Horizon, or stay on Dawn.

What Horizon actually is

Horizon arrived with Shopify's Summer 2025 Editions as the new default theme for stores created from then on, ending Dawn's run as the theme every new store started life in. It shipped alongside a family of sibling free themes built on the same foundation, each styled for a different kind of catalogue, so "Horizon" in practice means both one theme and the architecture underneath a group of them.

The marketing framing is about AI assistance in the editor. That is the least interesting part. The part that matters to anyone maintaining a store is that Shopify now recognises three theme generations, and Horizon is the first theme of the third one.

• Vintage. The original architecture. Sections on the home page only, no dynamic sources. Shopify's free vintage themes receive security fixes and nothing else.
• Online Store 2.0. Dawn's generation. Sections on every page, custom templates, app blocks, metafield dynamic sources, native filtering and predictive search.
• Theme blocks. Horizon's generation. Everything Online Store 2.0 has, plus blocks that can contain other blocks.

If you are still on a vintage theme, this guide is not your decision. Yours is covered in signs your store needs a rebuild, and the answer is almost always yes.

The one architectural change that matters

In an Online Store 2.0 theme, a page is a stack of sections. Each section is a self-contained unit of Liquid with its own settings, and it can hold blocks, but those blocks are flat: a slide, a menu item, a text field. You cannot put a block inside a block. When a designer asks for a two-column layout where one column holds an image and the other holds a stacked heading, body and button group, a merchant on Dawn cannot build that. A developer writes a new section for it.

That single limit is responsible for a large share of the custom section work agencies have billed for since 2021, and for the popularity of page builder apps, which exist mostly to route around it.

Horizon's theme blocks remove the limit. Blocks can be nested inside other blocks to a meaningful depth, so a group block holds a layout block which holds text and button blocks, and a merchant composes that arrangement in the editor without touching Liquid. Practically, this changes who does the work. Layout that used to be a ticket becomes a drag.

The corollary is the reason there is no migration button. A Dawn section and its settings have no equivalent object in a nested block tree. Shopify did not build an automatic converter because the two models do not map onto each other, and any tool claiming to convert one to the other is guessing at intent.

What you actually gain

• Layout changes stop being developer tickets. The team that runs your campaigns can build a landing section without a release. For a brand shipping weekly promotions, this is the whole business case.
• Fewer page builder apps. Page builders are heavy, they inject their own JavaScript and they are a common cause of the interaction problems covered in our INP guide. Native blocks do the same job with no third-party runtime.
• A theme that will keep receiving Shopify's attention. Dawn is not abandoned, but the platform's new capabilities land in the new architecture first. Over a few years that gap compounds.
• A cleaner starting point. Horizon is built with current performance and accessibility practice, which is a better foundation than a five-year-old theme carrying five years of patches. The floor is higher; it is not a ceiling. What you install on top still decides your scores.

What it costs

Treat a Horizon move as a rebuild, priced and scheduled like one, because that is what it is.

• Every custom section has to be rewritten. Not ported. Rewritten against the block schema. Stores accumulate more of these than anyone remembers, so count them before you scope anything.
• Every template and its settings are re-authored. Section settings do not carry over. Somebody sits in the editor and rebuilds each page.
• App blocks need re-checking one by one. Apps that use the app block standard usually drop in. Apps that shipped install instructions telling you to paste a snippet into a specific Dawn file need re-integrating, and some vendors have not updated their instructions.
• Anything you patched into Dawn's files is gone. The tracking snippet, the size chart hack, the shipping-threshold logic in the cart drawer. Inventory these before you start, because they are invisible until the day someone notices a number stopped moving.
• Metafields and dynamic sources need rebinding. The data survives. The bindings do not.

Where the store is close to stock Dawn with a handful of custom sections, this is a contained project. Where the theme has been customised for three years by four different developers, it is a full rebuild, and it is worth being honest about that up front rather than discovering it in week three.

Who should move now

• You are launching a new store. No decision to make. Start on Horizon.
• You are already committed to a redesign. If a rebuild is funded and scheduled, doing it on the old architecture is buying the same work twice. This is the single clearest case.
• You are on a vintage theme. You are due a rebuild regardless, so rebuild onto the current generation.
• Your bottleneck is genuinely layout velocity. If your marketing team is blocked on developer time for landing pages every week, or you are paying for a page builder app to work around Dawn, the editor flexibility pays for the migration in a way you can measure.

Who should wait

• Your Dawn store works and converts. There is no deprecation deadline forcing this. Online Store 2.0 themes are current, supported and fine. Migrating a healthy store to get an editor feature you will use twice a year is a poor trade.
• You are inside peak season. Nobody replatforms a theme in the run-up to a sale weekend. See the BFCM readiness plan for what the calendar should look like instead.
• Your theme is deeply customised and the customisations still earn their keep. Rebuilding forty custom sections to gain editor flexibility you already effectively have, because a developer builds what you ask for, is not a return.
• Critical apps have not confirmed support. Ask the vendors before you scope, not after. One unsupported subscription or reviews app can stall a launch.

How to do it without losing a week to surprises

The sequence that keeps this boring.

• Inventory first, decide second. List every custom section, every app that touches the theme, every snippet pasted into theme.liquid, every metafield binding, and every hard-coded piece of merchandising. That list is the scope. Building it usually takes a day and saves a fortnight.
• Confirm app support with each vendor in writing. Not a forum post. The vendor.
• Build on an unpublished theme, on a real copy of production data. Never in the live theme.
• Rebuild templates in traffic order. Product, collection, cart, home, then the long tail. Most of your revenue rides on the first three.
• Re-measure before you publish. Field Core Web Vitals from Search Console for the current theme, then a like-for-like lab comparison on the new one. Publishing and hoping is how a rebuild quietly costs conversion.
• Check analytics end to end. Every tag, every event, every consent banner. A theme migration is the most common cause of a suddenly broken conversion feed. Our GA4 setup guide covers what to verify.
• Keep the old theme published until the new one has passed on real traffic. Rollback is one click while the previous theme still exists in your library. Keep it there for a fortnight.

Common mistakes

• Treating it as a theme swap. It is a rebuild. Scoping it as an afternoon is how stores go live with a broken cart.
• Migrating for the AI features. Editor assistance is a convenience, not a business case. Layout velocity and removing a page builder are business cases.
• Assuming Dawn is being switched off. It is not. There is no forced date, and urgency here is manufactured.
• Forgetting the snippets nobody documented. The pixel, the affiliate tag, the schema block someone added for SEO. Inventory or lose them.
• Rebuilding the old design faithfully. If you are paying for a rebuild anyway, use the flexibility. Recreating Dawn pixel for pixel in Horizon gets you the cost with none of the benefit.

FAQ