Menus, collections, search bars, filters, breadcrumbs, and category pages exist because humans need ways to move through a catalog. They reduce the complexity of many products into understandable paths.
AI shopping agents introduce another navigation model: the customer can describe intent in natural language and let software traverse the commercial information on their behalf.
Does that mean navigation is dying? No. But navigation is becoming one interface to an information architecture rather than the information architecture itself.
Human navigation is a compression system
A merchant may have thousands of products and attributes, but a header can expose only a small number of choices. Collections and filters progressively narrow the catalog until the customer finds a relevant subset.
This works because the merchant predicts the dimensions customers are likely to use: category, price, color, size, brand, use case, and so on.
An agent can begin with a more complex combination of constraints at once. It can interpret “I need a compact option for travel, compatible with this model, under this budget, delivered next week” without forcing the customer through four separate navigation steps.
Intent can become the navigation layer
In an agentic journey, natural language functions as a dynamic filter. The agent converts the customer's request into attributes, conditions, and relationships, then searches for matching candidates.
This does not eliminate taxonomy. In fact, it makes good taxonomy more valuable. The agent still needs products to have categories, attributes, relationships, and states that correspond to the customer's intent.
The visible menu can become less important while the semantic information architecture underneath becomes more important.
Collections still matter—but their meaning should be clear
Collections remain useful for merchandising, campaigns, browsing, and human orientation. But merchants should distinguish between collections that express a genuine semantic grouping and collections that exist only as a temporary visual arrangement.
An agent benefits from understanding why products belong together: same category, same use case, compatible ecosystem, seasonal context, audience, price tier, or another meaningful criterion.
Filters are clues to the machine model
If customers frequently filter by an attribute, that attribute is probably important enough to model clearly in product data.
Size, material, capacity, compatibility, fit, technical specification, dietary property, service scope, or delivery eligibility should not exist only as frontend filter labels if they materially affect product choice.
Agentic commerce encourages merchants to treat filters as representations of real product meaning.
Search becomes conversational, but structured data still wins
Conversational search can make discovery feel dramatically simpler. Yet the quality of the answer depends on the quality of the underlying commerce information.
A natural-language interface cannot reliably compensate for a catalog whose attributes are missing, inconsistent, or embedded only in images. Better interaction does not remove the need for better information architecture.
Design for two navigation modes
Shopify stores increasingly need to support:
- Human-led navigation: menus, categories, collections, search, filters, links, and visual merchandising.
- Agent-led navigation: intent interpretation, semantic matching, relationships, constraints, and machine-readable actions.
These modes should converge on the same products and commercial truth.
What should merchants change?
Do not remove useful menus because AI exists. Instead, inspect the architecture underneath them.
- Make categories explicit.
- Model attributes customers actually use to choose.
- Clarify product relationships and compatibility.
- Reduce duplicate or contradictory classification.
- Ensure collection names describe meaningful concepts.
- Make important destination pages understandable when reached directly.
- Test natural-language customer intents against the catalog.
Navigation is becoming adaptive
Static navigation asks every customer to traverse the merchant's predefined path. Agentic navigation can construct a path around the customer's particular intent.
That does not make information architecture less important. It makes it more fundamental, because the agent needs an accurate map even if the human never sees the map directly.
The future is therefore not “no navigation.” It is navigation above a semantic commerce model—with humans browsing visually and agents traversing meaning.
For the changing role of the storefront's traditional entry point, read The Shopify Homepage After the Homepage. The next five articles move from theory into the advanced MX operating layer: auditing, technical debt, structured data, recommendation optimization, and the agent-ready product page.
