Your store is invisible to ChatGPT (here's the fix)

It's December 18th. A woman opens ChatGPT and types "best Christmas gift for a dad who likes cooking, under $80." ChatGPT shows her a clean carousel of seven products with prices and one-line descriptions. She clicks the third one. By the t

It's December 18th. A woman opens ChatGPT and types "best Christmas gift for a dad who likes cooking, under $80." ChatGPT shows her a clean carousel of seven products with prices and one-line descriptions. She clicks the third one. By the time she's clicked, your competitor has a sale. You sell exactly the same product. Yours isn't in the carousel.

The carousel isn't a search-engine results page. It isn't Google. It isn't a Reddit thread or a Facebook ad. It's a new surface entirely — one that picked seven products out of millions in roughly 800 milliseconds, and your store was not one of them. You weren't outbid. You weren't outranked. You were invisible. ChatGPT shopping has been running this carousel for nearly a year, and most stores I audit still haven't set up the basics that would put them inside it.

ChatGPT shopping is the only acquisition channel in 2026 with zero ad spend, zero auction, and a feed that 80% of merchants haven't even submitted. The window will close. It's open now.

ChatGPT is the new Christmas channel, and your store probably isn't on it

The shape of consumer gift-shopping shifted hard in 2025 and most retailers missed it. A meaningful share of "what should I buy for X" queries that used to start on Google now start in ChatGPT. The conversion behavior is different too — when an AI assistant recommends a product, the click-to-buy rate runs higher than a Google search result because the buyer arrived already convinced. They aren't comparing ten tabs. They're looking at three carousel cards.

The mechanic for getting into that carousel is mostly mechanical. ChatGPT pulls products from a few sources: an OpenAI-managed merchant feed (early access, waitlist-gated), a direct Shopify integration for stores already on that platform, and a web crawl conducted by an OpenAI bot called OAI-SearchBot that reads product pages and their structured data. The ranking layer on top weighs relevance to the query, structured-data quality, price competitiveness, and a social-proof signal pulled from places like Reddit and TikTok where the product is being discussed.

None of this is mystery. The pieces are documented. The reason most stores aren't in the carousel is that they never executed the setup, not that the setup is hard.

The OAI-SearchBot crawler is the gate, and most stores have it blocked

Open your store's robots.txt right now. If it looks like the default Shopify or WooCommerce file, you probably have User-agent: * followed by a Disallow directive somewhere that blocks at least one path the bot needs. Some stores block all User-agent: * traffic by default; some have legacy rules from a 2022 SEO audit that explicitly disallow unfamiliar bots.

The fix is one rule. Add the following block at the top of your robots.txt:

User-agent: OAI-SearchBot
Allow: /

That's it. No tokens, no auth, no API. The bot is documented in OpenAI's bots page along with two other crawlers — ChatGPT-User (used for user-initiated browse requests inside a chat) and GPTBot (used for training, which you can separately allow or disallow based on your training-data preferences). The shopping crawler is the first one. Don't conflate them.

I've audited fourteen stores in the last three months that thought they were "open" to ChatGPT and weren't. The blocker was always in robots.txt, sometimes inherited from a 2019 sitemap-generator plugin, sometimes from a security audit that wholesale-blocked anything not whitelisted. Two minutes to fix. Two minutes that 80% of operators haven't yet spent.

The four things ChatGPT actually pulls when it picks products

The carousel-selection logic isn't documented in full, but the inputs that demonstrably move the needle are these four.

Product structured data. ChatGPT reads JSON-LD Schema.org/Product markup on every product page. Title, description, price, availability, image URL, brand, GTIN if you have one, aggregateRating if you do. Without this, the bot has to infer everything from page content, which it can do, but the listing decision degrades fast when the inference is fuzzy. Schema.org markup is the single highest-leverage technical lift in the ranking flow.

The merchant feed. If you submit a product feed through OpenAI's early-access merchant program — which is the on-ramp non-Shopify stores use — the feed is the canonical source of truth for your products. ChatGPT trusts the feed over the crawl because it's cleaner. Shopify stores get a version of this for free because Shopify exports a feed to OpenAI automatically.

Price. Lower price doesn't beat higher price. Reasonable price for the category does. Outliers in either direction get suppressed because ChatGPT's safety layer doesn't want to recommend obviously-too-cheap or obviously-too-expensive products and absorb the customer-service blowback. Stay inside the category's competitive band.

Social-proof signals. This is the part most "ChatGPT SEO" guides miss. ChatGPT weighs whether your product is discussed positively on Reddit, in HN threads, on TikTok, in published reviews. The mechanism is the same one we've covered in the signal economy frame — public discussion of your product creates a credibility signal that the model picks up. A Shopify store with five strong Reddit threads about its product will out-rank an identical store with no community footprint, all else equal.

The Shopify path is automatic. The non-Shopify path is the waitlist

If you're on Shopify, most of the work is already done for you. Shopify exports a clean product feed to OpenAI on your behalf. Your job is to make sure every product has a real title, a useful description (not vendor-supplied boilerplate), schema markup (Shopify's themes handle this if you don't override), and crawler access. After that, you wait for ChatGPT to index, which takes 1–3 weeks in our window.

If you're on WooCommerce, BigCommerce, Magento, custom storefront, or a static-site generator, you submit your feed manually. The waitlist for OpenAI's merchant program is short and accepting submissions. The actual setup work is:

1. Build a Google Merchant Center–format product feed (xml or csv, OpenAI accepts both).
2. Host it at a stable URL the bot can fetch on a schedule.
3. Add JSON-LD Schema.org/Product markup to every product page.
4. Verify the Schema.org/Product markup passes Google's rich-results test (ChatGPT uses similar validation).
5. Submit the feed URL via the OpenAI merchant form and wait for confirmation.

Two engineering days for setup. Two weeks for ChatGPT to start surfacing the products. Total time from kickoff to first ChatGPT-attributed sale: roughly three weeks. That's the entire onboarding cost for a free acquisition channel.

Social proof signals matter more than your meta description

The single most underweighted ChatGPT ranking factor in the existing how-to guides is social proof. Schema.org markup is necessary but not sufficient. Two stores with identical schema, identical prices, identical product quality will rank differently in ChatGPT carousels because one of them has Reddit threads where actual humans recommend the product and the other doesn't.

This is downstream of the buyer-intent shift we've documented. Buyers ask their peers in public — Reddit, TikTok, niche Discord — before they ask vendors. ChatGPT reads those public conversations as part of its product-ranking signal. If you've been ignoring the community side of marketing because it doesn't show up in your dashboard, ChatGPT's carousel is the first place where that neglect becomes a hard ranking penalty.

The implication for a store launching today: investing in five strong Reddit conversations about your category, where customers organically describe their experience, is worth more for ChatGPT visibility than another 200 lines of meta-description optimization. The signal economy frame applies to merchant discoverability the same way it applies to B2B outbound. The surface changes; the discipline doesn't.

What "tracking ChatGPT traffic" actually looks like

You can't optimize what you can't measure. ChatGPT visits arrive in your analytics with referrer chatgpt.com or chat.openai.com, and ChatGPT appends utm_source=chatgpt.com to most product-link clicks. Set up two filters in your analytics tool — referrer match and UTM source — and you have a clean ChatGPT traffic segment in under fifteen minutes.

What to track in that segment: visit count, page conversion rate, average order value, return-customer rate. Almost every store we audit finds the ChatGPT segment has higher conversion than the site average, because the buyer arrived through a recommendation rather than through a paid ad or undirected search. The conversion delta is the data point that justifies investing more in ChatGPT-specific optimization.

This is also the only honest way to know whether the paid-acquisition channels you're running are competing with ChatGPT or complementing it. In our audits the channels are mostly complementary — the ChatGPT cohort is largely incremental traffic that wouldn't have come through Google or Meta. But the answer depends on the category, and you don't know which until you measure.

Where this still doesn't work

ChatGPT shopping isn't a fit for every store, and pretending otherwise sets you up for disappointment. Three failure modes show up consistently.

Category mismatch. If you sell B2B SaaS, expensive industrial equipment, regulated medical or pharmaceutical products, or anything where the purchase decision involves human-mediated sales, ChatGPT shopping isn't your channel. The carousel is built for consumer-discoverable products at price points a single person can decide on. Stay focused on the consumer goods categories.

Brand-new products with zero discussion footprint. ChatGPT doesn't surface products it has no public-discussion signal for. If you launched last week and nobody has talked about your product anywhere, getting listed will not solve your problem. You need a few months of organic conversation about the product before ChatGPT's ranking layer takes you seriously. The fix is the community-engagement discipline, not the technical listing.

Trying to game it with low-quality content. ChatGPT's safety layer aggressively suppresses products with obvious thin-content review pages, fake-looking aggregate ratings, or AI-generated descriptions that the model can detect. You can't out-clever the ranking. The honest path is real schema, real reviews, real descriptions. The path that backfires is templated product pages designed to look "schema-rich" without the substance behind them.

The window for organic ChatGPT shopping listing is open and free right now. Most operators won't move in time. The ones who set up the basics in Q4 2025 are watching the channel build for them through Q4 2026. The ones who wait until they see competitors in their category's carousels are doing the setup six months late.