Can AI See Your B2B Catalog? The Two Walls That Keep It Hidden

For twenty years, getting found meant ranking on a search results page. A buyer typed a query, scanned ten blue links, and clicked. Everyone optimized for that moment.

That crucial moment is being replaced. Buyers now ask an AI assistant, and the assistant answers with a shortlist, not a list of links. It names a few suppliers, describes a few products, and the buyer moves on from there. If your products are not in that answer, you are not in the running, and you may never see the traffic you lost.

For most large B2B catalogs, the reason is rarely weak SEO or thin content. It is that two specific barriers sit between the catalog and the AI model, and most teams are not looking at either. We mapped both in a new playbook produced with Prerender, and this article is the introduction to the full picture: what the two walls are, why B2B makes them taller, and how to tell which one is costing you visibility right now.

1. The Two Walls Between Your Catalog and AI
2. Why B2B Catalogs Are Especially at Risk
3. How to Tell If AI Can See Your Catalog
4. The Fix Is Not “Open Everything”
5. What to Do This Week
6. Your Questions, Answered

A machine can only recommend what it can reach and read. For most large B2B catalogs, it can do neither.

Wall 1: AI cannot read it. Modern catalogs run on JavaScript to give human shoppers a fast, dynamic experience. The problem is that AI crawlers largely do not execute JavaScript. When one lands on a JavaScript-heavy or single-page catalog, it often sees a near-empty shell: no products, no prices, no specs. The content is there for people, and invisible to machines. This is not a content-quality problem. The product description could be perfectly written and still never reach the crawler, because from the crawler’s point of view it was never sent.

Wall 2: AI cannot reach it. Even when a page renders perfectly, the parts that matter most in B2B are usually not on it. Prices, availability, assortments, and contract terms sit behind a login. The open web that feeds AI models never gets there. A single part number can be one price for one customer and another for the next, in stock in one region and unavailable in another. In B2B, a product does not fully exist until it sits inside a specific buyer, contract, and region.

Clear both walls and you are present in the answer. Clear neither and you are invisible, no matter how good your catalog is.

Two things make these walls taller for B2B than for a consumer store.

The first is scale. A catalog with millions of SKUs can be readable and still go substantially under-crawled, because no crawler will fetch every page. Crawlers work on a limited budget, so every page they do visit needs to deliver its content quickly and completely. Slow or empty pages waste those visits and leave more products undiscovered, unindexed, and missing from AI answers.

In one sample of roughly 3 million live product and category pages across 18 enterprise retail catalogs, 93% took longer than 5 seconds to fully load their content. Pages that heavy are at real risk of being indexed incomplete, or not at all.

The second is context, and this is the part specific to B2B. Because so much of the commercially important data is customer-specific and gated, the instinct is often to protect it by sharing as little as possible. That instinct is understandable, and it quietly hands the narrative to someone else. If the only product data an AI model can reach lives on your distributors’ sites, your distributors become the source of truth for your own catalog, and you inherit whatever they get wrong.

This plays out differently by industry. In auto parts, the pressure is mostly readability: enormous catalogs that have to be present for machines and fast enough to be crawled at scale. In fast-moving consumer goods, the pressure is mostly access: personalized pricing and gated assortments are the norm, not the exception, which is exactly where Wall 2 bites. If you sell in FMCG, our FMCG industry hub covers how these catalogs modernize without a full rebuild.

Start with the outcome. Pick 10 buying-intent queries for your category, the kind a real buyer would type, and run each through two or three AI assistants. For each one, record three things:

• do you appear at all,
• who gets named instead of you, and
• when you do appear, is your product described correctly.

That is your starting position, and the first look is usually uncomfortable.

If you come up short, three quick checks tell you whether a rendering problem is the reason.

• Disable JavaScript in your browser and reload a product page; whatever disappears is roughly what a crawler sees.
• Check Google Search Console for pages stuck in “Discovered, currently not indexed” or “Crawled, currently not indexed,” which is a strong signal of a rendering issue at volume.
• And check your average crawl response time: anything consistently over 500 milliseconds is high, and slow pages get re-crawled and re-indexed less often, so your AI-visible data drifts further out of date with every cycle.

The manual test builds intuition, but it is a snapshot, and AI visibility can change without warning. The practical next step is to put the monitoring on autopilot, so you catch a drop when it happens, not a quarter later when someone asks why traffic fell.

Once you can see the problem, the temptation is to swing to the other extreme and expose the whole catalog. In B2B, that is the wrong move, because most of your catalog cannot be published. It does not exist outside a buyer relationship.

The better approach is to think in two layers:

The public layer is what any visitor, human or machine, can see without logging in: that a product exists, what it is, its category, its core specifications, and general availability. This is the layer that has to render cleanly for crawlers and agents, and it is enough for an AI model to shortlist you.

The private layer is everything that only makes sense in context: customer-specific pricing, negotiated terms, contract assortments. That layer should never reach a crawler. It is not content, it is a commercial agreement.

Drawing that line is deliberate work, attribute by attribute. Expose too little and AI cannot shortlist you. Expose too much and you leak pricing or break a distributor agreement. It is also worth saying plainly that this is not cloaking. Everyone, human or machine, sees the same public layer; the commercial layer opens only after authentication, for people and authorized agents alike.

On the readability side, the mechanical fix is well understood. A prerendering layer sits between your site and incoming crawler requests, detects a crawler, and serves it a fully rendered page with products, specs, and categories already in the HTML. Human visitors never see that version; they get the same fast site as always. Virto Commerce works with Prerender exactly here, so the public catalog renders for machines while the commercial layer stays governed by the platform. You can see how that fits together on the Virto and Prerender integration page.

On the access side, the deciding factor is architecture. When catalog, price lists, and contracts are separate entities rather than fields stitched onto one product record, a single SKU can carry several valid versions at once, and the platform decides which one each visitor is entitled to see. A crawler gets the lean public version; an authenticated buyer or their authorized agent gets pricing and contract terms. The boundary is enforced by the platform, not rebuilt by hand for every new channel. It is the same logic B2B has used for machine buyers for years through punchout: first comes who you are and what you are allowed to see, then comes the content.

You can make real progress without a big program.

• Baseline your AI visibility with the ten-minute test.
• Diagnose which wall is hurting you: readability, reachability, or both.
• Confirm whether your catalog renders full HTML to bots today.
• Draw a first version of the line between your public and private layers.
• And decide who owns AI visibility going forward, because right now, in most companies, no one does.

Being visible to machines is the entry ticket, not the finish line. In B2B, the finish line is being visible on your terms: discoverable enough to make the shortlist, governed enough to keep your pricing and contracts where they belong.

The full playbook walks through both walls in detail, with the diagnostics, the case examples, and a readiness checklist you can run against your own catalog.