Alessandro Desantis

Every six months, Shopify ships an Edition. And every six months, our clients ask us the same thing: *So what's in it for us?*

It's a reasonable question. Roughly 150 updates ship in every Edition, and the average brand only needs to care about a handful of them.

But this time, if you step back from those 150 updates and look at the bigger picture, a pattern emerges: for two decades, Shopify has been selling you the software you needed to build the best possible e-commerce website.

This Edition is the strongest sign yet that they are now ready to turn into the digital commerce layer that powers the entire World Wide Web.

For DTC brands, that's both good news and bad news. Here's why.

## Your Catalog Is Now Our Catalog

When Shopify entered the agentic commerce race, they did so with their usual strategy: a new discovery channel was just emerging, and they wanted to be front and center, enabling their merchants to succeed in this new world. To accomplish this, they partnered with the major AI platforms (ChatGPT, Copilot, Perplexity) to syndicate each merchant's product catalog and, in some cases, allow direct checkout.

The technical approach included a few false starts, but ultimately it involved partnering with Google to release UCP (Universal Commerce Protocol), a protocol that enables LLMs (or any third-party system, for that matter) to seamlessly browse a merchant's catalog, manage a persistent cart, and check out. I will spare you the technicalities of the protocol, but if you're curious, I recently published [a more detailed breakdown](https://www.linkedin.com/posts/alessandro-desantis_ecommerce-ai-shopify-activity-7416461741892734976-2fjP/) of UCP and what makes it so cool (and better than OpenAI's ACP, in my humble opinion).

Back then, we speculated that this created the potential (and the danger) for the entire Internet to turn into a giant Amazon, where every single product on Earth is just one prompt away. The theory went that, due to the very nature of conversational shopping, this may allow an entire long tail of brands to emerge and compete with the incumbents without having to match their gargantuan marketing budgets. At the same time, we also speculated that this may lead to the decline of the DTC storefront, especially in some of the more commoditized categories.

These were very long-term bets, and we're still waiting to see how they play out. So far, it's turned out that the tricks that make a brand surface in AI conversations are very similar to the ones that make brands surface on Google, Meta, and all the other channels we already know and love. As for the demise of the DTC website, even though we work with tens of brands in eight- and nine-figure territory, we do not have a single client for which AI sales make up more than 0.5% of online GMV (if you exclude Shop, which Shopify disingenuously rolls up into their Agentic dashboard—more on that below).

However, what *has* changed with this Edition is that Shopify has made its entire merchant base more accessible and omnipresent than ever. In addition to syndicating their merchants' products to all the usual suspects, anyone can now integrate with the Catalog MCP, Cart MCP, and Checkout MCP (the three modules that make up the UCP specification) to build a full-fledged shopping experience powered by Shopify's infrastructure.

While they were at it, [Shopify also strengthened the Catalog MCP](https://shopify.dev/docs/agents/catalog/global-catalog-extension) with:

- Image and similarity search (both pretty self-explanatory).
- Offers from multiple sellers. The same product can be offered by multiple Shopify stores, each with its own listing.
- Affiliate placements (invite-only). By opting into the feature, platforms can surface paid placements in the search results and receive a commission on sales.

To show how some of the new functionality could be used in the wild (and, I guess, to let their dev team have some fun), Shopify released a few cool demos to go with it:

- [All Set](https://www.shopify.com/editions/demos/all-set), where you can tell an AI about your trip and get product recommendations.
- [Sourced](https://www.shopify.com/editions/demos/sourced), where you can upload a picture and buy the products in it.
- [Showroom](https://www.shopify.com/editions/demos/showroom), where you can buy the products featured in your favorite movies and TV series.
- [Starred](https://www.shopify.com/editions/demos/starred), where you can get product recommendations based on your daily horoscope.
- [Pippin](https://www.shopify.com/editions/demos/pippin), where the 21st-century version of Clippy gives you board game recommendations.

I encourage you to check them out, because they're really fun!

## The Shop App and the Buyer Identity Play

If the agentic infrastructure is the plumbing, Shop is where it becomes something a consumer can actually interact with. In this Edition, Shopify has improved the app in several ways.

First of all, Shopify is exposing new surfaces within the Shop app to help consumers discover the products they're looking for. This includes:

- Merchandised categories, curated both by Shopify and by third parties (e.g., [Selleb](https://www.selleb.com/)), that showcase select products and brands.
- [Shop Minis](https://shop.app/minis), third-party mini-apps that run within Shop and create new ways for the consumer to engage with the ecosystem (e.g., recipe trackers).
- The [Shop skill](https://shop.app/skill), which allows consumers to easily connect the Shop universe to an LLM/harness (e.g., Hermes, OpenClaw).

But the more interesting development, if you ask me, is how Shopify is starting to integrate Shop within the rest of its ecosystem.

Thanks to the years it has spent evolving the Shop app and pushing Shopify merchants to enable Shop-based sign-in in their storefronts and checkouts, Shopify is now sitting on what is probably (after Amazon's) the largest consumer identity graph in the digital commerce industry. They are now starting to activate that data in pretty interesting ways.

More specifically, Shopify has announced that the Catalog MCP will soon support [personalized searches.](https://shopify.dev/docs/agents/catalog/global-catalog#personalized-search) As long as you authenticate users via Shop (which you can now do even outside of a Shopify site), you can pass a buyer-linked token when making requests to Shopify's catalog. When that happens, the search results will be personalized according to the information Shopify has about the buyer's preferences and behavior.

What would be incredible—and this is a very long shot—is if merchants could leverage that same data on their storefronts as well. How cool would it be if you could have your customers authenticate via Shop and use Shopify's identity graph to seamlessly surface personalized collections and search results? Given the privacy issues involved and the architectural limitations of merchandising in Online Store (we have an article coming out about this soon), I don't see this coming any time soon—but hey, a man can still dream.

With these features and improvements, Shopify has turned Shop into the *de facto* demo for what can be accomplished with their catalog. In fact, I wouldn't be surprised if the Shop app had been recently rewritten, at least in part, to use the new UCP specification rather than Shopify's internal APIs, so that it can serve as a sort of reference implementation for the new capabilities.

## How Should Merchants Feel About All This?

For two decades, Shopify has told its merchants that it had no plans of becoming a marketplace (despite some marketplace-y functionality coming to Shop way before ChatGPT was ever a thing). Instead, they were focused on providing the best possible tools for each merchant to build their beautiful walled garden.

What no one could foresee is that a new shopping surface would come along that would force *the entire Internet* to turn into one huge marketplace.

When that happened, they decided the best strategy was to ride the wave rather than hide from it. And Shopify being the powerhouse it is, they nailed it: between their work on UCP and the identity graph they are building with Shop, it's not an exaggeration to say that they have become the commerce infrastructure for the open Web—so much so that they have now got Amazon [scrambling to catch up](https://www.readtheaisle.com/p/amazon-agentic-commerce-jobs-hiring-ai-shopping). Whatever shifts happen next, I fully trust that Shopify merchants will be in the best possible position to benefit from them.

At the same time, make no mistake: *some* disintermediation is coming. Just like marketplaces, agentic commerce offers merchants a new distribution channel but also requires them to give up control over the shopping experience. Sure, UCP affords some flexibility, but it will never be able to replicate your merchandising strategy or your in-cart upsells, nor should it necessarily: when a consumer is checking out via ChatGPT, they are often in a very different state of mind than when they are browsing your site.

This last point, I think (hope?), is *the* point we should all be focusing on: as agentic shopping channels get bigger and bigger, the website is not going away, but it will definitely evolve: rather than being a channel consumers *have* to interact with, it will gradually become one consumers *want* to interact with.

Because of this, DTC websites may end up getting less traffic than before, but (most of) that traffic will likely be much more engaged. Customers will end up on your site having already done their homework, and looking to really immerse themselves in your brand universe. This will allow merchants to create much richer, deeper, weirder DTC experiences that aim to create a connection, not just usher the customer out the door as quickly as possible—what our friends at Future Commerce called [Dork Mode](https://www.futurecommerce.com/posts/insiders-107-dork-mode-an-antidote-to-the-sea-of-sameness) in a prescient 2021 article.

And while no one knows what form this will take (more on our possible futures in an upcoming article) or how long it will take for us to get there (remember: on average, agentic sales make up 0.5% of GMV for our clients), that is something I can get excited about.

Shopify's Agentic Strategy and the Future of DTC

Alberto Vena

First, a disclaimer: this is NOT a feature recap. If that's what you're looking for, Shopify's [Editions site](https://www.shopify.com/editions/winter2026) is the place to be. Instead, what we're doing here is what we've done with every Edition for the last few years: analyzing what these updates actually mean for upper-midmarket and enterprise brands, where Shopify created genuine value, and what they could have done better.

In the Winter 2026 Edition specifically, AI stole the spotlight. Sidekick now generates admin apps! It generates theme sections! It supports Skills and third-party integrations! If you read the product announcements, it all feels like success is just one carefully engineered prompt away.

And truth be told, there is substance behind the hype. Shopify Magic, and Sidekick specifically, are powerful tools that are only getting more powerful. Agentic Storefronts are putting every Shopify brand in front of hundreds of millions consumers using LLMs to shop, all with a click of a button. And much-needed infrastructure improvements and platform additions (e.g., A/B testing and theme rollouts) are making life easier for merchants, developers, and consumers.

But when you evaluate these evolutions through the lens of an upper-midmarket brand moving serious volume, there are some nuances to consider. In this article, we're going to talk about those nuances, highlighting some of the biggest opportunities that still lie ahead for the Shopify ecosystem.

## Sidekick's Code Generation Is Powerful But Requires Governance

Sidekick's code generation capabilities, for both themes and custom apps, represent an interesting leap forward in operational velocity, but they're not the cure-all some industry pundits are making them out to be.

In our experience, Sidekick excels at rapid theme prototyping. It gets you to about 80% visual accuracy fast, which is valuable when testing concepts or exploring ideas quickly. But when you think about long-term maintenance, you'll run into challenges such as:

- Excessive configuration options for settings that you'll never use.
- Isolated CSS and JavaScript that doesn't use existing utilities or patterns.
- No consideration whatsoever for the build system used by your theme (if any).

Admin app generation comes with similar benefits but also similar challenges.

The app store is littered with low-quality apps solving simple problems for $9.99/month, and we're all happy to see a few of those go away. If merchants can generate custom solutions they actually own, that's a genuine win.

But without a clear governance framework, your stakeholders will start generating apps that solve very similar problems in very different ways, with no consideration of existing patterns or cross-app interactions. As this functionality expands to storefront-facing features, which seems inevitable given the direction Shopify has taken with AI code generation, the stakes will get even higher. While admin apps affect operational efficiency, storefront apps affect revenue.

Moving forward, we'd like to see Shopify invest not just in AI code generation, but also in stronger primitives and better development guidelines that guide the AI, so it can be more considerate of its surroundings and the long-term impact of its output. Shopify should also educate its merchants on proper stewardship of AI-generated functionality, helping them answer questions such as:

- Which challenges are good candidates for AI-assisted solutions and which are not?
- How do we ensure that generated code integrates with existing patterns rather than creating islands of functionality?
- When should we hand off an AI-generated solution to our engineers for polish and maintenance?

At Nebulab, we're starting to establish such processes with our clients and getting the best of both worlds: the speed of AI prototyping combined with the discipline of human oversight and integration.

Without clear handoff processes and governance frameworks, you accelerate the accumulation of technical debt rather than evolution of your store.

## Sidekick's Data Analysis Lacks Integration

Sidekick's business intelligence capabilities now come in two forms: the conversational chatbot that answers questions about your store data, and the new Pulse feature where Sidekick proactively suggests business initiatives. Unfortunately, both face the same fundamental limitations around data access and context.

For a $5M brand with a small team and straightforward operations, these tools can provide valuable direction. The suggestions might be generic, but generic advice often works when you're just starting out, your business model is simple, and your data lives primarily within Shopify.

For a $50M operation managing 1,500 SKUs across wholesale, marketplaces and DTC, coordinating 60 employees, and balancing a multi-year positioning shift against quarterly revenue targets, the calculus changes entirely. At this scale, context matters _a lot_. Generic optimization suggestions that ignore your specific constraints only create noise that your team has to filter through.

Sidekick can't tell you whether increasing email frequency will increase sales or damage your brand perception. It can't factor in that your warehouse is already at capacity and can't support that new sales channel. It doesn't know that your bestselling product line is being phased out next quarter and you should start weaning customers off of it.

More fundamentally, Sidekick is limited to the data that Shopify has. For businesses that integrate data from ERPs, warehouse management systems, customer service platforms, marketing attribution tools, and financial systems, Sidekick's view is necessarily incomplete. It sees transactions and customer behavior within Shopify, but lacks the full picture.

This is our problem with platform-specific chatbots and agents: they all only have a small piece of the truth and a small set of all the tools they could use, which prevents them from orchestrating larger business processes.

While Shopify is nice for smaller brands, we'd also like to see Shopify invest in better data portability—through official partnerships with ETL providers, data warehouses, BI tools, and external LLMs—so that the data can be integrated as part of larger operational workflows. Give enterprises the ability to bring Sidekick's analytical capabilities into environments where it has access to the full context, not just Shopify's limited view.

Without that broader integration, Sidekick is useful for surface-level insights but inadequate for strategic decisions.

## SimGym and A/B Testing: A Missed Opportunity

Shopify launched two experimentation features in the Winter '26 Edition: native A/B testing and SimGym.

Native A/B testing is new and useful. You can now natively split traffic between theme variations and measure performance differences directly within Shopify. For smaller brands without dedicated experimentation infrastructure, this means you don't have to use external experimentation platforms and lowers the barrier to entry significantly (although you will still need enough traffic to power your tests).

Still, we were disappointed that Shopify didn't launch an API that other tools could ubuild upon, like they did for example with the Subscriptions API. Instead, external A/B testing solutions will continue to use the same client-side hacks they've always relied on.

This is a shame because large-scale brands doing serious experimentation (multiple treatments, custom metrics, audience exclusions) will never be able to use Shopify's native A/B testing. Allowing third-party tools such as Shoplift or Intelligems to build upon Shopify's offering would have dramatically improved the experimentation experience for merchants (through a more seamless experience) and consumers (through faster page loads).

In addition to native A/B testing, Shopify launched SimGym, which takes a very different approach: rather than testing with real users, it uses AI to generate synthetic user behavior and predict experiment outcomes. The concept is interesting: in theory, brands could use synthetic data to run tens of experiments and move iterate much faster on their UX.

Unfortunately, research from Nielsen Norman Group and other UX practitioners has found that [synthetic users exhibit idealized behavior](https://www.nngroup.com/articles/synthetic-users/), lack the messiness of human decision-making, and produce overly optimistic results. AI-generated participants are overly agreeable and to miss the contextual nuances that shape actual behavior. In short, they're unreliable for concept testing and business decisions.

This may be a bit far-fetched, but we think Shopify missed a broader and much more interesting opportunity here.

Instead of generating synthetic predictions, imagine if Shopify had opened its A/B testing framework to third parties and then systematically collected and surfaced anonymized data about tests across its entire merchant network. With enough context about the business and the ability for the AI to understand what's being tested (i.e., not just surface metrics, but the strategic intent behind the test) Shopify could offer an invaluable best practices playbook to smaller brands and a sophisticated A/B testing engine to larger ones.

For example, a brand testing a new product page layout could receive insights like: "Among premium skincare brands with AOV above $75, similar layout changes increased conversion by 8-12% when combined with trust signals, but decreased conversion by 3-5% when the change removed product ingredient details." That's actionable intelligence grounded in real behavior seen in similar contexts.

This would have been a win-win-win for Shopify, third-party A/B testing apps, and merchants. The data would have been much more reliable than synthetic predictions and it would have been a true competitive moat for Shopify—proprietary insights that no other platform could match. Instead, we got synthetic predictions of uncertain value and a basic testing framework that doesn't serve the needs of more sophisticated merchants.

## Whatever Happened to Hydrogen?

Clearly, Hydrogen is not a focus at Shopify anymore.

If you haven't followed the story, Shopify has made some big moves in the headless space between 2021 and 2024:

- In 2021, they launched Hydrogen, their React-based framework for headless commerce experiences, and Oxygen, their platform to host Hydrogen storefronts.
- In 2022-2023, they acquired Remix and used it to rewrite Hydrogen from scratch, positioning it as a major advancement in their headless commerce strategy.
- In 2024, they announced the Hydrogen Visual Editor, an entry-level alternative to more sophisticated solutions like Sanity and Contentful.

At some point around 2024, Shopify quietly started shifting their focus from headless commerce to AI as the next big thing:

- Hydrogen's GitHub repository has seen minimal activity in the last few months. Most of it was API version updates and dependency maintenance
- The Hydrogen Visual Editor never made it past the early access phase. As far as we can see, the initiative was killed before seeing the light of day.
- The Winter 2026 Edition includes exactly _zero_ Hydrogen announcements.

Our own conversations with Shopify sources confirmed our suspicions: Shopify is discouraging brands from going headless unless they have a very strong reason to do it. In fact, they are telling Hydrogen brands to move back to Liquid.

This is not the disaster it seems. Hydrogen has always been plagued by a number of problems, most of them tied to the dynamics of the Shopify and JavaScript ecosystems rather than the framework itself:

- Most of Shopify's app ecosystem assumes that merchants use Liquid. Brands using Hydrogen often need to implement their own integrations (assuming the tool they're integrating with is kind enough to offer an API).
- While Hydrogen offers an excellent developer experience, the JavaScript ecosystem is famous for being extremely fragmented. It was not uncommon for development teams to ignore Hydrogen in favor of their preferred framework.
- Most Shopify product announcements would make it into Liquid way before making it into Hydrogen, which was a source of endless annoyance.

None of this is to say that you shouldn't go headless, if your brand is complex enough to need it. But you should know that, in 2026, headless is not the default choice anymore. Before going off the beaten path, make sure you have evaluated the pros and cons of doing so. We'll publish a much deeper dive about this in the next few weeks!

## Where To Focus Your Attention

If you're running an upper-midmarket brand, here's what actually matters from this release:

- **Use Sidekick for rapid prototyping**, but establish clear handoff processes before your team generates a dozen isolated apps solving similar problems in different ways. The speed is real, but so is the technical debt if you don't have governance in place.
- **Double check the AI's insights.** Sidekick is useful for quickly extracting raw numbers from Shopify, but it still hallucinates every now and then and it lacks important context about your business. Don't any results or conclusions at face value.
- **Adopt native A/B testing if you're only starting out**. It's a solid foundation for brands just starting to build experimentation capabilities. If you're already running complex tests with custom metrics and audience segmentation, you'll still need third-party tools.
- **Ignore synthetic data**. Synthetic user predictions are interesting in theory but unreliable in practice. If you have the traffic, run real tests with real customers. If you don't have the traffic, stick to e-commerce UX best practices until you do!
- **When it comes to headless, thread lightly**. Shopify's investment has clearly shifted elsewhere. Unless you have compelling architectural reasons to go headless, the default path is Liquid.

As always, if you need help sorting through adopting the new features or understanding what they mean for your business, [let's talk.](https://nebulab.com/contact-us)


Shopify's Winter 2026 Edition: The Needle Movers and the Missed Opportunities

Michela Frecchiami

Would you buy a $2,000 cashmere sweater without understanding how the fabric drapes on your body? Would you commit to a $5,000 handbag if you couldn't see the grain of the leather or how it sits on your shoulder? For most customers, the answer is no—and that's the fundamental challenge luxury brands face online. A poorly rendered material or unrealistic fit doesn't just lose a sale—it damages brand perception in ways that take years to rebuild. 

AI-driven virtual try-on is changing this equation. Augmented reality technology has existed for over a decade, but it's only now—thanks to its combination with advanced AI—that it's reached a quality level suitable for luxury. What's changed is that recent advances in AI, particularly generative models and diffusion algorithms, have transformed AR from a novelty into a genuinely useful tool. Today's systems can render materials with near-perfect accuracy, understand body types and movements, and account for real-world lighting conditions. 

But here's the critical insight: only a seamless, intuitive, and realistic experience can actually deliver benefits. The key is **realism over perfection**—virtual try-on needs to look authentic, not artificially polished. When rendering feels too perfect or synthetic, it triggers skepticism. AR models that slow down navigation, feel clunky to use, or produce artificial-looking results create a boomerang effect—damaging brand perception instead of enhancing it.

## The Business Case: Numbers That Actually Matter

Luxury and jewelry e-commerce typically sees conversion rates of around 1.0%–1.2%, which is significantly lower than the global average of 1.4%, according to Shopify's 2024 benchmarks. Yet retailers implementing AI-powered virtual try-on and 3D/AR experiences see consistent improvements: Shopify reports show that interacting with a 3D model increases purchase likelihood by 27%, while AR interactions can drive up to a 65% lift in conversion rates. 

More importantly, returns decrease measurably—Shopify case studies report 5–10% lower return rates on products using AR visualization, while third-party aggregated studies note reductions of up to 30–40% in specific categories. For a brand selling €800 shoes or €3,000 handbags, this doesn't just reduce logistics costs—it preserves inventory value and the ability to resell items at full price. 

Consider a brand doing €50M in annual e-commerce revenue. A 40% conversion lift on a 1.0% baseline brings you to roughly 1.4%—that's potentially millions in additional revenue. A 20–35% reduction in returns on high-ticket items can save hundreds of thousands of dollars in reverse logistics and preserve margins by ensuring products are sold at full price instead of being returned, reprocessed, and potentially discounted.

## How Leading Brands Are Implementing Virtual Try-On

The brands getting this right understand that virtual try-on must embody the same principles of realism over perfection that define their physical products. Here's how leading luxury houses are approaching implementation.

![](https://cdn.sanity.io/images/kqgmf0g2/production/ac99fdc24e43fdd74d8305bb7676cff3de11dc95-1999x1070.png?fit=clip)

**Gucci** has enabled sophisticated try-on across sneakers, accessories, watches, and beauty products with meticulous attention to material authenticity. Their AR rendering accounts for ambient lighting in real-time without looking artificially perfect—the leather on a Gucci loafer reflects light differently across its surface, just as it would in person. For beauty, Gucci offers virtual try-on of lipsticks and makeup through its dedicated beauty VTO platform on mobile devices.

**Chanel** focuses on beauty and accessories, enabling customers to virtually apply lipsticks and eyeshadows with remarkable color accuracy across different skin tones and lighting conditions. This is realism over perfection in practice: Chanel's renderings preserve real skin texture rather than over-smoothing it into synthetic perfection. For eyewear, their try-on feature captures a precise fit for different face shapes—critical in a category where how frames sit on your nose bridge or temples can make or break a purchase decision. 

**Luxottica** has explored multiple angles across its eyewear portfolio, including mobile apps, in-store kiosks, and smart glasses partnerships. Eyewear is particularly well-suited for virtual try-on, as customers need to see how frames complement their face shape, skin tone, and personal style. Their in-store kiosks let customers rapidly try on dozens of frame styles, demonstrating that digital try-on doesn't replace physical retail—it enhances it.

## What It Takes to Implement Successfully

The brands that succeed with virtual try-on share three non-negotiable principles: material fidelity that captures how products actually behave, interface design that feels native to the brand, and seamless integration across the entire customer journey. These aren't optional refinements—they're the foundation that separates implementations that strengthen brand equity from those that damage it.

Rendering quality separates successful implementations from those that damage brand perception. This requires extensive 3D modeling using photogrammetry techniques that capture products from hundreds of angles. Creating these digital assets properly requires significant upfront investment: specialized equipment, skilled 3D artists, and weeks of work per product line. The goal isn't perfection—it's fidelity. 

![](https://cdn.sanity.io/images/kqgmf0g2/production/9539e7101760bf9573222e41c2f94e2d9a5301dd-1600x1052.jpg?fit=clip)

Over-polished lighting, uniform textures, and idealized proportions tend to look synthetic. Imagine a cashmere sweater rendered with perfectly even lighting and no variation in the knit—it looks like CGI, not luxury. Strong implementations capture how products behave in the real world: the way leather reflects light differently across its surface, the natural movement of knitwear on the body, and the visible nuance of shades across different skin tones. That credibility is what distinguishes luxury-grade virtual try-on from mass-market implementations.

These digital assets become valuable inventory that can be reused across try-on, product pages, and marketing campaigns. However, there's no shortcut—low-quality 3D models produce experiences that are worse than having no try-on at all.

### Interface Design Must Align with Brand Language

High-quality 3D assets are necessary but insufficient. The interface itself must translate brand codes into interaction patterns that feel native to the luxury experience—not bolted on as an afterthought.

The interface should translate brand codes into the practical mechanics of try-on:

- **For minimalist brands**: Restrained UI with discreet prompts, few controls, and visual focus on the product
- **For more expressive houses**: Editorial layers and styling context, provided they don't distract from the core decision
- **Path to action**: One-tap camera activation, effortless switching between sizes, colors, and finishes, and simple save-to-wishlist or share-to-consultation flows
- **Continuity with the rest of the site**: Typography, tone of voice, motion, and microcopy should all feel native
- **Interaction patterns**: Mirror what customers already know from your PDPs and checkout to reduce cognitive load
- **Trust signals**: Explicit permissions for camera access, visible loading feedback, and gentle haptics that confirm key actions without breaking immersion

Above all, be selective: features like social sharing or gamification should only appear when they advance the brand's intent. In luxury, sophistication is often about what you remove; keeping the surface area clean allows the craftsmanship, materials, and fit to take center stage, making the experience feel effortlessly premium.

### Mobile Performance Is Mission-Critical

Mobile drives the majority of luxury e-commerce traffic. Performance is non-negotiable. Virtual try-on on mobile presents unique challenges: smaller screens, varied device capabilities, inconsistent camera quality, and suboptimal lighting conditions. The experience must be even more refined on mobile than on desktop, with rendering quality that cannot degrade—a luxury customer browsing on an iPhone expects the same material fidelity they'd see on a desktop.

Mobile-first design means rethinking the entire flow:

- Load times under two seconds
- Seamless camera access
- Flawless one-handed navigation
- Fewer steps to activate the camera
- Larger touch targets
- Simplified controls
- Aggressive optimization for bandwidth constraints

Most implementations fail here—the technology works beautifully in controlled desktop environments, but feels clunky and slow in real-world conditions on actual mobile devices.

### The Customer Journey Must Be Unified

Virtual try-on doesn't exist in isolation—it needs to connect seamlessly across every touchpoint where customers make decisions or seek reassurance.

Virtual try-on should sit exactly where decisions are made and then carry through the journey without friction:

- **On product pages**: Entry point should be prominent but unobtrusive, with states and preferences saved to wishlists so customers can resume their shopping later or discuss their purchases with a sales associate
- **During virtual consultations**: Associates should be able to reference the same try-on view to offer precise fit and styling advice, recreating the intimacy of a boutique experience at a distance
- **In-store**: Pre-built shortlists from at-home sessions can accelerate fittings and boost confidence
- **Behind the scenes**: Tying try-on engagement to personalization, size guidance, and CRM—always with explicit consent—turns a single interaction into a coordinated, cross-channel experience that feels like a natural extension of the brand

### Accessibility Is Non-Negotiable

Credible try-on must perform consistently across diverse customers—not just the narrow demographic often used in technology testing. This isn't just ethical; it's essential to building confidence and reducing returns across your entire customer base.

Requirements for inclusive virtual try-on:

- **For beauty**: Rigorous shade and undertone modeling that accounts for different lighting and camera conditions, with calibration that avoids over-smoothing and preserves real texture
- **For eyewear and jewelry**: Precise scaling, fit guidance, and clear, non-technical explanations that help users understand how size and proportion translate from screen to real life
- **Multiple modes**: Camera-on self try-on, try-on on models with similar features, and accurate static previews for customers who prefer not to use live AR
- **Navigation support**: Ensure each mode can be navigated with keyboard and screen-reader support
- **User control**: Transparent controls to adjust lighting, skin tone matching, and fit markers, so users can fine-tune results to their context rather than accept a one-size-fits-all output

Investing here is both the right thing to do and good business: it reduces friction, broadens reach, and builds trust with audiences often underserved by generic experiences, while improving confidence in purchase decisions and, ultimately, lowering return rates.

## Measuring Success Beyond Conversion Rates

Evaluating the success of virtual try-on requires a connected narrative that ties commercial impact to experience quality and brand equity. In luxury, it can be the right trade-off to accept slightly lower short-term conversions if the experience feels more authentically premium and strengthens brand equity over time. This is the most important tension to understand: **brand perception and conversion aren't always aligned, and in luxury, brand perception often matters more.**

Start with the financials to confirm tangible value—conversion rate lift, return-rate reduction, and AOV growth—then use experience signals to explain why those outcomes occur. Post-interaction surveys should assess whether the experience elevated brand perception, not just whether it drove conversions. 

Track the metrics that reveal whether virtual try-on is reinforcing your luxury positioning:

- **Engagement depth**: Time spent in try-on, number of products and variants explored, completion of key actions such as saving to wishlist or adding to cart
- **Downstream behaviors**: Do users who engage with try-on have higher LTV, lower return rates, or greater repeat purchase frequency?
- **Cross-channel continuity**: In-store visits that reference an at-home try-on shortlist demonstrate how digital confidence translates into physical conversion
- **Brand outcomes**: Pair pre/post brand perception scores (quality, innovation, trust) with sentiment from social and customer service that explicitly references try-on
- **Material authenticity correlation**: Analyze whether improvements in perceived fit and material authenticity correlate with reduced returns and higher AOV

When these layers align—stronger brand scores, deeper engagement, and better unit economics—you have the connective tissue that shows virtual try-on is not merely lifting conversion, but reinforcing the luxury positioning that sustains pricing power and long-term value.

## Technology Should Serve the Brand, Not Define It

AI-driven virtual try-on isn't a trendy feature—it's a tool to increase customer confidence, reduce returns, and strengthen brand positioning. But only when implemented with the same attention to detail that luxury brands apply to every other customer touchpoint. The brands succeeding with virtual try-on share a common principle: technology should serve the brand experience, never define it. The technology should feel invisible, allowing the product and brand to remain center stage. 

The gap between leaders and followers will only widen as technology continues to evolve. Brands that invest strategically now in high-quality implementations will establish competitive advantages that become increasingly difficult to replicate. But implementation quality matters far more than speed. A poorly executed virtual try-on can do more harm than not having one at all. 

The key is finding the right balance between innovation and brand integrity, between cutting-edge technology and timeless luxury values. When you strike that balance, virtual try-on becomes more than a feature—it becomes an extension of what makes your brand special. 

**Need help implementing a virtual try-on experience that actually serves your brand?** Let's discuss creating an experience that maintains your luxury positioning while delivering tangible business results.

Why Luxury Brands Are Betting on AI-Powered Virtual Try-On

Your potential customers are beginning to have conversations about your products with AI assistants like ChatGPT, Gemini, and Perplexity. When someone asks “What are the best wireless earbuds under $200 for running in all weather?” they’re getting instant recommendations complete with comparisons, specs, and reasons why.

[Nearly 60% of consumers](https://news.darden.virginia.edu/2025/06/17/nearly-60-use-ai-to-shop-heres-what-that-means-for-brands-and-buyers/) now use AI to help them shop, with [over half of Gen Z already using AI for product discovery](https://www.salesforce.com/news/stories/consumer-shopping-ai-trends-2025/).

This shift creates a new challenge for e-commerce brands: Generative Engine Optimization (GEO), the discipline of increasing your visibility within AI-powered discovery tools. Unlike traditional SEO, where you can pull up Search Console and see exactly where you rank, AI discovery is opaque. There’s no dashboard tracking how often ChatGPT mentions your brand, and no confirmed ranking factors. In other words, no one really knows what makes one brand show up over another.

Yet some brands consistently appear in AI responses while others don’t, and the industry is starting to make educated guesses about why that is.

In this post, we’ll focus on the high-level strategies for improving your AI visibility across platforms. We’ll explore the foundational tactics that work today and are most likely to continue to work tomorrow, regardless of how AI commerce evolves.

(Note: We’ll dive deeper into ChatGPT’s Instant Checkout feature and what it specifically means for Shopify brands in a separate, more technical post soon—stay tuned!)

## Setting Realistic Expectations About AI Commerce

Before diving into tactics, it’s important to understand where AI-powered commerce actually stands today versus where industry hype suggests it might be.

While nearly 60% of consumers report using AI to help them shop, usage doesn’t yet translate to transactions. A [comprehensive study of 973 e-commerce websites](https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5585812) analyzing 12 months of data (August 2024 — July 2025) with $20 billion in combined revenue reveals that organic LLM traffic currently:

- **Represents less than 0.2% of total e-commerce traffic**—roughly 200 times smaller than Google’s organic search
- **Converts at lower rates than all traditional channels except paid social**, including underperforming organic search, paid search, email, and direct traffic
- **Generates lower revenue per session** than established channels across the board
- **Shows favorable engagement signals** (like bounce rate, indicating relevance) but **shorter session durations and fewer page views**, suggesting users are still learning to trust AI recommendations

**The good news:** The same research shows conversion rates from AI traffic are improving over time, suggesting the channel is maturing even if slowly. The performance gap is narrowing, though [projections indicate it won’t reach parity with organic search within the next 12 months](https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5585812). This is a trend that will only accelerate with ChatGPT’s recent release of Instant Checkout.

In other words, GEO should be viewed as a long-term investment in future discoverability, not an immediate revenue channel. You’re building positioning for when—not if—AI-powered shopping reaches mainstream adoption. The brands that establish authority now will have significant advantages when the channel matures, but don’t expect meaningful traffic or conversion impact in 2025.

Think of this as 2005-era SEO: the fundamentals you build today will compound as the channel matures. The brands that act now—while the channel is still developing and the playing field is relatively open—will likely dominate their categories once adoption reaches critical mass.

## Measuring Your Share of Answer

Before you can improve your AI visibility, you need to understand where you currently stand. But unlike with Google’s Search Console, there’s no dashboard showing how often your brand appears in AI responses.

The solution is a surprisingly trivial “Share of Answer” approach. Here’s how it works.

Start by creating a comprehensive list of 25-50 prompts that represent how customers discover and research products in your category. If you’re struggling to generate this list, consider enlisting an LLM to help: you can provide context about your product category, target audience, and typical customer journey stages, and ask it to generate relevant queries. These should span different stages of the shopping journey:

- For product discovery, include queries like “Best laptop for graphic design students” or “What mattress should I buy for back pain?”
- For comparison and research, test prompts such as “[Brand A] vs [Brand B] comparison” or “Alternatives to Popular Product X.”
- For problem-solution matching, try “What helps with back pain while working from home all day?” or “How to choose the right standing desk.”

Once you’ve built your query set, it’s time to test your LLM by asking it each of the queries you’ve come up with. We recommend testing with ChatGPT, which [holds over 60% of the AI assistant market](https://gs.statcounter.com/ai-chatbot-market-share) and [ranks among the top 10 most visited websites globally](https://www.similarweb.com/website/chatgpt.com/). While it’s worth periodically checking other platforms like Gemini, Perplexity, and Claude to understand the broader landscape, your measurement and optimization efforts should center on where your customers actually are—and right now, that’s predominantly ChatGPT.

Calculate your Share of Answer score: the percentage of relevant queries where your brand receives at least one mention. The scoring breaks down like this:

- **Below 20%:** You’re invisible to the AI. Most queries in your category don’t mention your brand at all.
- **20-40%:** Limited presence. You appear occasionally, but competitors dominate most conversations.
- **40-60%:** Competitive visibility. You’re in the mix, showing up regularly alongside other established players.
- **60-80%:** Category leadership. You appear in most relevant queries, often with favorable positioning.
- **Above 80%:** You dominate AI discovery and are the go-to reference in your space.

When testing each prompt, don’t just check whether your brand appears. Document where it appears in the response, how it’s described, which competitors are mentioned alongside you, and what sources the AI cites. Context matters because it shapes how people perceive your brand.

This baseline becomes your benchmark. Without it, you’re making changes without knowing what’s working. With it, you have a clear starting point and a way to track progress over time.

## Prioritize Product Information Completeness

AI systems strongly prefer brands with detailed product specifications, clear imagery, accurate pricing and availability, comprehensive size guides, and compatibility information. Products with incomplete information are frequently skipped entirely, even from well-known brands.

Think of product data as table stakes. Without structured, complete information, LLMs have a much harder time understanding or recommending your products. To fix that:

- **Complete your specifications.** Every product should have comprehensive details covering dimensions, materials, compatibility, and performance characteristics. Include all details that help customers (and chatbots!) make confident decisions. As an example, if you sell a a standing desk, that means not just height range, but weight capacity, motor type, noise level, and cable management features.
- **Improve your imagery.** High-quality images showing products from multiple angles and in use help both customers and AI systems understand what you’re selling. Include lifestyle shots that demonstrate use cases, detail shots that highlight key features, and size comparison images that provide context.
- **Ensure pricing transparency.** Accurate product pricing, taxes, shipping costs, and availability clearly stated helps AI systems recommend products that customers can actually purchase.
- **Add detailed size and fit information.** For apparel, furniture, and any product where fit matters, comprehensive size guides are essential. Include measurements, fit guidance, and ideally customer photos showing the product on different body types or in different spaces.

Start with an audit of your top products. Where are the gaps? What information is missing or outdated? Prioritize fixing the products that drive the most revenue or that represent your brand in competitive searches. Then expand systematically across your catalog.

## Create Educational Content That Demonstrates Expertise

Educational content demonstrates genuine expertise in ways that product pages can’t. When you publish in-depth guides, detailed comparisons, and thoughtful how-to content, you signal to LLMs that you understand your space deeply, and you’re more likely to show up in long-tail queries.

- **Write comprehensive buying guides.** Address the real questions customers have when choosing products in your category. Not just “which product is best” but “what should I consider?” and “how do different features impact performance?”
- **Answer technical questions thoroughly.** If customers in your category wonder about compatibility, installation, maintenance, or optimization, create content that addresses these topics with real depth. Surface-level answers won’t do, as LLMs tend to prioritize sources that go deep.
- **Create comparison content.** Develop honest comparisons that help customers understand tradeoffs between different product types or approaches. When you can fairly compare alternatives in your category, you build credibility with both humans and machines.
- **Document use cases and applications.** Show how products work in real situations. Case studies, setup guides, and application-specific content help AI systems understand not just what you sell, but how and when customers should use it.

Start with your most frequently asked questions. What do customers want to know before buying? What problems are they trying to solve? What comparisons do they make? Create content that answers these questions comprehensively, and you’ll naturally build the educational foundation that AI systems look for.

## Build Authority Through Authentic Engagement

The tough truth about GEO is that your website is just one data point. What matters more is what the rest of the internet says about your brand.

LLMs train on and reference content from across the web. Even though [ChatGPT is citing Reddit less nowadays, ](https://www.investors.com/news/technology/reddit-stock-chatgpt-advertising-google/?utm_source=chatgpt.com) Reddit content is likely to still carry heavy weight in core model training. The same goes for YouTube videos showing your products in action, or social media content that demonstrates real customer satisfaction with your brand.

When your brand shows up consistently across Reddit, YouTube, industry publications, social platforms, and other sources, ChatGPT notices. It’s looking for patterns of trust across multiple sources. One mention is just noise. Ten consistent mentions across different platforms are signal.

- **Encourage authentic customer sharing.** Make it easy and natural for satisfied customers to share their experiences. Send post-purchase emails with prompts like “How are you using your new standing desk? We’d love to see your setup!” Include shareable moments in your product experience—unique unboxing, thoughtful packaging notes, or surprising product features that make people want to post.
- **Participate meaningfully in community discussions.** Find where your customers congregate online (e.g., industry forums, Reddit communities, Facebook groups) and find ways to amplify those channels. In some cases, this might mean contributing directly. In others, it will mean fostering the organic contributions that already happen (e.g., Reddit is notoriously hostile to brands hijacking the conversation).
- **Create shareable resources.** Develop tools, calculators, guides, or templates that others want to reference and link to naturally because they’re genuinely useful. An ergonomic desk setup calculator, a home office lighting guide, or a remote work productivity toolkit: these resources get shared, cited, and linked to organically, building the distributed presence that LLMs will recognize.

In other words, focus on building genuine value where your customers already spend time. The visibility will follow naturally.

## Monitor Your Progress and Iterate

Just like SEO, GEO isn’t a one-time project, but rather an ongoing practice that requires regular measurement and iteration. The AI landscape evolves quickly, and what works today may need adjustment tomorrow:

- **Re-run your Share of Answer analysis quarterly**. Track how your optimization efforts are actually impacting AI visibility. Use the same query set you established in your baseline, but don’t be afraid to add new queries as customer behavior evolves or as you expand into new product categories.
- **Track quality, not just quantity.** Look not just at whether your Share of Answer score is improving, but at how the nature of your mentions is changing. Are you appearing earlier in responses? Are the descriptions more accurate? Are you being mentioned alongside stronger competitors?
- **Test new approaches systematically.** When you hear about a new tactic or strategy, don’t just implement it everywhere. Test it on a subset of products or categories, measure the impact on your Share of Answer, and expand only if you see positive results. A disciplined approach prevents wasted effort on unproven tactics.
- **Document what you learn.** Keep a log of changes you make, results you observe, and patterns you notice. Over time, you’ll build institutional knowledge about what works for your specific brand, category, and customer base. This documentation will be invaluable when new team members join or when you need to justify continued investment.

Rather than implementing every possible tactic, you’ll want to systematically improve your Share of Answer score over time while building assets that serve customers across all discovery channels.

## Building AI Visibility Is Building Future-Proof Discovery

As covered in the expectations section above, AI-powered product discovery is still early-stage but showing clear momentum. The trajectory is undeniable: [Gartner predicts that traditional search engine volume will drop 25% by 2026](https://www.gartner.com/en/newsroom/press-releases/2024-02-19-gartner-predicts-search-engine-volume-will-drop-25-percent-by-2026-due-to-ai-chatbots-and-other-virtual-agents), and conversion rates from AI traffic are [steadily improving over time](https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5585812).

This creates a strategic window. The brands that build AI visibility now—while the channel is still developing—will likely dominate their categories once adoption reaches critical mass.

To do that, start with your baseline measurement, then systematically work through product data completeness, educational content creation, and authentic engagement. Finally, monitor quarterly and iterate based on what you learn. The approach is straightforward, but it requires commitment—a trait that is quite rare in an industry that tends to be very short-term oriented.

Need help navigating this transition? At Nebulab, we work with mid-market and enterprise e-commerce brands to build the technical infrastructure that drives visibility across all discovery channels—including emerging AI channels. We take a pragmatic, data-driven approach to GEO that balances long-term positioning with near-term ROI from your core acquisition channels. We can help you measure your baseline, prioritize optimization opportunities, and build a systematic approach to GEO that integrates with your broader e-commerce strategy. [Get in touch](https://nebulab.com/contact-us/) to discuss how we can help your brand show up in the conversations that matter most.

The Nebulab

Shopify's Agentic Strategy and the Future of DTC

Shopify's Winter 2026 Edition: The Needle Movers and the Missed Opportunities

Why Luxury Brands Are Betting on AI-Powered Virtual Try-On

The Early Mover’s Guide to AI Product Discovery