2026-07-06 · agent-readiness · the next visibility layer

From ungooglable to unagentable

Search made products findable. Agents ask a harder question: can this offer be understood well enough to enter the decision?

TL;DR — For years, ecommerce visibility was not about being online — it was about whether a search engine could represent you well enough to rank you. Almost anything can be crawled; being legible was the real problem. AI agents raise the same problem at a new layer. An agent has to understand an offer, compare it, and act on it under a budget. A catalog can be crawled, indexed, and still lose consideration because the agent cannot cheaply or confidently turn it into commerce state. Call that being unagentable.

For years, ecommerce visibility was not really about being online.

Being online was the minimum condition. The real question was what Google could make of you.

Almost anything can be crawled. A weak product page, a badly structured category, a thin description, a duplicated listing, a messy URL: all of it can end up somewhere in the index. But being in the index is not the same thing as being visible.

The real problem was representation.

Could the search engine understand what the page was about? Could it associate the product with the right queries? Could it distinguish the canonical version from duplicates? Could it extract price, availability, reviews, brand, category, and intent? Could it trust the page enough to rank it where a buyer would actually see it?

That is what ecommerce SEO was really about.

Not simply making products indexable, but making them legible enough to be represented, ranked, and selected inside the search layer.

A store could exist, be crawled, and technically be indexed, while still being commercially invisible.

Call that the old problem: ungooglable.

Not absent from the web. Absent from meaningful discovery.

The new visibility problem

AI agents introduce a similar problem, but at a different layer.

An agent does not just retrieve a list of pages and hand the work back to a human. It has to understand the offer, compare alternatives, reason about constraints, check availability, interpret variants, evaluate tradeoffs, and decide what is worth showing, recommending, or buying.

That means the question changes.

It is no longer only:

Can this product be found?

It becomes:

Can this offer be understood well enough to enter the decision?

This is where many ecommerce catalogs fail.

Not because they are offline. Not because they cannot be crawled. Not because the product data does not exist.

They fail because the data is not shaped for reasoning.

Call it unagentable

Unagentable — a catalog is unagentable when it is available to machines, but too noisy, ambiguous, incomplete, inconsistent, or expensive for an AI agent to understand and act on with confidence.

This distinction matters.

In search, a bad page could still be indexed but rank poorly because the search engine could not extract a strong enough representation of it.

In agentic commerce, a bad catalog can still be accessible but lose consideration because the agent cannot cheaply or confidently turn it into commerce state.

That is the new visibility problem.

Old problem — ungooglable — crawled, but not represented well enough to rank New problem — unagentable — accessible, but not computable well enough to be chosen

Most ecommerce platforms were not designed for this

Most ecommerce platforms were designed to render storefronts, power admin panels, feed marketplaces, support plugins, synchronize inventory, and expose generic APIs.

That history leaves traces in the data.

Verbose payloads. Duplicated fields. Platform-specific attributes. Variant information split across names, options, metadata, and descriptions. Stock signals that are technically present but not decision-ready. Shipping rules hidden in policy pages. Product descriptions written to persuade a human, not to support machine comparison.

A human can often resolve this with a glance.

An agent cannot glance.

It has to spend context, tokens, latency, and reasoning effort. Every ambiguity becomes either a cost or a risk. If size is buried in prose, it has to infer. If availability is unclear, it has to hedge. If variants are inconsistent, it has to reconcile. If the payload is full of irrelevant fields, it spends attention before reaching the signal.

This is why "more data" is not the answer.

The three resolutions of truth

Agents do not need the entire platform object every time they look at a product.

They need the right resolution of truth for the step they are on:

Discovery
Compact summaries
Comparable at a glance. Enough to rank a candidate, honest about what it omits.
Comparison
Normalized facts
Attributes, constraints, stock, price, shipping signals, provenance — on comparable terms.
Decision
The full record
Live availability, terms, policies, and canonical paths to act — for the finalist only.
Not compression — a different contract: the right resolution of truth for the step the agent is on.

That is not compression.

It is a different contract.

The new visibility layer

SEO made products legible to search engines.

Agent-readability makes offers computable for decision-making systems.

And computable does not mean simplified. It means structured enough to be compared, constrained, trusted, and acted on.

The catalog stops being only a set of pages to render.

It becomes a reasoning surface.

This applies beyond WooCommerce. WooCommerce makes the problem obvious because it is open, extensible, plugin-driven, and often messy. But the same issue exists across Shopify, Magento, PrestaShop, custom ecommerce stacks, marketplace feeds, ERP exports, and headless systems.

Different platforms produce different kinds of noise.

The underlying problem is the same: ecommerce data was built for pages, feeds, dashboards, integrations, and search engines, not for agents that must make decisions under budget.

Search is not going away. Stores still need to be found. Ranking still matters.

But being found is no longer the whole game.

A product can be crawled.
A page can be indexed.
A store can rank.
An API can respond.

And the offer can still be too hard for an agent to use.

The next visibility problem is not simply being ungooglable.

It is being unagentable.

← All posts