When it comes to ecommerce product discovery, most online stores still expect you to know what you want before you arrive. You type a keyword into a search box, scroll a grid of thumbnails, apply a few filters, and hope. That worked when catalogues were small. It breaks down completely when a shopper is browsing for inspiration, can't name what they're looking for, or simply wants to see what's possible.
Product discovery — the journey from "I'm interested" to "this is the one" — is where ecommerce is changing fastest. AR, visual search and AI personalisation are converging into something that looks less like a catalogue and more like a knowledgeable shop assistant. Here's how it works, what's real today, and what retailers should actually do about it.
Why Product Discovery Is Broken on Most Stores – Ecommerce Product Discovery
The standard discovery toolkit — keyword search, category trees, faceted filters — assumes the shopper can describe the product in words the catalogue understands. In practice, shoppers describe things vaguely ("a nice lamp, kind of industrial, not too big"), misspell terms, or don't know the category name at all. Internal site search on many stores is genuinely poor: it matches keywords rather than intent, buries relevant products, and gives up entirely on synonyms.
The cost is invisible but large. Shoppers who can't find what they want don't complain; they leave. Retailers see it as bounce rate on category pages and search exits, rarely diagnosing the cause. Before investing in anything futuristic, it's worth saying plainly: fixing basic search relevance and navigation usually pays back faster than any AI feature. Discovery innovation builds on solid foundations — it doesn't replace them.
Visual Search: Point Your Camera Instead of Typing
Visual search flips the input: instead of describing a product in words, the shopper shows a picture. Snap a photo of a chair in a café, upload it, and get visually similar products from the retailer's catalogue. Pinterest, Google Lens and Amazon have trained shoppers to expect this, and the technology — image recognition matched against a product database — is mature. Branded AR filters on social platforms extend visual discovery further, letting shoppers encounter products through playful camera experiences that link back to the catalogue.
For retailers, visual search works best in visually-driven categories: furniture, fashion, décor, eyewear. The requirements are straightforward but non-trivial: clean product imagery (consistent backgrounds, multiple angles), rich visual tagging, and a search index that compares images rather than text. The quality of the results is almost entirely a function of catalogue data quality — which is why retailers with messy product data get disappointing results and blame the tool.
Visual search also pairs naturally with AR: find a similar product visually, then place it in your room to confirm. That two-step — visual match, then spatial validation — is becoming the standard discovery flow for home and furniture retail, and it's worth designing for deliberately rather than bolting together afterwards.
AR as a Discovery Channel, Not Just a Validation Tool
AR is usually discussed as the last step before purchase — the "check it fits" moment. But it's increasingly a discovery mechanism in its own right. When a shopper can place products in their space with one tap, browsing becomes experimental: try the blue sofa, then the green one, then the whole styled set. Each placement is a micro-discovery that flat thumbnails can't deliver, because the shopper sees the product in their context rather than the retailer's.
This changes what "browsing" means. WebAR experiences embedded in category pages let shoppers audition dozens of products in minutes, filtering by feel rather than by specification. Retailers we've worked with find that AR entry points on collection pages — not just product pages — materially change browsing behaviour: shoppers explore more of the catalogue because each exploration costs one tap instead of a page load and a leap of imagination.
The implication for catalogue strategy is significant. Products that were hard to sell through thumbnails — unusual shapes, bold colours, items whose appeal is spatial — become discoverable once shoppers can try them in situ. AR doesn't just validate choices; it widens the consideration set.
AI Personalisation: From Segments to Individuals
Recommendation engines are old news — "customers also bought" has been around for decades. What's changed is the granularity. Modern personalisation builds a taste profile per shopper from browsing behaviour, purchase history and stated preferences, then shapes discovery around it: the homepage, the search ranking, the email content and the "complete the look" suggestions all reflect that individual.
Done well, this feels like a good shop assistant who remembers you. Done badly, it feels like being stalked by a pair of shoes you looked at once. The difference is restraint and transparency: explain why something is recommended, make it easy to dismiss, and don't personalise so aggressively that the shopper never encounters anything new. Discovery needs serendipity — a feed of pure "more of the same" is a filter bubble that shrinks the catalogue over time.
The honest limitation: personalisation quality depends on data volume. It works brilliantly for retailers with large catalogues and repeat traffic, and poorly for small stores where every shopper is effectively new. Smaller retailers should invest in excellent guided discovery (see below) before chasing algorithmic personalisation — it's cheaper and often more effective at their scale.
Guided Discovery: Quizzes, Finders and Conversational Browsing
Some products need a conversation, not a search box. Skincare, mattresses, running shoes, complex electronics — categories where the right choice depends on the shopper's situation — benefit from guided discovery: short quizzes or chat-led flows that ask a few questions and narrow the catalogue to a sensible shortlist.
This is unglamorous, highly effective, and entirely within reach for any retailer. A well-designed finder that asks four questions and recommends three products will outperform a clever AI on a messy catalogue every time. The questions encode the expertise of the retailer's staff; the tool just delivers it at scale. Combine the output with AR validation — "here are the three sofas for your space and budget, now see each in your room" — and you have a discovery journey that genuinely mirrors good in-store service.
Conversational interfaces powered by large language models are the newest variant. They handle open-ended questions well ("I need a gift for someone who likes cooking but has a tiny kitchen"), but they need careful grounding in the actual catalogue — an AI that confidently recommends products you don't stock is worse than no AI at all. Retrieval-augmented setups that constrain the model to real inventory are the responsible way to deploy this.
What This Means for Your Catalogue and Content
Every discovery technology above runs on the same fuel: structured, rich product data. Visual search needs consistent imagery and visual attributes. Personalisation needs behavioural and product-attribute signals. AR needs 3D models and accurate dimensions. Conversational discovery needs product knowledge it can quote accurately.
That means the highest-return investment most retailers can make in "the future of discovery" is boring: clean up product data, standardise imagery, write attributes properly, and build the 3D asset pipeline once rather than per campaign. Our AR product visualisation work starts with exactly this foundation, because the fanciest AR experience in the world can't sell a product whose dimensions are wrong in the model.
It also means thinking of content as discovery infrastructure. Buying guides, comparison content, styled looks and video demonstrations are what feed both human browsers and AI systems. A product with a good guide, clean data and a 3D model is discoverable through every channel; a product with a thumbnail and a two-line description is discoverable through none.
A Realistic Roadmap for Retailers
You don't need all of this at once. A sensible sequence:
- First, fix the basics. Audit internal search relevance, navigation and filtering. This is where most discovery value is currently leaking.
- Second, build the data foundation. Standardise imagery, complete product attributes, and start 3D modelling your hero products. This unlocks everything downstream.
- Third, add one discovery upgrade matched to your category: visual search for visually-driven catalogues, a guided finder for considered purchases, AR placement for spatial products, or virtual try-on for personal products.
- Fourth, connect the steps. Visual match → AR validation → personalised follow-up is a journey; design it as one, with consistent data throughout.
- Finally, layer personalisation once you have the traffic and data volume to make it meaningful.
Each stage should pay for the next. If stage one doesn't move search-to-purchase conversion, don't buy stage three — fix stage one properly first.
Product discovery is becoming the real battleground in ecommerce — not price, not even product, but how easily the right shopper finds the right thing. If you'd like an honest assessment of where your store's discovery is leaking value and what to fix first, talk to our team for a free consultation.
Frequently Asked Questions
What is ecommerce product discovery?
Product discovery is the process by which shoppers find products they want to buy — through search, browsing, recommendations, visual search and increasingly AR. Good discovery matches shoppers to products even when they can't describe exactly what they're looking for.
How does visual search work for ecommerce?
The shopper uploads or snaps a photo, image recognition analyses it, and the system returns visually similar products from the retailer's catalogue. It works best with clean, consistent product imagery and rich visual tagging.
Is AI personalisation worth it for small retailers?
Usually not as a first investment. Personalisation needs significant traffic and data volume to work well. Smaller retailers typically get better returns from guided discovery tools like product finders and quizzes, plus solid search and navigation.
How does AR help customers discover products?
AR lets shoppers audition products in their own space with one tap, which makes browsing experimental — they try more products because each try is cheap. It also makes hard-to-photograph products (unusual shapes, bold colours) discoverable by showing them in context.
What's the first step to improving product discovery?
Fix the foundations: internal search relevance, navigation, filtering and product data quality. Discovery innovation built on messy data disappoints; the same technology on clean data transforms the experience.
For more information, see Shopify Enterprise.