How to Tell If Your Products Are Invisible to ChatGPT

A customer asks ChatGPT to recommend a running shoe under $150. Or a standing desk for a small apartment. Or the best sunscreen for sensitive skin.

Some products get recommended. Three or four of them, usually, with a short explanation for each. No ads. No blue links. Just an answer.

If your product is not one of them, you have not lost a ranking. You were never in the consideration set.

OpenAI’s own economic research team has cited around 50 million daily shopping queries in ChatGPT alone. The Google, Perplexity and Amazon AI assistants also have their own numbers. This is not an emerging channel that you can explore next year.

I do ecommerce optimization audits for clients, and I see the same problems over and over again in almost every project. They are not obscure. They are often hard to see.

Let’s start with the 30-second test: Go to your top-selling product page. Click View Page Source, not Inspect Element. Find your product’s price tag in the source code.

If it is not there — you guessed it — most AI systems cannot see it there either

Why? Google figured out this issue years ago. It executes JavaScript used on modern sites to generate prices, inventory, specifications and reviews. The AI crawlers generally do not. They read the initial HTML and stop. Research by Vercel and MERJ shows that the large majority of AI crawlers cannot execute JavaScript at all.

As a result, you get a catch for many good teams — your page ranks well in Google. It returns nothing to AI assistants when they try to read it. Content is visible to humans and Google, but not to anything else.

The solution is simple: Discuss it with your development team. It’s expensive on some platforms, but you don’t have a choice if you care about this channel.

If you sell exclusively on Amazon, you have a serious problem

Amazon has spent the last year expanding the list of AI bots listed in the robots.txt file — a standardized format for letting automated visitors know what they can access. Currently, this list includes the search and browsing agents of OpenAI, Anthropic and Perplexity. You can view it yourself at amazon.com/robots.txt.

The consequence is that Amazon product listings are invisible to ChatGPT’s organic shopping recommendations.

It makes perfect sense from Amazon’s standpoint. It spent two decades controlling the moment of a purchase decision. An AI assistant that compares products across all stores makes this advantage disappear. So Amazon restricts external agents while developing its own, currently available via Alexa for Shopping.

Think about the implications for a brand relying exclusively on Amazon listings: no direct-to-consumer product pages. No press coverage. No community discussion. Such a business is close to invisible in a channel with tens of millions of shopping questions every day.

Two consequences here: First, having a product page on your own site is no longer a debate about channels. It is about infrastructure. Second, selling on Amazon has turned into an optimization problem of its own with its own rules, because its AI assistant reads listing copy, reviews and customer questions to make decisions.

And here is one important thing you should know, too: Walmart, Target, Best Buy, Home Depot and Etsy made a different decision and allowed AI shopping surfaces to index their listings. If you sell through those retailers, their listings can help you reach an audience that Amazon is hiding from.

It’s also worth noting that this is not permanent. The CEO of Amazon has stated publicly that the company is negotiating with third-party shopping agents and is sure they’ll find a way to partner eventually.

Third-party reviews are a key gatekeeper

This is the insight that is most likely to change how you operate on Monday.

Trustpilot commissioned research from Seer Interactive analyzing over 800,000 AI answers on ChatGPT, Gemini, Perplexity and Google’s AI Mode, based on over 15,000 buying-journey prompts and roughly 1,900 brands.

Brands without a third-party review profile showed up in 1% of AI answers. Brands with one to 13 reviews — in 53%. Brands with active profiles and regular response rate — in 75%.

Trustpilot funded this study, so take the exact numbers with a grain of salt. But look at the graph — this is the interesting part here. Almost all of the difference happens from zero reviews to a handful of reviews. This is not a volume game. This is a presence game.

Which means the barrier is a lot lower than most business owners assume. You don’t need thousands of reviews. You just need to be present in two to three review sites relevant to your niche.

AI amplifies criticism, but does not average it

Search-conditioned business owners tend to think about the rating as an average. Some bad reviews are diluted by a lot of good ones.

Not so with generative AI. Search engines average the data. AI assistants tell a story. As the Seer research showed, even in the case of overwhelmingly positive signals, AI systems surfaced the outliers.

A vivid complaint can shape the AI description of your brand for a potential customer about to purchase.

This creates three priorities for your business:

Collect reviews constantly, not periodically, because recency counts, and a burst of six-month-old reviews is aging out.

Answer public complaints, pointing out the specifics of your situation, since it shows that your business is responsible to both people and AI.

And don’t try to remove the negative reviews, because this practice has both legal and technical risks.

And one more thing: Ask customers for details, not just ratings. Two hundred reviews mentioning fit, size, use case and durability provide the AI system with valuable information. Five hundred one-liners provide very little.

And here is one last point about where not to invest

In late 2025, OpenAI introduced a checkout-in-chat feature, and the retailers rushed to integrate. In March 2026, OpenAI shut down the checkout and moved to product discovery. Meanwhile, Google chose the opposite route and introduced a cross-retailer cart featuring Nike, Sephora, Target and Walmart.

The takeaway is not about the victory of one approach over the other. It is about the fact that the destination changes, but the underlying requirements are always the same. Complete product information. Real-time accurate pricing. Current inventory. Pages readable by machines.

The businesses that optimized their product data for a feature that got cancelled got all the benefits. Those that integrated with a specific destination wasted their money.

Start with View Page Source. The rest will follow from there. For a more comprehensive guide, follow this eCommerce AEO playbook I have put together. This is part of the audit and strategy we do in-house for our own ecommerce properties and for our clients.

A customer asks ChatGPT to recommend a running shoe under $150. Or a standing desk for a small apartment. Or the best sunscreen for sensitive skin.

Some products get recommended. Three or four of them, usually, with a short explanation for each. No ads. No blue links. Just an answer.

If your product is not one of them, you have not lost a ranking. You were never in the consideration set.