How to get your Shopify store recommended by ChatGPT

The shape of the problem is different from SEO

Search gave you a second chance. If you ranked eleventh, you were on page two — invisible in practice, but present, and one good quarter of work from being visible.

An answer engine does not work like that. Someone asks for the best merino base layer for cold-weather running, and the assistant names three or four stores in a paragraph. There is no fifth position. There is no page two to climb onto later. Either your catalogue is in the answer, or a competitor's is, and the buyer never learns you existed.

That is why this is worth attention now rather than when it feels mature. The mechanics reward whoever is machine-readable at the moment the question is asked, and retrofitting that after your category has been answered a few million times is a harder job than getting it right first.

What actually decides the answer

Four things, roughly in order of how much they matter and how much control you have.

1. Whether the crawlers are allowed in

The first thing to check is the least interesting and the most commonly broken: your robots.txt.

Assistants reach the open web through named crawlers — GPTBot and OAI-SearchBot for OpenAI, ClaudeBot and Claude-SearchBot for Anthropic, PerplexityBot, Google-Extended for Gemini's training use, and a handful of others. Some of these tokens are opt-out: if you say nothing, you are in. Others are opt-in for specific uses.

Two failure modes are common. A store blocks them all with an over-broad rule copied from a privacy-hardening blog post, then wonders why it never appears. Or a store blocks nothing and assumes that is the same as being read well, which it is not — being crawlable is the entry requirement, not the achievement.

Check your live file at yourstore.com/robots.txt and read it as a list of who you have quietly excluded. Which token does what, why naming a bot in order to allow it can switch off every other rule you have, and how to change the file on Shopify without freezing it, is a post of its own: what your Shopify robots.txt tells the AI crawlers.

2. Whether your product data is structured

An assistant that can read your page still has to work out what is being sold, who it is for, what it costs, whether it is in stock, and whether anyone liked it. Structured data answers those questions in a format that requires no interpretation.

For a Shopify store that means valid JSON-LD on every product page: a Product node with the name and description, an Offer with the price, currency and availability, and review data as Review and AggregateRating where you genuinely have it.

Most themes ship some of this. Very few ship all of it correctly, and apps routinely break it — two apps each injecting their own Product node on the same page is a common and quietly damaging pattern, because a page asserting two different prices for one item is a page an engine has reason to distrust.

Run your product URL through Google's Rich Results Test. It reports what a machine actually extracted rather than what your theme intended.

3. Whether your copy answers the question

This is the part most stores get wrong, and it is not a technical problem.

Product copy written for brand voice tends to open with atmosphere: a sentence about the mountains, a word like considered, a paragraph that establishes mood before it establishes fact. That copy can be excellent for a human who has already arrived and is deciding whether they like you.

It is close to useless to a machine trying to decide whether to recommend you, because it never states the extractable facts. What is it made of. What temperature range is it for. Who is it not for. What does it cost. What is the return window.

Answer-shaped copy states those things in plain sentences, near the top, in text rather than in an image or a tab that loads on click. You can keep the atmosphere — put it second. The test is whether someone could answer a buyer's question by quoting one sentence from your page.

4. Whether anyone else corroborates you

Assistants weight sources they did not get from you. A store that describes itself as the best in its category is making a claim; a review site, a forum thread, a comparison article and a retailer listing that all describe it similarly are evidence.

This is the slowest of the four to move and the least susceptible to a technical fix, which is exactly why it is worth starting before you need it. It is also the reason a store with weak product pages and strong third-party presence sometimes gets recommended over a better-built competitor nobody has written about.

Test it yourself in ten minutes

You do not need a tool or a subscription to find out where you stand.

1. Write down the three questions a real buyer would type before they knew your brand existed. Not your brand name — the problem. "Best waterproof dog coat for a whippet." "Alcohol-free aperitif that is not sweet."
2. Ask each one in ChatGPT, Gemini and Perplexity. Nine prompts.
3. Record what came back verbatim: which stores were named, in what order, and which sources the answer leaned on.
4. Now check the mechanics. Open yourstore.com/robots.txt. Run one product URL through the Rich Results Test. Read your best-selling product page and ask whether the first hundred words state what it is, who it is for, and what it costs.

The result is usually uncomfortable and always specific. Most stores discover they are absent from all nine answers, that their schema has a warning they have never seen, and that a competitor they do not rate is named in every one.

Run it again in a month. Engine answers move week to week, and a single reading is a data point rather than a trend.

What to fix first

In this order, because that is the order of leverage:

• Crawler access, if anything is blocked. It is a five-minute fix and everything else is worthless without it.
• Schema validity on product pages, including removing whichever app is duplicating a node. This is a build task, usually a small one.
• The first hundred words of your top ten product pages. This is a copy task and it is the one that actually changes whether you get named.
• Third-party presence, which is a slow campaign rather than a fix, and belongs in a quarterly plan rather than a sprint.

The honest limits

Nobody can guarantee inclusion in an answer. There is no submission form, no ranking factor list, and no support ticket. Anyone selling certainty here is selling something they do not have.

What can be done is remove every reason for an engine to skip you, and make the facts about your products trivially extractable. That is a real body of work with a defined end state, and it is measurable in the only way that matters: ask the questions again in a month and see whether your name has appeared.

If you would rather have that written down for your store, with the prompts run and reported verbatim and the fixes ranked by what they are worth, that is section five of the Store Teardown — and it is the section nobody else's audit includes.