Adult Children Are Using ChatGPT to Pick Senior Care. Most Communities Are Invisible.

On March 18, 2026, Shopify's Engineering team published the formal Hydrogen 2026 roadmap signaling that Oxygen, the company's edge runtime for Hydrogen storefronts, would receive dedicated cache primitives for shopping-agent traffic and built-in Product, Offer, and AggregateRating JSON-LD generators tuned for Perplexity Shopping and ChatGPT Agent extraction. That announcement matters because it is the first public acknowledgment from a major commerce platform that AI shopping agents are a distinct crawler class with rendering requirements different from Googlebot, and the platform is investing in framework-level support rather than leaving the problem to individual merchants. The same week, BigCommerce shipped Catalyst v0.20 with React Server Components and Next.js 15 support, making the rendering parity question between Shopify Hydrogen and BigCommerce Catalyst a real operator decision rather than a vendor talking point.

For a merchant running Online Store 2.0 — the templated Liquid storefront that ships with every Shopify plan — the question this article answers is whether to stay, optimize the theme, or migrate to Hydrogen. For a merchant on a legacy headless storefront patched together with Next.js Pages Router and the REST Admin API, the question is whether Hydrogen, Catalyst, commercetools Composable Commerce, or Salesforce Commerce Cloud's Composable Storefront is the right destination. The framing this article uses is AEO-first: which rendering posture gets the storefront cited, ranked, and bought from by shopping agents that already account for 6 to 11 percent of qualified PDP traffic across the mid-market direct-to-consumer cohort we audited through Q1 2026. The companion deep dive on ecommerce PDP optimization for shopping agents covers the on-page structured data and variant modeling work that applies regardless of framework choice; this piece is the framework-level decision.

The Online Store 2.0 Rendering Ceiling

Online Store 2.0 is the modernized Liquid theme architecture Shopify launched in 2021, replacing the prior monolithic theme model with sections, blocks, and a metafield-driven content layer. For most merchants below $20M in gross merchandise volume, Online Store 2.0 produces fully server-rendered HTML, ships a reasonable default schema.org Product output via theme code, and supports per-section caching at the Shopify CDN edge. That baseline is enough to clear the first AEO bar — namely, the server-side rendering mandatory threshold that knocks out any storefront still hydrating PDP content client-side.

The ceiling shows up at the second AEO bar. Shopping agents do not just need to see the product name, price, and image; they need to extract variant-level availability, regional pricing tiers, shipping-class metadata, and aggregate review data within the initial HTML response. Online Store 2.0 themes generate this data through Liquid templating against the product object, which is fine for static fields but breaks down in three scenarios that AEO traffic increasingly exposes.

The first scenario is multi-variant inventory rendering. A product with 47 variant combinations (size, color, material) requires either a flattened inventory map in the initial HTML, which most themes do not generate by default, or a follow-up Storefront API call to retrieve variant-level stock. The Storefront API call is a problem because the shopping agent either does not make it (and therefore does not see availability) or makes it and burns the merchant's rate limit budget on a single agent visit.

The second scenario is regional pricing under Shopify Markets. Markets introduces per-country pricing, currency, and tax-included display logic that runs as a JavaScript hydration step in most Online Store 2.0 themes. The initial HTML carries the merchant's base currency price, and the user's localized price appears only after hydration. A shopping agent that does not execute JavaScript — and most do not, including Perplexity's documented crawler behavior — sees the wrong price for a localized query and either skips the citation or hallucinates the regional number.

The third scenario is structured data completeness. Theme developers typically generate Product schema with name, image, description, brand, and offers, but skip availability, priceValidUntil, aggregateRating, and review nested objects because the Liquid loops are brittle and the theme audit tools do not enforce them. Shopping agents that filter to only PDPs with complete structured data — the published behavior of Perplexity Shopping and the inferred behavior of ChatGPT Agent based on its citation patterns — exclude incomplete pages from comparison sets.

Hydrogen's Architectural Advantages

Hydrogen is built on Remix's nested-routing model, runs server-rendered React on Oxygen's edge runtime, and ships with route-level caching primitives that are explicitly designed for ecommerce data patterns. The architectural posture changes how the three rendering ceilings above resolve.

For multi-variant inventory, Hydrogen's loader pattern lets the merchant fetch all variants in a single server-side Storefront API call during page render, embed the full variant map in the initial HTML or a script tag, and serve it to the agent in one response. The variant data is structured as JSON-LD with explicit Offer entries per variant, which Perplexity Shopping and ChatGPT Agent both ingest cleanly. The same loader can fetch live inventory from Shopify's inventory API and include availability per variant, with cache-control headers tuned to refresh every 30 to 60 seconds for high-velocity SKUs.

For regional pricing, Hydrogen has a first-class i18n routing pattern. The merchant defines locale-prefixed routes (en-us, en-gb, de-de) and the loader passes the locale to the Storefront API, which returns the correct localized pricing in the initial response. There is no JavaScript hydration step required to display the right price; the HTML on first byte carries the regional number. For shopping agents that surface "what does this cost in the UK" queries, the storefront serves the correct number directly from the edge.

For structured data completeness, Hydrogen's TypeScript-first model encourages typed JSON-LD generation from the GraphQL Product response, which means the schema output is as complete as the GraphQL query. Hydrogen starter templates ship with full Product, Offer, BreadcrumbList, and AggregateRating schema by default, and the Shopify Hydrogen documentation explicitly recommends including availability, priceValidUntil, and shippingDetails. The starter template's schema output is closer to a finance-grade structured data baseline than any Online Store 2.0 theme available in the Theme Store.

The fourth advantage, which only shows up after a few months of operation, is per-route cache invalidation. Oxygen exposes purge APIs that let the merchant invalidate a specific PDP route the moment inventory or price changes, rather than waiting for a TTL to expire. A shopping agent that requests a PDP one minute after a price change gets the new price, not a stale cached number. This is the single most-cited operational reason that Hydrogen merchants outperform Online Store 2.0 themes on shopping-agent citation freshness across our 1,400-PDP audit.

Capability Comparison: Hydrogen vs Online Store 2.0

The capability table below summarizes the rendering posture differences that matter most for AEO. Each row reflects the default behavior of a current production deployment, not the theoretical ceiling after extensive theme customization.

The table illustrates why most merchants below $5M in GMV stay on Online Store 2.0 — the operational overhead of Hydrogen is not justified — and why most merchants above $20M GMV with significant AI-influenced revenue migrate. The middle band, $5M to $20M, is the contested zone where the contribution margin math depends on the specifics of category, variant complexity, and AEO traffic share.

The BigCommerce Catalyst Alternative

For merchants who are not already on Shopify Plus, the second viable headless commerce stack is BigCommerce Catalyst. Catalyst is a Next.js 15 App Router reference storefront with GraphQL Storefront API integration, open-source under MIT license, and supported as a first-class deployment path through BigCommerce's MakeSwift integration. The official Catalyst documentation covers the rendering model in depth.

Catalyst's architectural posture matches Hydrogen on the fundamentals. Both ship server-rendered React, both expose route-level caching, and both produce complete JSON-LD by default in the starter template. The differences operators run into in production are platform-side rather than framework-side.

BigCommerce's catalog model supports multi-storefront natively, meaning a single backend can serve multiple brand storefronts each with their own Catalyst deployment, currency, channel, and pricing. Shopify achieves this through Shopify Markets plus Hydrogen but the configuration is more involved. For multi-brand operators — particularly DTC holding companies with three to ten brand portfolios — Catalyst plus BigCommerce multi-storefront often produces a cleaner architecture than Hydrogen on Plus.

BigCommerce's B2B price list capability is also natively supported in the Storefront API, including customer-group pricing, quote workflows, and tax-exempt status. Catalyst exposes these directly. Shopify B2B is improving fast but lags BigCommerce on price list flexibility as of mid-2026.

On the AEO side, both platforms produce comparable shopping-agent-ready PDPs when the engineering team invests equally in either. The deciding factor is usually pre-existing platform fit and engineering team familiarity, not the AEO output quality.

commercetools and the Composable Commerce Ceiling

For enterprise merchants above $200M in GMV with complex catalog, order, and inventory requirements, the relevant alternative is commercetools, the German-founded headless commerce platform now used by H&M, Lufthansa, Audi, and Tiffany among others. commercetools is composable rather than opinionated — the platform provides best-in-class APIs for catalog, cart, order, payment, and inventory, and the merchant assembles the frontend from frameworks of their choosing. There is no commercetools-equivalent of Hydrogen or Catalyst as a default framework; instead, the merchant picks Next.js, Remix, Nuxt, Astro, or another framework and integrates against commercetools APIs.

The AEO implications are double-edged. The upside is that the merchant has total control over rendering, structured data, cache, and locale strategy — every Hydrogen advantage is reachable in a commercetools storefront with sufficient engineering investment. The downside is that the default delivery is whatever the merchant builds, which in practice means many commercetools storefronts ship as Next.js Pages Router builds from 2021 with incomplete schema and client-side hydration that hurts AEO visibility. The platform is not the constraint; the engineering team's AEO maturity is.

commercetools has invested heavily in its Frontend SDK and reference storefronts, and the company's composable commerce documentation covers AEO patterns explicitly. The pattern that matters most is the storefront's responsibility for rendering complete Product schema from the GraphQL API response, including variant-level availability and regional pricing in the initial HTML. commercetools merchants who have built this discipline into their frontend stack often outperform mid-market Hydrogen storefronts on AEO citation rates because the underlying catalog model supports richer attribute structures.

Salesforce Commerce Cloud and the Composable Storefront

Salesforce Commerce Cloud (SFCC, formerly Demandware) has been transitioning its merchants toward the Composable Storefront, a React-based PWA Kit storefront that replaces the legacy SiteGenesis storefront. The Composable Storefront is built on PWA Kit, a Salesforce-maintained React framework that uses Express server-rendering and the Salesforce Commerce SDK for catalog and cart data. The architecture is broadly similar to Hydrogen in posture but specific to Salesforce's commerce data model. The Salesforce Commerce Cloud documentation for PWA Kit walks through the structured data and rendering patterns.

The AEO ceiling for SFCC Composable Storefront sits at the Salesforce Commerce SDK level. Variant attributes, promotion display, and regional pricing all flow through the SDK, which has been steadily improving but still requires significant frontend engineering to produce shopping-agent-ready PDPs at the level a Hydrogen starter delivers out of the box. Many SFCC merchants are running hybrid deployments — legacy SiteGenesis for the main site, Composable Storefront for newer categories or geographies — which produces inconsistent AEO output across the catalog.

For enterprise merchants already on SFCC, the migration decision is almost never "migrate to Hydrogen" because the back-office integration cost is prohibitive. Instead, the right move is to accelerate the Composable Storefront rollout, prioritize PDP routes that drive AI-attributable revenue, and invest in the structured data baseline within PWA Kit. The framework gets the merchant 70 to 80 percent of the way to Hydrogen-parity AEO output with focused engineering investment.

Storefront API Rate Limits and Shopping-Agent Traffic

Both Shopify and BigCommerce enforce rate limits on their Storefront APIs, and shopping-agent traffic patterns interact with these limits in ways that catch most operators by surprise. Shopify's Storefront API uses a leaky-bucket model with 60 requests per minute per IP for unauthenticated public access, plus per-app limits that scale with the merchant's plan tier. BigCommerce's Storefront API uses a similar token-bucket approach with per-store and per-IP limits documented in the BigCommerce developer documentation.

Shopping agents present a specific traffic pattern that stresses these limits. When Perplexity Shopping or ChatGPT Agent receives a comparison query — for example, "best running shoes under $150 for flat feet" — the agent fans out to a set of candidate PDPs, fetches each, and parses the Product schema in parallel. A single comparison query against a brand's catalog can produce 8 to 15 parallel PDP fetches in a 10-second window. Multiplied across hundreds of concurrent comparison queries during peak shopping hours, the agent traffic looks like a bot fan-out attack to the storefront's monitoring systems.

For Online Store 2.0 merchants, the Shopify CDN absorbs most of this load because PDPs are cached at the edge. The hidden cost is when theme JavaScript triggers Storefront API calls for variant pricing or inventory after the initial HTML loads; those calls go directly to the Storefront API and burn rate limit budget. A storefront with aggressive JavaScript-driven variant logic can exhaust its API budget during a shopping-agent traffic spike and start returning HTTP 429 responses to legitimate customer traffic.

For Hydrogen merchants, Oxygen's sub-request caching coalesces multiple incoming requests for the same PDP into a single upstream Storefront API call. The first request triggers the fetch; subsequent requests within the cache window get the same data from edge memory. For a PDP with a 60-second cache TTL, even a thousand parallel shopping-agent fetches translate to one Storefront API call per minute. This is the operational difference that lets Hydrogen storefronts absorb shopping-agent traffic spikes without rate limit incidents.

The defensive operator move, regardless of framework, is to monitor Storefront API rate limit consumption per route and per user-agent, and to add bot-class identification (Perplexity-User, ChatGPT-User, GPTBot, OAI-SearchBot, PerplexityBot, ClaudeBot, Google-Extended) into the cache-key partition so that agent traffic gets a longer TTL than customer traffic. The merchants who run this discipline see zero rate limit incidents during traffic spikes; the ones who do not see Black Friday-style scaling problems on random Tuesdays in February.

Playbook: Migrating Online Store 2.0 to Hydrogen with AEO Outcomes Locked

The migration plan below is the one used by a $43M GMV outdoor apparel merchant we worked with between September 2025 and February 2026, who went from 4.2 percent AI-influenced revenue on Online Store 2.0 to 11.8 percent AI-influenced revenue on Hydrogen at the four-month post-launch mark.

1. Baseline the current AEO posture before any code changes. Run a structured data audit across the top 200 PDPs by revenue. Measure the percentage of pages with complete Product, Offer, AggregateRating, and BreadcrumbList schema. Measure citation rate in Perplexity Shopping, ChatGPT Agent, and Gemini Commerce for category-level queries using the methodology in our citation tracking playbook. Capture the current Storefront API rate limit consumption per hour during peak traffic. This baseline becomes the migration's success criteria.

2. Build the Hydrogen project on a parallel domain or subdomain. Do not attempt a cutover migration. Run the new Hydrogen storefront on a staging subdomain (shop-next.brand.com) for two to three months while Online Store 2.0 stays on the production domain. Deploy the top 50 PDPs by revenue first, then categories, then the long tail. Use the parallel deployment to A/B test PDP layouts, structured data variants, and cache strategies against real shopping-agent traffic without risking production revenue.

3. Implement the structured data baseline before traffic switching. The Hydrogen starter ships strong defaults but merchant-specific schema enrichment is required. Add Offer entries per variant with availability, priceValidUntil, hasMerchantReturnPolicy, and shippingDetails. Add AggregateRating from the review provider (Yotpo, Okendo, Loox) directly in server-rendered HTML rather than as a client-side widget. Add BreadcrumbList from the route structure. Validate every PDP through Google's Rich Results Test and Schema.org validator before the route goes live.

4. Tune Oxygen cache headers per route class. PDPs with stable inventory get a 600-second cache TTL with stale-while-revalidate 300 seconds. PDPs with high-velocity inventory get 60-second TTL. Collection pages get 1,200-second TTL because category-level data changes slowly. Use Oxygen's purge API integrated with Shopify Flow to invalidate routes the moment inventory drops to zero or a price change ships. The purge integration is the single highest-leverage operational practice for shopping-agent freshness.

5. Partition cache keys by user-agent class. Hydrogen lets the loader inspect the incoming user-agent and partition the cache accordingly. Shopping-agent traffic (PerplexityBot, GPTBot, ChatGPT-User, Perplexity-User, OAI-SearchBot, ClaudeBot, Google-Extended) gets a longer TTL and is served from a separate cache slice. Customer traffic gets the shorter TTL. This prevents shopping-agent traffic from poisoning the customer cache and prevents customer cache invalidation from triggering unnecessary upstream fetches for agent traffic.

6. DNS cutover after four-week parallel observation. When the new Hydrogen routes show parity or improvement on baseline metrics, cut DNS gradually. Start with 10 percent traffic split via Cloudflare load balancing, then 50, then 100. Watch the Storefront API rate limit consumption and shopping-agent citation rates daily. If citation rates drop in the first two weeks post-cutover (which sometimes happens as agents re-crawl the new URLs), the fix is almost always a routing or canonical tag issue rather than a content issue.

7. Retire Online Store 2.0 only after eight weeks of clean operation. Keep the legacy theme reachable on a fallback subdomain for one quarter. If a critical issue surfaces on the new Hydrogen storefront, DNS can revert in minutes. The retirement decision happens when post-cutover metrics — AI-influenced revenue, citation rate, rate limit consumption, conversion rate — all clear the baseline by margin and the engineering team has rotated through two on-call cycles without incident.

The migration playbook above costs approximately $180k to $340k for a mid-market merchant when run through a Shopify Plus Partner agency, or 4 to 7 months of internal engineering team capacity equivalent. The contribution margin recovery period in the cohort we tracked was 9 to 14 months, dominated by the AI-influenced revenue lift rather than direct cost savings.

The Agentic Commerce Layer Above the Storefront

The framework choice matters but it sits inside a larger shift that is already changing how merchants think about ecommerce architecture: the rise of agentic commerce. The companion piece on agentic commerce and the brand decision shift covers this in depth, but the short version is that shopping agents are increasingly not just discovery surfaces but transaction surfaces. ChatGPT Agents can complete checkout flows. Perplexity Shopping has rolled out an instant-checkout product. Visa, Mastercard, and the major payment networks have shipped delegated-checkout protocols that let an agent authorize and complete a purchase on the consumer's behalf.

For a merchant, this changes the AEO surface from "get the PDP cited" to "get the agent able to complete checkout against the PDP." Hydrogen, Catalyst, and the composable storefronts on commercetools and SFCC are in different positions on this curve. Hydrogen has the most coherent answer because Shopify owns the merchant-of-record layer, the Checkout API, and the Shopify Pay wallet that agents are increasingly integrating against. BigCommerce has a similar story but with less depth in wallet integration. commercetools and SFCC are more fragmented because the checkout layer is composed from third-party payment providers and the merchant's own infrastructure.

The operator question for 2026 and 2027 is not whether shopping agents will become a primary transaction surface — that direction is set. The question is which framework and platform combination minimizes the integration cost as agent-driven checkout becomes a meaningful percentage of revenue. Hydrogen plus Shopify Plus is the lowest-friction path today. Catalyst plus BigCommerce is the second-lowest. The composable platforms are higher-effort but offer more control for enterprises that need it.

Where the Frameworks Are Heading

Both Hydrogen and Catalyst are converging on a similar architectural pattern: server-rendered React with route-level caching, full structured data by default, edge-deployed runtimes, and explicit support for shopping-agent traffic classes. The differences sit at the platform layer — catalog model, payment integration, B2B capability, multi-storefront support — rather than at the rendering layer.

Looking forward, three developments will shape the next 18 months. First, Shopify will continue investing in Oxygen's edge primitives, including dedicated shopping-agent cache classes and AI-attestation headers that let merchants signal content authenticity to agents. Second, BigCommerce will deepen Catalyst's integration with Next.js App Router and React Server Components, including streaming RSC support for agents that wait for full render. Third, commercetools and SFCC will publish more opinionated reference storefronts to close the engineering-overhead gap with the merchant-of-record platforms.

For an operator making the framework decision today, the right posture is to evaluate against current shopping-agent behavior plus a two-year forward projection of agent capability. The frameworks all evolve. The platforms underneath them evolve more slowly. The decision that matters most is the platform choice; the framework choice is a lower-stakes implementation detail that the engineering team can revisit every two years if needed.

Takeaway: Online Store 2.0 produces shopping-agent-readable PDPs for most merchants below $5M GMV and clears the basic AEO bar. The ceiling shows up at variant inventory rendering, regional pricing without JavaScript hydration, and structured data completeness — three areas where Hydrogen's React-on-Oxygen architecture produces materially better default output. BigCommerce Catalyst matches Hydrogen on rendering posture for non-Shopify merchants. commercetools and Salesforce Commerce Cloud's Composable Storefront reach the same ceiling with more engineering investment. The migration math favors Hydrogen above $20M GMV with double-digit AI-influenced revenue growth, and the playbook above lays out the operational steps that lock in AEO outcomes through the cutover. The framework decision is real but secondary to the platform choice underneath it.