Should I build a headless Shopify or Liquid?

For this guide we also read 16,914 Shopify Community threads across the checkout, Liquid, migration, Plus, metafields, Flow, inventory and sales-channel forums, and pulled the 257 that discuss headless, Hydrogen, Oxygen or the Storefront API. That data appears throughout, because it shows what merchants actually get stuck on rather than what agencies say they get stuck on.

What each approach actually is

Liquid and the online store

Liquid is Shopify's server-side templating language. Your storefront runs on Shopify's own infrastructure, is edited through the theme editor, and renders complete HTML before it reaches the browser. Since Online Store 2.0 it has supported JSON templates, sections everywhere, metafields, metaobjects and, since 2024, theme blocks: reusable, nestable components defined once at theme level and dropped anywhere.

A "custom Liquid build" in 2026 does not mean editing a bought theme. It means a bespoke theme built from Shopify's Skeleton starting point with your own component system, which is what we do on most Shopify builds.

Custom and headless storefronts

A custom storefront replaces that frontend with your own application. It pulls product, collection and cart data from the Storefront API and customer data from the Customer Account API, and you host and deploy it yourself. Shopify's admin, checkout, payments and order management stay exactly where they are.

The main routes:

Route

What it is

Hosting

Hydrogen

Shopify's own stack, historically React Router. Rebuilt in June 2026 with Vercel as a framework-agnostic SDK, currently in developer preview

Oxygen, free on paid plans

Next.js, Nuxt, Astro, SvelteKit

Bring your own framework against the Storefront API

Vercel, Netlify, your own infrastructure

Headless plus a CMS

Any of the above with Sanity, Contentstack, Storyblok or Contentful for editorial

As above, plus CMS subscription

Storefront Web Components

One script tag that embeds Shopify commerce into an existing site

Wherever that site already lives

If the choice of CMS is the real question behind your architecture decision, start there instead.

The middle ground is now real

The framing that dominates search results is binary, and it stopped being accurate. You can run a Liquid theme and render one complex surface from your own stack. You can embed Shopify commerce into an existing WordPress or Next.js site with Storefront Web Components. You can go headless on one market and leave the rest on the theme. Most brands we talk to should be looking at these before they scope a full rebuild.

What changes, and what does not

A lot of the confusion in this debate comes from people arguing about things that are not on the table.

Stays the same either way: checkout stays hosted by Shopify (you get a checkoutUrl, not the DOM), Shop Pay and your payment providers work, PCI compliance is inherited from Shopify, the admin is unchanged, and orders, inventory, discounts and Flow all behave identically.

Changes when you go custom: you own the rendering layer, the routing, the caching, the deploy pipeline, the SEO plumbing and every third-party integration that used to be a one-click app install. You also lose the theme editor.

That last point is the one that gets underestimated. Shopify's own developer docs put the condition plainly: before committing, be sure the merchant is comfortable with "the added costs and complexity of managing a custom storefront solution" and has development resources for ongoing integration work after launch.

What merchants actually get stuck on

Of the 257 headless-related community threads we analysed, the topics that come up most are not the ones in the sales deck:

Theme

Share of headless threads

Checkout behaviour and the checkout domain

42%

Domains, hosting and deployment (404s, DNS, Oxygen)

18%

Customer accounts and authentication

14%

SEO, sitemaps, canonicals and redirects

11%

Tracking, pixels and attribution

8%

Content editing and CMS

6%

Almost nobody is posting about design freedom. They are posting about plumbing that used to be free.

[QUOTE SLOT: Arthur] One or two sentences on how this decision usually arrives at 6th Man. Suggested prompt: "When a founder calls us asking for a headless build, what are they actually trying to fix, and how often is headless the right fix?"

The 2026 shift nobody has priced in: Liquid moved

Most comparison articles still run on 2023 logic, where Liquid was a templating language for brochure-style stores and Hydrogen was the escape hatch. That gap has narrowed sharply.

Since 2024, Liquid has picked up theme blocks, the Horizon theme family built entirely on them, and native page-level composition. In 2026 Shopify shipped standard storefront events and actions, which kills the DOM-scraping hacks that made theme apps fragile. In July 2026 it opened a developer preview of {% block %} and {% partial %} tags: named regions of server-rendered HTML that JavaScript can refresh without reloading the page. That is islands-style interactivity inside Liquid, and it removes one of the classic reasons to leave.

March 2026 added Rollouts, native scheduling and A/B testing for themes, available on Basic plans and above. We covered the wider picture in our breakdown of the Shopify Winter '26 update.

The agentic-commerce layer landed with no work on your side. Shopify's WebMCP tools (catalogue search, cart, checkout, order and policy lookup for AI agents) went live on 5 August 2026 on every Liquid storefront and on the Hydrogen developer preview, with, in Shopify's words, "nothing to install or configure". On a bring-your-own-stack build, you implement it yourself.

Hydrogen, meanwhile, has been a moving target. It moved from Remix to React Router 7, then in June 2026 was rebuilt again with Vercel as a framework-agnostic SDK rather than a framework, currently in developer preview. That rebuild is genuinely good news for lock-in, since the roadmap covers Next.js, Nuxt and Svelte rather than React Router alone, though only React is available on the preview branch today. It also means anyone quoting a fixed-price Hydrogen build right now is quoting against a spec that has changed twice.

[QUOTE SLOT: Wout] Suggested prompt: "Hydrogen has changed architecture twice in a year. What does that actually cost a client who launched on the old version, and how do you decide what to build on today?"

Speed: the number everyone quotes is three years old

Almost every article on this topic cites the same study. Shopify published it in October 2023: 59.5% of Liquid origins passed all Core Web Vitals, against 35% for Hydrogen, 28% to 29% for Next.js and Remix, and 22% for Nuxt. It has never been refreshed, it was written by the company that sells Liquid themes, and it lumps the entire non-commerce Next.js web into the comparison. Treat it as directional, not decisive.

The newer, more defensible numbers:

HTTP Archive Web Almanac 2025 (July 2025 crawl, published January 2026): 76% of Shopify origins pass all Core Web Vitals on mobile, against 48% for the web overall and 35% for WooCommerce.

Shopify's live theme data (7 days to 4 August 2026): the pass rate ranges from 34% to 98% depending on which theme you run. Dawn sits at 86.2%, Horizon at 83.6%, median 85.7%.

A 1,166-store audit (April 2026, 1,086 with usable Lighthouse data) found an average TTFB of 53ms while 45% of all requests went to third-party origins, with 120+ script resources per store.

What those three numbers mean together

Read them together and the conclusion is uncomfortable for both camps. Shopify's server layer is not the bottleneck. The variance is almost entirely in theme quality and app payload, and a 64-point swing within one architecture is far larger than any measured gap between architectures.

The practical implication: going headless does not delete your apps. Merchants re-add Klaviyo, GA4, reviews and chat to the new frontend, and the scripts follow you. If your Lighthouse score is 30, a rebuild will not fix it. A script audit might, which is where our ecommerce SEO checklist starts, and it is usually a faster route to a better conversion rate than a replatform.

For reference, the current Core Web Vitals thresholds are LCP under 2.5s, INP under 200ms and CLS under 0.1, measured at the 75th percentile of real users. INP replaced FID in March 2024 and the metric set has not changed since.

The honest counterpoint, from an agency that builds both: "Liquid performs well by default with a higher floor. Headless has a higher ceiling, but reaching it requires deliberate engineering effort."

SEO: what you inherit versus what you rebuild

For Google, JavaScript rendering is close to settled. A study of over 100,000 Googlebot fetches found that 100% of HTML pages resulted in full-page renders. The cost is latency: a median of 10 seconds between crawl and render, with the 90th percentile around three hours. Survivable, but not free if your prices and stock move.

The bigger headless SEO tax is administrative. On Liquid you get the following automatically, generated from your store and Markets configuration:

Signal

Liquid

Custom storefront

sitemap.xml

Automatic, always current

You own generation, pagination and the 50,000-URL split

robots.txt

Automatic, overridable via robots.txt.liquid

You own it, and staging leaks are the classic failure

Canonical tags

Emitted on every template

Configured per route, easy to miss one

hreflang

Auto-generated from Markets

Fully manual. Hydrogen's SEO docs do not cover it

Per-market sitemaps

Automatic, with hreflang annotations

Yours to build

URL redirects

Native redirect table in admin

Middleware or edge config, and 301 chains are the common bug

Image dimensions

Set on every product image, which is why Shopify leads on CLS

Yours to enforce

Server logs

Not available

Available, which is a genuine headless win for crawl budget analysis

None of this is impossible. All of it is work that was previously free, and it is the part of the quote that gets compressed when a project runs late.

Shopify's URL constraints apply either way

Both routes inherit Shopify's mandatory /products/, /collections/ and /pages/ prefixes if you keep the standard structure, and one product can be reachable at several collection-scoped URLs. Shopify canonicalises those for you on Liquid. Go custom and you must reproduce that logic, or you generate the duplicates without the canonicals, which is how keyword cannibalisation and indexability problems start. The same discipline applies to faceted navigation, which we covered in should filter pages be indexed.

If you are moving between architectures, the redirect map is the whole game. Our guide to technical SEO migration covers redirects, canonicals and crawl budget, and it is the same discipline we use on every website migration, whether that is from WooCommerce, from PrestaShop or from Tilroy. We have moved 170,000+ variants without losing rankings. The full approach is on our replatforming page.

AI search: the argument that actually changed in 2026

This is the one genuinely new factor since the last generation of articles on this topic.

Non-Google AI crawlers do not execute JavaScript. Vercel's analysis of AI crawler traffic found that "none of the major AI crawlers currently render JavaScript", naming OpenAI's GPTBot, OAI-SearchBot and ChatGPT-User, Anthropic's ClaudeBot, and PerplexityBot. They fetch JavaScript files without running them. Google's Gemini, by contrast, uses Googlebot's rendering infrastructure and does render.

Surface

Renders JavaScript?

Googlebot, AI Overviews, AI Mode, Gemini

Yes

GPTBot and ChatGPT

No

ClaudeBot

No

PerplexityBot

No

So a client-side-rendered custom storefront is fully visible to Google and structurally invisible to ChatGPT, Claude and Perplexity. If being cited by an LLM is part of your 2026 plan, that matters more than it did last year.

This is not an argument against custom storefronts. It is an argument against client-side rendering. Server-render your product data and you are fine. Liquid just happens to be server-rendered by construction, so it never has this problem.

Two things help regardless of architecture: server-rendered content, and clean structured data on product and collection pages.

Do this test today: disable JavaScript, load a product page, and check whether price, availability and reviews survive. That is roughly what an AI crawler sees. If they vanish, you have an AEO problem regardless of which architecture you are on.

What your marketing team gains or loses

This is where the decision is usually made, and it is barely covered anywhere else.

Task

Liquid theme

Custom storefront

Build a landing page

Theme editor, minutes, no developer

Only if the CMS schema and matching component already exist. New layout is a dev ticket.

Swap hero copy or imagery

Theme editor

CMS field if modelled, otherwise a developer

Add a reviews or upsell widget

App block, no code

Custom API integration

Schedule a sale theme with auto-revert

Native Rollouts (Basic plan and up)

Build it yourself

A/B test or multivariate test a page

Native Rollouts

Rollouts is theme-based, so it does not apply. Intelligems has an official headless SDK; most alternatives need custom work.

Repeatable content pages (stores, guides, ambassadors)

Metaobjects plus dynamic sources, no developer

CMS document type plus a frontend renderer

Shopify announced a Hydrogen visual editor at Summer '24 Editions. Two years later there is still no public release. Do not budget as if one is arriving.

A practitioner thread on Shopify's community forum in August 2026 put it well: it is not that Hydrogen cannot do something, it is that "you end up building the layer that makes it practical for the merchant to do it themselves."

Attribution is the quiet killer

The performance-marketing side is worse than most teams expect. Because checkout sits on a different domain from a custom storefront, attribution breaks by default.

One merchant running three Next.js storefronts reported every paid purchase landing in GA4 as direct/none, adding that Shopify Support could not help. Another in the same thread rebuilt tracking by hand with GTM plus an orders/create webhook. A media brand running Hydrogen and Sanity across 75,000 pieces of editorial content described exactly this in February 2025: "our headless site is creating challenges with ad-tracking. We spend most of our ad budget on Meta, and we don't get the right data back."

All of it is fixable, usually with server-side tracking and a checkout domain that matches your primary domain. None of it is free, and it is not usually in the quote.

[QUOTE SLOT: Sjonlee] Suggested prompt: "From a design point of view, what can you actually build in Horizon and theme blocks in 2026 that you could not two years ago, and where do you still hit a wall that only a custom frontend solves?"

The team you need to run it

Liquid

Custom storefront

Core skills

Liquid, HTML/CSS, vanilla JS, theme architecture

React/TypeScript, a framework, GraphQL

Data

Metafields and metaobjects, configured in admin

Storefront API, Customer Account API (OAuth), caching strategy

Infrastructure

None

Edge runtime, CI/CD, secrets, observability

Maintenance

Shopify ships platform changes

You own framework upgrades, CVEs, API version bumps

Handover risk

Any Shopify partner can pick it up

Requires fluency in three systems at once

The hiring-pool objection cuts both ways. In the UK, ITJobsWatch showed 1,866 permanent vacancies mentioning React over six months, against 11 for "Shopify Developer". React talent is not scarce. React talent that also understands Shopify commerce semantics is, and that intersection is who you are actually hiring.

Two platform changes worth planning for: legacy customer accounts were deprecated in February 2026, so older custom builds have auth rework coming, and a final sunset date is still to be announced.

The other question nobody asks in the pitch is what happens when the agency leaves. A Liquid theme can be downloaded from the admin and picked up by any Shopify partner. A custom storefront is a bespoke application plus a CMS schema plus a deploy pipeline plus N custom integrations. Ask about code ownership in writing before you sign. On our builds it transfers on final payment, and we think that should be the norm rather than a differentiator.

What it costs, over three years rather than one

Benchmarks, with the caveat that most published figures come from agencies with something to sell.

Build

In the Benelux, a professional custom Liquid theme runs roughly €10,000 to €25,000 in our experience, with replatforming projects at €15,000 to €75,000 and up. Our own builds typically run 8 to 14 weeks.

Published headless build ranges sit at $10,000 to $25,000 for a lean build, $25,000 to $60,000 for a growth build, and $60,000 to $150,000+ at scale, with agency-quoted Hydrogen and Next.js projects reaching $250,000 to $300,000.

Belgian rates for context: roughly €40 to €85 per hour freelance, €75 to €120 at a small agency, €120 to €200+ at a large one.

Run

Shopify plans in 2026: Basic $39/month, Grow $105, Advanced $399, and Plus from $2,300 on a three-year term.

Hosting is where the intuition misleads people. Liquid hosting is included, and Oxygen is free on paid plans too (Starter through Plus, though not on Agentic plans). The incremental cost of headless is not hosting. It is engineering.

Theme retainers in the Benelux run €750 to €3,000 per month. Published headless maintenance models start at $4,000 per month, and agencies with a headless practice openly say not to go headless if you cannot commit $5,000 to $25,000 per month to it.

The parts nobody quotes

A headless CMS. Sanity, Contentful and Storyblok all put SSO and a contractual SLA behind Enterprise pricing, so "€15 per seat" is not the real number if your IT policy requires SSO.

Monitoring and observability.

Accessibility re-audits after each framework migration.

Rebuilding analytics and attribution.

$7,000 to $25,000 to rebuild what apps used to do for you.

Engineering time for undocumented behaviour. Shopify's docs state there are no rate limits on the number of Storefront API requests, yet merchants hit THROTTLED responses on cartCreate and had to add exponential backoff. Third-party apps are frequently hardcoded to expect the Online Store channel and will error or refuse to install without it, which is why experienced builders keep a password-protected Liquid theme alive alongside the custom storefront rather than deleting the channel.

One worked year-one model: a $50,000 build totals $114,948 in the first twelve months once you add the plan, CMS, tooling, engineering and contingency. Multiply the recurring lines by three and compare that against a custom theme plus a retainer before you decide.

[QUOTE SLOT: Arthur] Suggested prompt: "What is the honest three-year number a €5M brand should budget for each route, and where do the surprises usually land?"

Two EU-specific factors most articles skip

Markets, hreflang and multi-currency

Shopify Markets generates hreflang tags, per-market canonicals and per-market sitemaps automatically, but only for markets with a distinct domain, subdomain or subfolder. Sell into DE, FR and NL on one undifferentiated URL and you get no hreflang at all, with no error shown.

Subfolders consolidate authority and are the lowest-complexity option in Shopify's own ranking. Go custom and every bit of this becomes engineering you own, including the country selector and locale routing.

Two Shopify quirks that survive either way and catch out EU stores: policy page URLs stay in English across markets, and product handles do not translate unless you manage them deliberately.

The European Accessibility Act

In force since 28 June 2025, with e-commerce explicitly in scope. The technical standard is EN 301 549, which incorporates WCAG 2.1/2.2 Level AA.

The micro-enterprise exemption for services applies to businesses with fewer than 10 people and either turnover or a balance sheet total under €2M, so once you are past €2M or ten staff you are in scope. Belgium is stricter still: reporting on the Belgian transposition says micro-enterprises selling online are not exempt but get a transitional period to 2030. Enforcement runs through the FPS Economy's Economic Inspection via the ConsumerConnect portal, with a 30-day response obligation on complaints. Published penalty ranges conflict badly across sources, so treat the ceiling as a question for counsel rather than a blog post.

The architectural point. Shopify's Theme Store requirements cover only an estimated 16% to 22% of WCAG 2.2 AA criteria, so Liquid is a partial baseline you inherit and top up. A custom storefront gives you a 0% baseline. Every criterion is yours to build and re-verify after each framework migration. In the EU this is now a legal obligation, not a nice-to-have, and it belongs as a line item in any custom-storefront quote.

The middle ground most brands should look at first

Liquid with headless sections. Keep the theme, render one complex surface (a configurator, a store locator, a B2B portal) from your own stack.

Storefront Web Components. A single script tag puts Shopify commerce on any existing site. Note the gaps: search and filter components are not available, and the components are not covered by Shopify Support.

The Shopify plugin for WordPress. A first-party route if your content team already lives in WordPress.

One headless sub-storefront. Prove the case on a single market or brand line before rebuilding everything.

Each of these buys you most of the flexibility for a fraction of the ownership cost, and each is reversible in a way a full replatform is not.

Score your own store

Give yourself a point for each true statement.

Points for Liquid

Marketing publishes pages weekly without a developer.

You run six or more apps, including any of reviews, subscriptions, loyalty or bundles.

You sell into multiple EU markets and rely on Shopify Markets for hreflang and pricing.

You cannot commit a permanent frontend engineer or an equivalent retainer.

Your product page is images, specs, variants and a buy button.

Accessibility compliance is a board-level risk you want a maintained baseline for.

Points for a custom storefront

You already have a non-Liquid frontend or a corporate site you must integrate with.

Your product experience genuinely cannot be expressed in theme blocks (configurators, complex made-to-order, heavy real-time logic).

You have in-house engineering that will still be there in three years.

Shopify is one of several commerce backends you serve.

Editorial and content are the product, and you already run a headless CMS.

You need server logs and full control of the rendering and caching layer.

Four or more on the Liquid side and the answer is almost certainly a better Liquid build. Four or more on the custom side and it is worth scoping properly. A split score usually points to the hybrid route.

On reversibility

Brands publish "we went headless" case studies constantly, and almost nobody publishes "we went back". That absence is not evidence that reversals do not happen, it is evidence that nobody wants to write that post. They do show up in the forums, just without a logo attached. From April 2024: "My current website design is headless and we are facing multiple issues getting channels tracking to work properly. So we thought about moving back to our old template which was a native Shopify design."

The reversal is technically straightforward, since your catalogue, orders and customers never left Shopify. You are simply paying for a second frontend. Ask any agency what the reversal costs before you commit, and treat reversibility as a decision criterion in its own right.

Questions to ask before you sign

Who owns the code, and when does it transfer? Get it in writing.

What happens to hreflang, canonicals, sitemaps and redirects? Ask for the specific plan, not reassurance.

Which of my current apps survive, and what replaces the rest? Ask for the list, with a cost per replacement.

How will my marketing team publish a landing page in month four? If the answer involves a ticket, price the ticket.

How does attribution work when checkout is on a different domain? If the answer is "we'll set up GTM", ask about server-side.

Who does WCAG 2.2 AA conformance, and who re-tests it after each framework upgrade?

What is the three-year cost, not the build cost?

What does going back look like?

If you want a second opinion on the answers you get, that is what a storefront architecture audit is for.

[QUOTE SLOT: Wout or Arthur] Closing line. Suggested prompt: "What is the one question you would ask a brand before letting them sign off on a custom storefront?"

Method

Community figures in this article come from an analysis of 16,914 topics scraped from eight Shopify Community categories (checkout, Liquid, migration, Plus, metafields, Flow, inventory, sales channels), covering threads from 2019 to August 2026. Of those, 257 topics reference headless, Hydrogen, Oxygen, custom storefronts or the Storefront API. Theme percentages are keyword-classified per topic and overlap, so they do not sum to 100%. Quoted merchant comments are reproduced verbatim and linked inline.

Cost figures marked "in our experience" come from 6th Man project data. All other figures are linked to their primary source at the point of use.

About the author

Arthur Lauwers is the founder of 6th Man, a Belgian agency that builds custom Shopify storefronts and runs SEO and CRO on the same stores. The team has migrated catalogues of over 170,000 variants to Shopify without losing rankings, including T'Hooft Buitengewoon, a 100,000 SKU move from Tilroy to a fully custom Liquid theme with three custom Shopify apps, and the TROC Leuven redesign for second-hand design furniture. More of our work is here.

More field notes, playbooks and definitions are in Learn.

If you are weighing this decision, we will give you a straight answer on which route fits your store, including when the answer is that you do not need us.

Sources

Shopify primary Enterprise blog: headless commerce · Custom storefronts: getting started · Hydrogen and Oxygen fundamentals · Storefront API rate limits · Theme blocks · Standard storefront events and actions · Storefront Web Components · WebMCP on Liquid and Hydrogen · Liquid block and partial developer preview · Legacy customer accounts deprecated · Rollouts · Horizon themes · Markets SEO · Metaobjects · PCI compliance · Pricing · Plus pricing

Performance and rendering Shopify theme performance data · Shopify: Liquid vs headless real user performance, Oct 2023 · HTTP Archive Web Almanac 2025: ecommerce · Web Almanac 2025: performance · web.dev: Core Web Vitals · Hyperspeed: why the average PageSpeed score is 30 · Vercel and MERJ: how Google handles JavaScript, Jul 2024 · Vercel: the rise of the AI crawler · Vercel and Shopify: rebuilding Hydrogen

Accessibility and EU law European Commission: European Accessibility Act · Directive (EU) 2019/882 · Craftzing: EAA in Belgium · Converge Accessibility: accessible Shopify themes

Costs, apps and ecosystem Weaverse: Shopify headless pricing · Roboto Studio: Shopify headless guide · Commerce-UI: the state of Shopify, headless vs Liquid · Klaviyo: integrate with a Shopify Hydrogen store · Recharge API · Sanity pricing · Contentful pricing · Storyblok pricing · ITJobsWatch: React · ITJobsWatch: Shopify Developer

Shopify Community threads Headless vs theme for content-heavy site · Moving a headless storefront back to a Shopify template · Cross-domain tracking for a headless Shopify store · cartCreate returning THROTTLED despite no documented limits · Hydrogen sales channel: do I keep the online store? · Hydrogen and Shopify Live View · Custom domain 404 on Hydrogen · Custom checkout with headless Shopify and Next.js · Devs running Hydrogen in production: what's still missing in 2026? · Hydrogen Visual Editor

Related reading

Which CMS best boosts your SEO: WordPress, Shopify or headless

Technical SEO migration: redirects, canonicals and crawl budget made simple

10 things that surprised me in Shopify's Winter '26 update

Ecommerce SEO checklist: speed, schema and content fixes

How does server-side tracking work?

Should you index filter pages in e-commerce?

How to implement a post-purchase survey in Shopify

All articles and the full glossary live in Learn

Replatforming from something else? We have specific guides for moving from WooCommerce, from PrestaShop and from Tilroy.