ChatGPT Keywords: How to Find the Prompts Shoppers Use

ChatGPT keywords are the questions and phrases shoppers type into ChatGPT when they research a product, compare options or ask what to buy. To find them, start from the Google keywords you already rank for, rewrite each one as the full question a customer would ask an assistant, test those prompts in ChatGPT, and build pages that answer them clearly enough to be cited.

This guide is for Shopify and e-commerce teams that want their products to show up when buyers ask ChatGPT for a recommendation. You'll learn how prompts differ from search keywords, what the published research says about how people use ChatGPT, a seven-step process to build a prompt list for your store, the page changes that make you citable, how to use ChatGPT itself for keyword research, and how to measure the traffic that comes back.

What are ChatGPT keywords?

"ChatGPT keywords" is shorthand for two related things, and most guides only cover one of them:

• The prompts your customers use. The conversational questions people ask ChatGPT, such as "what's the best non-toxic cookware for a gas stove under $200?" These are the terms you want your brand to appear for.
• The search queries ChatGPT runs. When ChatGPT searches the web, it doesn't send the whole conversation to a search provider. OpenAI's help center explains that ChatGPT search can rewrite a prompt into a more specific query, and can send additional, more specific queries to other search providers after reviewing the first results. Those shorter rewritten queries decide which pages get read and cited.

A useful ChatGPT keyword strategy covers both layers. You map the long prompts your buyers ask to the short queries ChatGPT is likely to run, then make sure your product, collection and guide pages answer both.

There's also a third meaning: using ChatGPT as a keyword research tool. That's covered in its own section below, along with the limits you need to know before you trust its output.

How ChatGPT prompts differ from Google keywords

The biggest difference is length. Semrush analyzed more than 80 million lines of clickstream data from July to December 2024 and found that prompts submitted without ChatGPT's search feature averaged 23 words, while queries with search enabled averaged 4.2 words. The same study found that about 70% of the ChatGPT queries it analyzed didn't fit a classic search intent label, so Semrush created a new "unknown" intent category for them.

That gap matters for how you plan. A 23-word prompt carries context a Google keyword never does: budget, use case, who it's for, what the shopper already tried, and which brands they're weighing. Your pages need to answer those details, not just repeat a head term.

The good news: the two overlap more than they differ. The short queries ChatGPT runs when it searches look a lot like Google keywords, so the keyword research you already have is the right starting point. If you need a refresher on building that list, our guide to Shopify SEO keywords walks through it.

What people actually ask ChatGPT

The best public data on ChatGPT usage comes from OpenAI's own researchers. In the NBER working paper "How People Use ChatGPT" (Chatterji and coauthors, September 2025), the team classified a sample of about 1.1 million conversations. By July 2025, ChatGPT users were sending 18 billion messages a week, and 73% of messages were not related to work, up from 53% a year earlier.

The topic mix shifted fast. Seeking Information, the category the paper describes as a very close substitute for web search and which includes questions about products, grew from 14% to 24% of all usage between July 2024 and July 2025. Writing fell from 36% to 24%.

The paper also sorted messages by what the user wanted. About 49% were "Asking" (seeking information or advice to make a decision), 40% were "Doing" (asking ChatGPT to produce something, like an email or code) and 11% were "Expressing." Asking was growing faster than Doing.

For a store, the Asking share is the one to care about. Those are the moments a shopper is deciding: which product, which brand, whether a feature is worth paying for. They're also the prompts where your brand can be named.

Scale is no longer a reason to wait. TechCrunch reported in February 2026 that OpenAI announced 900 million weekly active users, up from the 800 million it reported in October 2025. And in March 2026, Shopify announced that products from eligible merchants become discoverable in ChatGPT by default through Agentic Storefronts, with orders arriving in the Shopify admin with ChatGPT referral attribution. Shoppers can now go from a ChatGPT prompt to a purchase, which makes prompt research a revenue question, not just a content one.

How to find ChatGPT keywords for your store

There is no ChatGPT equivalent of Google Keyword Planner. OpenAI doesn't publish prompt volumes, so you build your list from the signals you do have. This process takes a few hours for a focused catalog.

1. Start with your money keywords. Pull the queries that already drive impressions and revenue from Google Search Console. Our free Search Console analyzer can help you sort them. Focus on category and comparison terms such as "best", "vs", "for" and "under $X", because those map most directly to buying prompts.
2. Collect the questions customers already ask you. Support tickets, live chat logs, post-purchase surveys, product reviews and sales call notes contain the exact language buyers use. Look for constraints: body type, skin type, budget, room size, compatibility, allergies.
3. Mine community language. Read Reddit threads, niche forums and the "People also ask" boxes for your category. Note how people phrase problems before they know a product exists.
4. Rewrite each keyword as three prompts. For every money keyword, write one discovery prompt ("what helps with..."), one comparison prompt ("X vs Y for...") and one decision prompt ("is X worth it if..."). Include the context a real buyer would add.
5. Run the prompts in ChatGPT and record what you see. Use a logged-out or fresh session, with search on. Note which brands are named, which pages are cited, and what follow-up questions ChatGPT suggests. Repeat each prompt a few times, because answers vary between runs.
6. Extract the short queries. From the cited pages and the wording of the answer, infer the shorter queries ChatGPT likely ran. Check that you have a page that would rank for each one.
7. Prioritize by revenue, not novelty. Score each prompt by how close it is to a purchase, whether a competitor is being recommended instead of you, and whether you already have a page that could answer it with a few changes.

An example prompt map

Here's how one keyword expands into a prompt map for a hypothetical cookware store. The pattern works for any catalog.

How to turn prompts into pages ChatGPT cites

Finding the prompt is half the work. ChatGPT can only cite pages its search crawler can reach, and it tends to quote pages that answer a specific question plainly. Here's what to change.

Let OpenAI's search crawler in

OpenAI's crawler documentation says sites that opt out of OAI-SearchBot will not be shown in ChatGPT search answers. It's a separate bot from GPTBot, which collects training data, so you can block training and still appear in search. Check your robots.txt with our robots.txt tester, and read GPTBot vs OAI-SearchBot if you're unsure which rules you need. OpenAI notes that changes can take about 24 hours to take effect.

Answer the prompt in the first lines

Put a direct, quotable answer at the top of the relevant section: who the product is for, the key spec, the price range and the trade-off. Long prompts include constraints, so state yours explicitly, such as "works on gas, electric and induction" or "fits wide feet up to size 14."

Cover comparisons and objections

Prompts often name a competitor or ask "is it worth it." Pages that address alternatives honestly, including who should not buy your product, give ChatGPT the material it needs to recommend you in context.

Make product facts machine-readable

Keep price, availability, materials, dimensions and reviews in visible text and in Product structured data. On Shopify, also keep your catalog data clean, because Shopify Catalog is how products surface through Agentic Storefronts. Our schema validator checks your markup.

For a broader walkthrough of visibility in ChatGPT, see how to appear in ChatGPT.

How to use ChatGPT for keyword research

ChatGPT is a strong brainstorming partner for keyword research, but a weak source of numbers. Use it to expand and organize your list, then validate everything in a tool with real search data.

What ChatGPT does well

• Expanding a seed list. Give it a product and audience and ask for the questions buyers ask at each stage, from problem to purchase.
• Grouping keywords. Paste a keyword export and ask it to cluster terms by intent and suggest which page should own each cluster. This helps prevent two pages competing for the same term.
• Rewriting keywords as prompts. Ask it to turn each keyword into the full question a specific customer persona would ask.
• Spotting missing angles. Paste your page and a competitor's page and ask what questions the competitor answers that you don't.

What it can't do

• It doesn't have reliable search volume, keyword difficulty or ranking data, and it may produce plausible-looking numbers anyway. Never copy a volume figure from ChatGPT into a plan.
• It can't see your Search Console or analytics unless you paste or upload the data.
• Its suggestions skew toward generic phrasing. Real customer language from reviews and support tickets is usually more specific and converts better.

Prompts to copy

For more on the content side, see using ChatGPT for content and SEO.

How to measure ChatGPT keyword performance

You can't see ChatGPT impressions the way you see Google impressions, so measure in three layers.

1. Referral traffic. OpenAI's publisher FAQ says ChatGPT automatically adds utm_source=chatgpt.com to referral links from ChatGPT search. In GA4, filter sessions by that source and by the chatgpt.com referrer, then look at landing pages, conversion rate and revenue. Shopify also reports orders from Agentic Storefronts with ChatGPT attribution in the admin.
2. Prompt tracking. Re-run your priority prompts on a schedule, every two to four weeks, and log whether your brand is named, whether you're cited, and which competitors appear. Because answers vary, track the share of runs that mention you rather than a single result.
3. Supporting Google signals. Watch the short queries you mapped in Search Console. Rising impressions on those terms are a leading indicator that the pages ChatGPT may read are getting stronger.

Our free AEO checker gives you a quick read on whether a page is structured to be quoted by AI answers.

Where to start this week

Pick your ten best-selling products or categories. For each one, write three prompts using the discovery, comparison and decision pattern, run them in ChatGPT, and note who gets recommended. Fix crawler access first, then update the pages closest to a purchase with direct answers, honest comparisons and complete product facts. Set up the GA4 filter so you can see whether it's working.

If you want a second set of eyes, our ChatGPT SEO team builds prompt maps and citable pages for Shopify brands. You can also check where your site stands with a ChatGPT SEO audit, or get in touch to talk through your catalog.