- llms.txt
- A plain-text file published at a site's root that describes its content in a format intended to be easily read by AI language models and shopping agents.
- JSON-LD schema
- Structured data embedded in a page's code that explicitly tells search engines and AI systems what the content represents, such as a product, article, or FAQ.
Things they pointed at.
Lines you could clip.
“You know me as a YouTuber. I don't really take sponsors. I'm really bad with sponsors.”
“You cannot use any words that you do not find on the site itself. Be extremely diligent.”
“This is the kind of content that will rank so, so well on Google. I don't think there's a better way to make these pages.”
“Five minutes' work could be 10,000 clicks a month.”
Word for word.
Don't just watch it. Burn it in.
See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.
The bait, then the rug-pull.
A screen-recorded, unscripted session where an SEO YouTuber connects Claude Code to a live Shopify store and a keyword-research tool, then watches it draft, publish, and index new pages before the video is even half over.
Named ideas worth stealing.
Site-only content guardrail
An explicit instruction given to the model every time it drafts new customer-facing page copy: do not use any word that isn't already on the site, and do not embellish or invent product details.
How they asked for the click.
“Go and check them out, guys. They have made a really, really good tool, and it's not too expensive either... SEMrush is definitely the best option and choice.”
A roughly 80-second dedicated pitch inside a longer ~3-minute sponsor block (03:15-04:38), read over static SEMrush promotional slides, with an explicit 14-day free-trial link in the description and pinned comment.
























































