The short answer: what AI changes in Shopify theme development

AI makes Shopify theme development faster, not automatic. At Vikn Studio, we use AI to build working concept prototypes, scaffold sections and blocks, and write settings schema boilerplate. Our developers own the architecture, brand decisions, accessibility, performance and final code review, and nothing goes live without human sign-off. For merchants, that means more creative directions to choose from and quicker revisions within the same budget, without trading away quality.

By Alec Ferriby, Founder of Vikn Studio.

How Shopify theme building has changed

Shopify themes are now modular systems that merchants edit visually, and AI tools are now built into both the editor and the developer’s workflow. Here is how we got here.

From Online Store 2.0 to theme blocks and Horizon

Online Store 2.0 replaced most hard-coded Liquid templates with JSON templates. That meant merchants could add, remove and reorder sections on almost every page, not just the homepage. Dawn, Shopify’s long-standing reference theme, showed developers how to build in that way.

Theme blocks went further. Blocks can now be defined once at theme level and reused across sections, and they can be nested inside other blocks. A product card block might contain a price block, a badge block and a swatch block, each with its own settings.

Horizon, launched in Shopify’s Summer ’25 Edition, is Shopify’s new theme foundation built around this block-based approach. It was the first public theme built entirely on theme blocks, and merchants can group and arrange blocks far more freely than in older themes.

Shopify’s built-in AI: Sidekick and AI-generated blocks

Merchants can now generate simple theme blocks themselves. In Horizon-based themes, you describe a block in the theme editor and Shopify’s AI generates it. Sidekick, Shopify’s AI assistant in the admin, and the wider Shopify Magic features cover tasks such as writing product descriptions and answering questions about your store.

These tools are useful for small additions, such as a promotional banner or an FAQ-style content block. They don’t know your wider design system, your catalogue structure or how your apps and integrations behave. Nobody is reviewing the output for performance or accessibility before it goes live, either.

AI coding assistants and the Shopify Dev MCP server

Developers use AI coding assistants inside their editors, and Shopify now offers a way to ground those assistants in its own documentation. The Shopify Dev MCP server connects assistants such as Claude, which is what we use, to Shopify’s docs and API schemas. Suggestions are then based on current documentation rather than the model’s memory.

This reduces errors, but it doesn’t remove them. That is why review still matters.

Can AI build a Shopify theme on its own?

AI can generate parts of a Shopify theme, such as sections, blocks and Liquid snippets. A production-ready theme still needs a developer to design the architecture, fix errors, and test performance and accessibility.

The same applies to the wider question of whether AI will replace Shopify developers. For custom builds, it won’t. AI handles repetitive code, but someone still has to make the judgement calls. Someone also has to take responsibility for performance, accessibility, integrations and what goes live on your store.

What AI does well

AI is reliable for work that is repetitive and well documented:

  • Boilerplate Liquid, such as loops over collections or conditional logic for sold-out states.
  • Section and block schemas, the JSON that defines which settings appear in the theme editor.
  • Repetitive markup, such as variations of a card or grid layout.
  • Test content, such as placeholder products, copy and metafield values for staging stores.
  • Quick visual prototypes that clients can click through in a browser.

Where AI-generated theme code goes wrong

AI-generated Liquid fails in predictable ways. These are the most common problems we see:

  • Invented Liquid filters and objects. The model writes {{ product.price | format_money }}, but no format_money filter exists. The correct filter is money.
  • Deprecated syntax. AI often uses {% include %}, which Shopify has deprecated, instead of {% render %}. It also tends to use the older img_url filter instead of image_url.
  • Accessibility gaps. Images are output without alt text, and icon-only buttons such as “close” or “next slide” are missing ARIA labels.
  • Heavy JavaScript. AI may pull in a full slider library and load it unminified in the <head>, which blocks rendering, when a few lines of deferred script would do.
  • Settings merchants can’t use. Schemas with labels like “Setting 1” or “Toggle”, and no help text, leave your team guessing in the editor.

Here is a typical example from an AI first draft, followed by our corrected version:

liquid
{%- comment -%} AI first draft {%- endcomment -%}
{% include 'product-card', product: product %}
<img src="{{ product.featured_image | img_url: '800x' }}">
<span>{{ product.price | format_money }}</span>
liquid
{%- comment -%} After developer review {%- endcomment -%}
{% render 'product-card', product: product %}
{{ product.featured_image | image_url: width: 800 | image_tag: loading: 'lazy', widths: '400, 600, 800' }}
<span>{{ product.price | money }}</span>

The corrected version uses supported syntax. It also serves responsive, lazy-loaded images and outputs the image’s alt text from Shopify admin. Your team still needs to fill that alt text in, which is why we cover it in training.

Settings schemas need the same care. A label that makes sense to the developer who wrote it often means nothing to a merchant:

json
{
  "type": "checkbox",
  "id": "show_vendor",
  "label": "Show brand name on product cards",
  "info": "Uses the Vendor field from each product.",
  "default": false
}

How Vikn Studio builds Shopify themes with AI (our workflow)

We use AI at specific stages where it saves time without affecting quality. Every other stage is led by people. These are the six stages of a typical build.

1. Discovery and brief (human-led)

We learn how your business works before any design or code starts.

We cover your brand, catalogue structure, variants and metafields, the apps you rely on, the markets you sell into and your commercial goals. We also look at your existing analytics and search performance. This stage sets the constraints that everything else has to respect.

2. Rapid creative concepting (AI-assisted)

You review working prototypes of several directions, not static mock-ups.

We use AI to produce several concept directions as clickable prototypes in the browser. You can scroll a homepage, open a product page and see how each direction feels on your phone. Our designers set the direction for each concept and edit the results. The AI simply makes it practical to show you more options in the same time.

3. Architecture and design system (human-led)

Our developers decide how the theme is structured, and we set a performance budget.

We plan which sections and blocks the theme needs, how they nest, and which settings each one exposes. We also define design tokens for colour, type and spacing. The performance budget caps JavaScript, fonts and third-party scripts. These decisions shape how fast and how editable your store will be for years, so we don’t hand them to a tool.

4. AI-assisted build (AI-assisted)

Developers use AI to scaffold sections and blocks, then refine every file by hand.

Using Claude, connected to the Shopify Dev MCP server, we generate first drafts of schemas, markup and repetitive Liquid. Developers then rewrite, trim and test that code against the architecture. All work is version-controlled through Shopify’s GitHub integration, so every change is tracked.

5. Human review and testing (human-led)

Every change is reviewed by a developer and tested before it reaches your live store.

Our checks include:

  • Code review of every pull request.
  • Shopify CLI’s shopify theme check for errors and deprecated syntax.
  • Lighthouse audits against our performance budget. We use Shopify Theme Store requirements as a minimum benchmark: an average Lighthouse score of at least 60 for performance and 90 for accessibility.
  • Automated accessibility testing, plus manual keyboard and screen-reader checks against WCAG 2.2 AA.
  • Testing across phones, tablets and desktop browsers.

6. Handover and editor training (human-led)

We make sure your team can run the store without calling a developer.

We walk your team through the theme editor, explain what each section and block does, and provide handover documentation to refer back to. We also explain which jobs the editor’s AI blocks are suitable for and when it’s better to ask us.

Want to see this process applied to your store? Start a Shopify project with our web development team in Hull.

What this means for merchants

Using AI in this way gives you more choice and faster feedback, while the checks that protect your store stay the same.

More concepts before you commit

Building prototypes is quicker, so trying a bolder idea costs less. You can compare a safe direction with an ambitious one side by side before committing your budget, rather than finding out halfway through the build that the safe option feels flat.

Faster feedback rounds

Scaffolding a revised section takes less time than writing it from scratch, so changes come back to you sooner.

What AI doesn’t make cheaper

Some costs stay the same because AI can’t do the work:

These are the parts of a build that protect your revenue, so we don’t cut corners on them.

Custom theme, premium theme or AI blocks: which is right for you?

A premium theme suits a new or smaller store on a tight budget. AI blocks suit small changes to an existing Horizon-based theme. A custom theme suits brands that need a distinctive experience, specific functionality or tight control over speed.

Custom themePremium themeAI-generated blocks
Upfront costFixed price, scoped upfrontTypically $100 to $500 on the Theme Store, one-offNo extra cost beyond your theme and Shopify plan
Time to launchSeveral weeks, depending on scopeDays to a few weeks, depending on setupMinutes per block
UniquenessBuilt around your brandShared with other stores using the same themeLimited to what the theme supports
Performance controlFull control, with a set budgetDepends on the theme and apps addedNot reviewed for performance
Editor flexibilityDesigned around your team’s needsBroad, generic settingsSingle blocks only
Ongoing maintenanceYou or your agencyTheme developer releases updatesYour team, with no developer review
Best forEstablished brands, complex catalogues, custom featuresNew or smaller storesSmall additions to an existing theme

Choose a premium theme if you’re launching or validating a store, your catalogue is straightforward and budget matters more than standing out. Note that Theme Store licences generally cover a single store.

Choose AI blocks if you already have a Horizon-based theme you’re happy with and need a small addition, such as a promotional banner or content block.

Choose a custom theme if your brand needs to look different from competitors, your catalogue or integrations are complex, or page speed directly affects your revenue.

Does an AI-assisted theme affect SEO, speed and AI visibility?

Unreviewed AI code can hurt SEO and page speed, because bloated scripts and poor markup slow pages and weaken Core Web Vitals. Once it has been reviewed and optimised, it performs the same as hand-written code.

We build against three Core Web Vitals:

  • LCP (Largest Contentful Paint): we keep hero images lean using Shopify’s image_url and image_tag filters with responsive widths.
  • INP (Interaction to Next Paint): we keep JavaScript light and deferred.
  • CLS (Cumulative Layout Shift): we reserve space for images and embedded app content so pages don’t jump as they load.

Clean, semantic markup and accurate structured data matter just as much. We build Product, Offer and BreadcrumbList structured data into the theme, so prices, availability and page hierarchy are clear to search engines.

Build quality also affects whether AI assistants cite your products. AI Overviews and ChatGPT shopping answers rely on content they can crawl and parse. Product details hidden behind heavy scripts, or inconsistent structured data, make your pages harder to understand and less likely to be referenced. Our AI visibility optimisation service builds on a well-built theme. Our guide to the metrics that matter for AI visibility explains how we measure it.

Two compliance points also belong in the theme build. Cookie consent needs to meet UK GDPR and PECR, which means non-essential scripts shouldn’t load before consent is given. If you sell into the EU, the European Accessibility Act has applied since June 2025, and we target WCAG 2.2 AA on every build. This is general guidance, not legal advice.

Questions to ask any Shopify agency that uses AI

Before you sign with any agency that uses AI, these questions will tell you how much checking actually happens:

  • Who reviews AI-generated code, and how many people check each change before it goes live?
  • Which automated checks do you run? Look for shopify theme check, Lighthouse and an accessibility tool as a minimum.
  • Which accessibility standard do you build to, and do you test manually as well as automatically?
  • Which AI tools do you use, and is my store data or are my credentials ever shared with them?
  • Who owns the finished code, and is it stored in a repository I can access?
  • How do you make sure the theme editor settings are usable by my team, not just by developers?
  • What happens if AI-generated code causes a problem after launch?

FAQs

Can AI build a Shopify theme?

AI can generate parts of a theme, such as sections, blocks and Liquid snippets. Horizon-based themes also let merchants generate blocks in the editor. A production-ready theme still needs a developer to design the architecture, fix errors, and test performance and accessibility.

Is AI-written Shopify code safe to use?

Only once a developer has reviewed it. AI tools can invent Liquid filters, use outdated syntax or add unnecessary scripts. At Vikn, every AI-assisted change goes through human code review and automated theme checks before it reaches a live store.

Does using AI make a custom Shopify theme cheaper or faster?

Mainly faster, in concepting and scaffolding. In practice, our clients get more creative directions and quicker revisions for the same budget, not a stripped-back build.

What are Shopify’s AI-generated theme blocks, and do I still need a developer?

Horizon-based themes let merchants describe a block and have AI generate it in the theme editor. That works well for small additions. A distinctive brand, complex catalogue or custom functionality still needs a developer.

Does AI-generated code hurt SEO or page speed?

It can if nobody checks it, because bloated scripts and poor markup slow pages and weaken Core Web Vitals. Once reviewed and optimised, it performs the same as hand-written code.

Who owns the code in an AI-assisted build?

You do. Every custom theme we build is handed over in full at the end of the project, along with ownership of the code. Premium themes are different, because Shopify Theme Store licences generally cover a single store.

Do you put client data into AI tools?

No. We use Claude to help write theme code, and that’s all it sees. Your store logins, customer details and order data never go into it.

How long does a custom Shopify theme take, and what does it cost in the UK?

It depends on your catalogue, integrations and how many custom features you need. Every build is fixed-price and scoped upfront, so you know the timeline and the full cost before any work starts.

Planning a Shopify build? Talk to us

Tell us about your store and we’ll show you what an AI-assisted, human-checked build could look like. We’re based in Hull and work with merchants across the UK.

Start a Shopify project to book a discovery call, or see our recent builds first. If your main concern is being found in search and AI assistants, read about our AI visibility optimisation service.