How to Get Your Products Recommended by ChatGPT in 2026

Published: August 24, 2026

Last updated: August 24, 2026

10 min read

Somewhere today, a buyer asked ChatGPT what to get their dad for his golf trip.

It answered with three products. They bought one.

No search results page. No ads. No visit to the seventeen stores that would have ranked for the query.

If your products were in that answer, you just made a sale you'll struggle to attribute. If they weren't, you never knew the buyer existed.

That's the whole game now, and most Shopify stores are playing it with default settings and a shrug.

This guide is the practical version of the AI search section in my complete Shopify SEO guide: where AI product recommendations actually come from, the Shopify settings to check this week, the data that decides who gets picked, and the content play that took my client REHAUS to number one in ChatGPT for their category.

Platform facts in this guide were verified in August 2026. This corner of search changes monthly, so if you're reading it much later, check the current state before you act on anything dated.

Short on time? Here are the key takeaways

Two surfaces, two playbooks: the shopping carousel is won with data quality, the conversational answer with citations and consistent language.
Shopify already syndicates eligible stores: Agentic Storefronts are on by default, so check your channel settings.
Google-Extended is not a crawler: blocking it does nothing to AI Overviews, which run on Googlebot.
Recommendations are earned, not bought: clean data, one canonical description, and third-party proof.

Where AI recommendations actually come from

First, a distinction almost every guide muddles.

AI assistants recommend products on two different surfaces, and they're won in two different ways.

The shopping surface: the product carousel with images, prices and buy links. This is fed by structured product data: catalogues, feeds and marketplace integrations. It's won with data quality.
The conversational answer: "what's the best second hand designer furniture?" answered in prose, with brands named and linked. This is fed by what the model reads across the web. It's won with citations, canonical language and third-party corroboration.

Different plumbing, different playbooks. You want both.

And it's bigger than ChatGPT.

As I write this, Google's AI Mode has passed a billion monthly users and leans on a Shopping Graph of 60 billion product listings, Microsoft Copilot has run checkout with a Shopify integration since January, and Perplexity's Instant Buy handles purchases for thousands of merchants through PayPal.

The good news? The work below feeds all of them at once, and your Google Merchant Center feed does a chunk of the lifting on the shopping surface. That's covered in my Merchant Center guide.

The Instant Checkout lesson

A quick story, because it explains why this article is about data before it's about tactics.

In September 2025, OpenAI launched Instant Checkout: buy inside ChatGPT, powered by the Agentic Commerce Protocol it built with Stripe.

The agentic future, arrived.

By March 2026 OpenAI had pulled it back, moving purchases into merchant-run ChatGPT apps instead.

The reasons were plural and unglamorous. Barely a few dozen merchants live. Conversion inside the chat running well below click-out. A checkout that couldn't do multi-item carts, promo codes or shipping promises.

And underneath it, product data that couldn't reliably back a sale. One analyst noted inventory was simply absent from the plan.

Read that again.

The agents showed up ready to sell. The merchant data wasn't ready to be sold from.

That's the entire thesis of this cluster in one product decision.

The stores that win the AI era are the ones whose machine-readable layer tells the truth: live prices, honest availability, one source of schema.

Buyers now discover in AI and buy on site, and the discovering is done entirely from your data exhaust.

Check your Shopify agent settings this week

Here's the step nearly every store owner has missed, because Shopify shipped it quietly.

Since 24 March 2026, eligible Shopify stores have Agentic Storefronts on by default.

Your products already syndicate through Shopify's catalogue toward ChatGPT, Google's AI Mode and Gemini, Microsoft Copilot and Meta, and the platform handles the protocol plumbing (Google and Shopify's UCP, OpenAI and Stripe's ACP) so you don't have to.

Note what isn't on that list. Perplexity runs its own merchant programme and needs a separate opt-in, so assuming Shopify has you covered there is a mistake.

Your job is twenty minutes of checking:

Review the channels: in Shopify Admin, under Sales channels, confirm which AI surfaces you're syndicated to and that nothing you sell is accidentally excluded. Eligibility varies, and Google's surfaces were still rolling out when I wrote this.
Fix what the agents will read: titles that name the product, descriptions that state facts, GTINs populated, at least three decent images. Agents rank on attribute completeness, not adjectives.
Publish the policy pages: delivery, returns, guarantees. Recommendation engines check them for credibility before featuring you, which is the same reason they sit in my llms.txt template.
Verify the trust triad: price, availability and ratings, agreeing everywhere they appear. That's schema work, and my Shopify schema markup guide is the single-source setup.

OpenAI also runs a merchant feed programme for approved partners who want direct control, with a daily catalogue snapshot plus API updates through the day.

Worth exploring once the basics above are honest. Pointless before.

Win the conversational answer

Now the surface where content strategists earn their keep, because the shopping carousel is table stakes and the conversational answer is where brands get made.

When someone asks an assistant a category question, the model assembles an answer from what it can read and retrieve: your pages, other people's pages about you, reviews, comparisons, forums.

You influence that answer three ways.

Say what you are, identically, everywhere: one canonical sentence on your homepage, about page, Organization schema and llms.txt. Machines quote consistency. The full method is in my llms.txt on Shopify guide.
Write the shortlist content yourself: category roundups, honest brand comparisons, alternatives pieces. Models answering "best X" lean on exactly this format, which is why it's the format I used across 28 REHAUS articles and my own store's brand-review content. My guide to comparison page SEO has the template.
Earn third-party corroboration: most of what an AI says about your brand comes from pages you don't own. Reviews platforms, press, community threads. You can't astroturf this at any scale worth having, so earn it: a distinctive product, review volume, and a founder visible enough to be mentioned.

Does this actually move an answer?

Ask ChatGPT for the best second hand designer furniture.

REHAUS comes back number one, "Best Overall", above Vinterior, Pamono, 1stDibs and eBay UK.

The answer quotes the site's canonical sentence almost word for word: a specialist reseller that collects, authenticates, prepares and stores.

That sentence appears on the homepage, the about page, the schema and the llms.txt, and every one of the 28 articles reinforces it.

The receipts are in the REHAUS case study.

Nothing about that was luck. It was repetition, structure and third-party proof, pointed at one question.

Crawlers: let retrieval in, decide on training

One nuance worth getting right, because the crawler list isn't one decision anymore.

Retrieval crawlers fetch your pages to answer a question someone is asking right now. OAI-SearchBot and ChatGPT-User for ChatGPT, PerplexityBot and Perplexity-User, Claude-SearchBot and Claude-User.

Block these and you've opted out of the answer. Let them in, no debate.

Training crawlers collect data to teach future models. GPTBot and ClaudeBot are the main ones.

Blocking those is a legitimate choice some large publishers make.

My take, for a store rather than a newspaper: allow both.

A small brand's problem is obscurity, not appropriation.

Being in the training data is how a model knows you exist when nobody's retrieving.

One correction worth making here, because it costs publishers real visibility.

Google-Extended is not a crawler. It's a control token with no user-agent of its own, and it governs Gemini app training and grounding.

It does not control whether you appear in AI Overviews or AI Mode. Those run on Googlebot, because they live inside Search.

So blocking Google-Extended to stay out of AI Overviews achieves nothing except cutting yourself out of Gemini. Plenty of sites have made exactly that trade by accident.

Both lists live in robots.txt, editable on Shopify via robots.txt.liquid, and check 37 of my 40-point Shopify SEO checklist covers the verification.

Measure it properly

Time for proportion, because this channel gets oversold weekly.

AI-referred traffic is still a small slice of retail visits, typically under a few percent even for stores doing this well.

Two things stop that being a reason to ignore it.

First, the buyers convert.

Adobe's data had AI-referred shoppers converting 31% better than other traffic over the 2025 holidays, and 42% better by March 2026, with AI traffic up 393% year on year.

They arrive pre-sold, because the assistant already did the selling.

Second, the attribution is broken in your favour's disguise.

Buyers discover in ChatGPT, then search your brand on Google or type the URL. Your analytics credits "organic" or "direct" for a sale the assistant created.

The channel is bigger than your GA4 says.

So measure it with a ritual, not a dashboard.

Monthly: ask ChatGPT, Perplexity and Google's AI Mode the three questions your store should win.

Screenshot the answers. Track whether you're in them, and whether the brands beating you deserve it.

My AEO readiness score tells you in two minutes which foundations you're missing before the ritual starts.

The order of operations

A disqualifier before the FAQ, because this article will attract shortcut hunters.

If your collections run default metadata and your schema disagrees with itself, AI recommendations are not your next job.

The assistants read the same store Google does.

Fix the foundations in the checklist, then come back here.

The AI layer rewards stores that were already telling the truth cleanly. It doesn't rescue the rest.

Final Thoughts

Every era of search had a moment where early effort was embarrassingly cheap.

This is that moment for AI recommendations, and it's the least mystical work imaginable.

Check the Shopify agent settings everyone ignores. Make the data honest. Say what you are in one sentence, everywhere. Write the shortlists yourself. Earn the mentions you can't write.

Then ask the assistants your three questions every month and watch yourself climb into the answer.

REHAUS went from invisible to "Best Overall" in one season of exactly this.

If you want to know where your store stands first, my AI visibility audit maps it properly.

Get in touch, or email me directly.

Author Bio

Graeme Whiles is an independent SEO and AEO consultant at GWContent. He has worked with enterprise and SaaS brands, including Originality.ai, Connecteam, 6sense, Practice Better, and Peppr, growing organic traffic and AI search visibility across some of the most competitive categories in B2B. He also built Three Putt Golf Clothing from a blank domain as a live proof of concept for his methodology.

ChatGPT Product Recommendation Frequently Asked Questions

How does ChatGPT decide which products to recommend?

Two systems. Its shopping surface ranks structured product data: attribute completeness, pricing transparency, reviews, availability accuracy. Its conversational answers assemble from what it reads and retrieves: your declarative content, your consistent brand description, and what third parties say about you. Different inputs, so work both.

Can I pay to be recommended by ChatGPT?

Not into the organic recommendation itself as I write this, which is precisely why this work matters: the answer is earned or absent. Advertising formats around AI assistants are arriving, so watch that space. Treat any agency selling "guaranteed AI placement" the way you'd treat guaranteed rankings.

Do I need to submit a product feed if I'm on Shopify?

Mostly no. Agentic Storefronts syndicate eligible stores' catalogues automatically, so your job is verifying the channel settings and the quality of what's syndicated. OpenAI's merchant feed programme is open to approved partners wanting tighter control, and Perplexity requires its own opt-in.

What happened to ChatGPT Instant Checkout?

OpenAI pulled it back in March 2026 and moved purchasing into merchant-run ChatGPT apps, citing a lack of flexibility in the first version. Adoption was tiny, conversion lagged click-out, and the feature couldn't handle multi-item carts or promo codes. Discovery inside ChatGPT continues, and OpenAI's work with Stripe on the Agentic Commerce Protocol continues with it.