AI for Shopify SEO: Meta Titles, Descriptions and Schema

Key takeaways

• AI Shopify SEO works best as a drafting tool, not an autopilot. Claude writes the first pass, you review before publishing.
• Keep meta titles near 60 characters and descriptions near 155 characters. Google truncates longer ones to fit the device width.¹²
• Give Claude your real product data so it does not invent fields. Hallucinated JSON-LD values are the main risk with AI-generated schema.
• JSON-LD is Google’s recommended format for structured data. Place it in a <script type="application/ld+json"> tag in your theme.³
• Validate every schema block with the Rich Results Test before you ship.³

This guide covers an AI Shopify SEO workflow: using Claude to draft and refine meta titles and meta descriptions at scale, and to generate Product, FAQ, and BreadcrumbList structured data for your store. The manual, click-by-click versions live in separate guides. This one is about running the same work faster with an AI model, and doing it without introducing the mistakes AI tends to make.

The goal is speed with a human check at the end. Claude handles the repetitive drafting. You handle the review and the publish.

Why you can trust us

Jacques has over 15 years of development experience and has worked with hundreds of Shopify stores. We built Fudge - an AI-native Shopify store editor with a 5.0 rating and a Built for Shopify badge. We write SEO metadata and structured data into live themes as part of our daily work, so the advice here reflects what actually holds up in production.

Where AI fits in a Shopify SEO workflow

An AI model is good at three parts of on-page SEO:

1. Drafting meta titles and descriptions across many pages at once
2. Generating JSON-LD for Product, FAQ, and BreadcrumbList schema
3. Rewriting duplicates into unique variants

It is not good at deciding what is true about your store. That part stays with you. Claude will happily write a five-star rating into Product schema even if you have no reviews, because it is predicting plausible text, not reading your catalog. The whole workflow below is built around that limitation.

For the manual foundations, see our guides on how to edit meta titles in Shopify and how to edit meta descriptions in Shopify. This piece assumes you know where those fields live and want the AI-assisted version.

Drafting meta titles with Claude

A meta title is the clickable headline in Google’s results. Google does not publish a hard character limit, but it truncates the title link to fit the device width, so a practical target is around 60 characters.¹

Give Claude the raw product data and a tight brief. Vague prompts produce vague titles.

Example prompt:

Write a meta title for this Shopify product, under 60 characters.
Lead with the product name, then one specific detail. No brand suffix.

Product: Merino Wool Runner
Category: Men's running shoes
Key detail: machine washable, 4 colors

Ask for a batch when you have many pages:

Here are 20 products as a table (handle, title, key feature).
Write one meta title per product, each under 60 characters,
each unique. Return as a two-column table: handle, meta title.

Rules to put in every title prompt

• Under 60 characters, or Google may cut it off.¹
• Unique per page. Google warns against identical or boilerplate titles repeated across a site.¹
• No keyword stuffing. Repeating the same term does not help and can look spammy.¹
• Lead with the keyword or product name, not your brand.

Claude will often overshoot the character count. Ask it to report the count next to each title so you can spot the ones that run long.

Drafting meta descriptions at scale

A meta description is the snippet under the title. It does not directly affect rankings, but it affects whether people click. Google truncates descriptions to fit the device width, so aim for around 155 characters.²

Google’s own guidance: make each description unique, keep it relevant to the page, and avoid long strings of keywords that do not describe the content.²

Example prompt:

Write a meta description for this product page, under 155 characters.
Name one concrete benefit and one specific detail. No generic CTA
like "shop now" or "buy today". Include the product name once.

Product: Merino Wool Runner - machine washable, 4 colors, $95

Fixing duplicate descriptions

Duplicate descriptions are common on Shopify stores where the theme falls back to the same template text. This is a good AI task because it is high-volume and low-judgment.

These 15 collection pages share the same meta description.
Rewrite each into a unique description under 155 characters
that reflects the specific collection. Return handle + description.

Review the output for accuracy. Claude does not know your inventory, so it can describe a collection as having “hundreds of styles” when it has twelve. Cut anything you cannot verify.

Generating structured data (JSON-LD)

Structured data is code that tells Google what a page is about. Google recommends JSON-LD as the format, because it is the easiest to add and maintain at scale.³ You place it inside a <script type="application/ld+json"> tag, either in theme.liquid or in the relevant template.⁴

Before adding anything, check what your theme already outputs. Modern Shopify themes generate Product, Organization, and BreadcrumbList schema automatically. Adding a second copy creates duplicates. View a page’s source and search for application/ld+json to see what is already there. Our structured data in Shopify guide covers what Shopify includes by default.

Product schema

Claude can generate Product JSON-LD, but the values must come from you. Product structured data supports properties like name, description, image, price, currency, and availability.⁵

Give Claude the data, not a guess:

Generate Product JSON-LD for this product. Use ONLY the values I give you.
Do not add aggregateRating, review, or any field I have not provided.

name: Merino Wool Runner
price: 95.00
priceCurrency: USD
availability: InStock
sku: MWR-001

A minimal, honest result looks like this:

{
  "@context": "https://schema.org",
  "@type": "Product",
  "name": "Merino Wool Runner",
  "sku": "MWR-001",
  "offers": {
    "@type": "Offer",
    "price": "95.00",
    "priceCurrency": "USD",
    "availability": "https://schema.org/InStock"
  }
}

The instruction “do not add fields I have not provided” is the single most useful line in the prompt. Without it, models tend to pad schema with aggregateRating blocks and review counts that do not exist. Google’s guidelines say not to add structured data about information that is not visible to the user, even if it were accurate.³ Fabricated ratings break that rule outright.

BreadcrumbList schema

BreadcrumbList is still a supported rich result in Google Search.⁶ Each entry needs a position, a name, and an item URL. The last item in the trail can omit the URL.⁶

{
  "@context": "https://schema.org",
  "@type": "BreadcrumbList",
  "itemListElement": [
    {
      "@type": "ListItem",
      "position": 1,
      "name": "Home",
      "item": "https://example.com/"
    },
    {
      "@type": "ListItem",
      "position": 2,
      "name": "Men's Shoes",
      "item": "https://example.com/collections/mens-shoes"
    },
    {
      "@type": "ListItem",
      "position": 3,
      "name": "Merino Wool Runner"
    }
  ]
}

Ask Claude to build the trail from your real collection structure. Google recommends breadcrumbs that reflect a typical user path, not a raw copy of the URL.⁶

FAQ schema and a change worth knowing

You can still add FAQPage JSON-LD, and Claude can generate it. But be clear on what it does today. Google removed the FAQ rich result from Search in 2026, so FAQ markup no longer produces the expandable dropdowns it once did.⁷

FAQ structured data can still help other systems parse your content, and clear on-page FAQ text remains useful for search and for AI answer engines. If you are writing content with AI assistants in mind, see writing for AI search in Shopify and our take on answer engine optimization. Just do not expect FAQ schema to change how your listing looks in Google anymore.

Validate before you publish

Every schema block goes through the same check. Paste the page URL or the code into Google’s Rich Results Test. Google recommends the Rich Results Test during development and the Search Console rich result reports after deployment.³

A pass is not the finish line. The test confirms the syntax is valid. It does not confirm the values are true. A Product block with a fabricated rating can pass validation and still violate Google’s content guidelines. Read the values, then run the test.

The AI pitfalls to watch for

AI-generated SEO fails in predictable ways. Here is what to catch in review.

The common thread: an AI model produces confident, plausible text whether or not it matches your store. Treat every draft as a proposal that needs a human sign-off.

A repeatable review checklist

Run this on each batch before it goes live.

1. Character counts. Titles near 60, descriptions near 155.¹²
2. Uniqueness. No two titles or descriptions identical across pages.¹²
3. Accuracy. Every schema value maps to real store data.
4. No invented fields. No ratings, reviews, or IDs you did not supply.
5. Rich Results Test. Every schema block validates clean.³
6. Visible content. Schema reflects what a shopper can actually see.³

Where Fudge fits

The workflow above is manual prompting plus manual placement into theme files. That is fine for a developer. It is a lot for a merchandiser.

Fudge generates and edits storefront code directly in your theme from plain prompts, with drafts and previews before anything goes live. For SEO work specifically, that means you can propose metadata and schema changes, see them, and publish on your terms rather than pasting JSON-LD into theme.liquid by hand. If you are running a store rather than writing Liquid, the Shopify store editor is the surface built for that.

For a wider view on how AI-driven stores present to crawlers and assistants, see agent-ready Shopify storefronts.

Summary

AI Shopify SEO is a drafting accelerator. Claude writes meta titles under 60 characters, descriptions under 155, and clean Product, BreadcrumbList, and FAQ JSON-LD, at a speed no manual process matches. The value comes from pairing that speed with a strict review: real data in, no invented fields out, and a Rich Results Test pass on every block before you publish.

Keep the human at the end of the loop, and AI turns hours of metadata work into minutes without putting fabricated claims in front of Google.

FAQ