How to Pitch an Ecommerce Brand With the Products ChatGPT Recommends
Ask ChatGPT, Gemini and Perplexity a shopper's question, record which products they recommend, and pitch the ecommerce brand on what the answer left out.
Show an ecommerce brand the products an AI assistant recommends to their shopper, with the reasons it gave and their own product absent. The method:
- Pick the shopper's question: Choose one product the store sells and write what a shopper would ask: product type, use, size and budget.
- Ask it as someone else: Phrase the prompt as a purchase for another person, in an unpersonalized chat, so your own history does not shape the answer.
- Record brands, products and reasons: Run it in ChatGPT, Gemini and Perplexity. Capture each answer with the date and copy the reason given for every product.
- Find the missing information: Check that the product page copy, the Product structured data, the feed and the reviews in the HTML state what the shopper asked for.
- Send one finding and ask: One capture, one likely cause, one fix, one question. The brand knows its shoppers better than you do.
CrawlRaven does not run the prompts. Its crawl explains the cause you report in step 4.
This is a sales method, and the first three steps need nothing but an AI assistant. CrawlRaven comes in at step 4, where you have to explain why the product was left out, and again once the brand is a client and you can join the crawl with their Search Console and GA4 data. Try CrawlRaven free for 7 days: 1 website, no credit card →
A shopper asks an AI assistant for a waterproof school backpack for an 11-year-old, under $60, with free returns. The assistant answers with three products and a reason beside each one. The store you want to pitch sells that backpack, and its name is absent from the answer.
The short version: to pitch an ecommerce brand, show them the products an AI assistant recommends to their shopper, name the one piece of information it could not find about their product, and ask which shoppers matter most to them. I have run an SEO agency for more than 15 years, and I call this the AI shopping test.
What you send the ecommerce brand
I am looking for a waterproof school backpack for my cousin's 11-year-old son. The budget is under $60, it needs free returns in case the size is wrong, and it should arrive within a week. Which ones would you recommend, and why?
Here are three backpacks that fit what you described:
Sized for ages 8 to 12 and listed under your budget.
Fits the budgetFree returns within 30 days, stated on the product page.
Free returnsReviewers say it keeps books dry in heavy rain.
Reviews mention rainYour prospect's backpack is missing from the answer
It costs $54 and has free returns as well. The age range is only inside a size-chart image, and the returns page has no link from the product page.
Illustration with placeholder names and prices. The green tags are the reasons you copy down in step 3.
The mock above uses placeholder names and prices. Brand A is recommended at $49 because it is sized for ages 8 to 12 and fits the budget. Brand B, at $58, is there for free returns within 30 days. Brand C, at $55, is there because reviewers mention rain. The prospect's backpack costs $54 with free returns and is missing.
Why a product recommendation gets a reply from a brand
Ecommerce teams receive cold SEO audits all the time, and most read the same for every store. A product recommendation is about their catalog, in plain sentences, with a competitor's product where they expected their own.
- It needs no explanation. The founder or marketing lead reads it the way a shopper would.
- It is specific. It names a product they sell, a price, and the brands they compete with.
- It comes with reasons. Each reason tells you what the assistant looked for, and that points to a fix.
It is also an uncrowded angle. On 11 October 2026 I checked Google in the United States for “how to get products recommended by chatgpt”. The results were a Reddit thread, OpenAI's own merchant page, and guides from Profound, HubSpot, Bazaarvoice and Impact Plus. All of them are written for brands. I found none written for an agency using the test as a pitch.
Google Ads data for the United States on the same date shows the interest:
- “ai shopping”: 1,300 searches a month
- “chatgpt shopping”: 1,000 searches a month
- “perplexity shopping”: 480 searches a month
The AI shopping test in five steps
From the shopper's question to the first reply
Choose one product the store sells and write what a shopper would ask: the product type, the use, the size and the budget.
Phrase the prompt as a purchase for another person, in an unpersonalized chat, so your own history stays out of the answer.
Run it in ChatGPT, Gemini and Perplexity. Capture each answer with the date, and copy the reason given for every product.
Match each thing the shopper asked for to the product page copy, the Product structured data, the feed and the reviews in the HTML.
One capture, one likely cause, one fix, one question about which products and shoppers matter most to the brand.
Step 5 loops back to step 1: the brand's answer becomes the next prompt.
What the vendors say about how products get picked
You need a few facts before you explain a gap to a brand. These come from the vendors' own pages, listed in the sources at the end.
- ChatGPT results are selected on relevance. OpenAI's help center says product results “are selected independently by ChatGPT and are not ads”, and that a product appears when ChatGPT perceives it to be relevant to the shopper's intent.
- Product data is an input. The same page says ChatGPT considers structured metadata from first-party and third-party providers, such as price and product description, and other third-party content.
- The budget changes the weighting. OpenAI's example: if you specify a budget of $30, ChatGPT will focus more on price.
- Review summaries come from public pages. OpenAI says its review summaries are based on reviews from public websites.
- Shopify catalogs are already connected. OpenAI says product data for Shopify merchants is integrated through Shopify Catalog. Its merchant page says a feed is optional when ChatGPT already crawls the site, and that a feed gives more control over how products appear.
- Perplexity says its product cards are unsponsored. Its 2024 launch post says a merchant who shares product data has a better chance of being recommended, because more detail helps Perplexity judge relevance.
The common thread is product information. Every vendor describes matching a shopper's request against details about the product. Step 4 checks whether the store states those details.
Step 1: Pick the question a shopper would ask
Start with one product. If you do not have a prospect yet, build a list of stores first and qualify each store before you spend time on the test.
- Open the store and find a product it pushes: on the homepage, in the main menu or at the top of a collection.
- Write one sentence a shopper who needs that product would say out loud.
- Add the details shoppers give: what it is for, who it is for, the size, the budget, and anything about shipping or returns.
- Leave out the brand name. You are testing whether the assistant finds the product without being told.
- Label it a first guess. The brand will give you a better question after the first reply.
Step 2: Ask on behalf of someone else
AI assistant answers are skewed by your own account history. OpenAI's shopping help page says ChatGPT considers your query and context such as Memory or Custom instructions. Your account is full of SEO work, so your answer will differ from a shopper's.
My way around this is to phrase the test prompt on someone else's behalf. The prompt describes the person the product is for, and that pulls the answer toward a real shopper's request.
I am looking for a waterproof school backpack for my cousin's 11-year-old son. The budget is under $60, it needs free returns in case the size is wrong, and it should arrive within a week. Which ones would you recommend, and why?
- Open a temporary chat in ChatGPT and choose Unpersonalized before the first message. OpenAI says this mode does not use memory or custom instructions.
- Write the prompt as a purchase for another person, with their age and what they need the product for.
- State the budget and any shipping or returns condition.
- Ask for a recommendation and the reason, so the answer tells you what it valued.
- Do not mention SEO, marketing or the prospect's brand.
Step 3: Record brands, products and reasons in three engines
Run the same prompt in ChatGPT, Gemini and Perplexity. Their answers vary, and a product missing from all three is your strongest finding.
- Screenshot each full answer, with the prompt visible.
- Note the date and the engine on each screenshot.
- List the brands and products recommended, in the order they appear, with the price shown.
- Copy the reason given for each one, such as fits the budget, free returns or reviews that mention rain.
- Note anything the assistant says it could not confirm, such as a delivery time.
- Run the prompt a second time to see whether the list holds.
The reasons are the useful part. They show which details the assistant treated as proof, and those are the details you look for on the prospect's store in the next step.
Step 4: Find the information the assistant could not find
A common reason a business is missing from AI recommendations is incomplete information the AI cannot reach, such as a page that exists but is not linked. The store sells what the shopper asked for, and the assistant cannot tell.
Take each thing the shopper asked for and find where the store states it. There are four places to look:
- Product page copy. The use, the size and the age range written as text. My guide to product page SEO covers what belongs there.
- Product structured data. Google's merchant listing documentation requires a name, an image and an offer with a price. It recommends availability, shippingDetails, hasMerchantReturnPolicy, aggregateRating and review.
- The product feed. Google's Merchant Center specification requires title, description, price and availability, and has attributes for size and age group.
- Reviews in the HTML. Review text that is present in the page source, where a crawler can read it.
A block of code on a product page that states the product name, price, availability, shipping, returns and rating in a fixed format that machines can read without interpreting the page design.
What the shopper asked, and where the store must say it
| The shopper asked | The store must say it here | What to check |
|---|---|---|
| Use: waterproof, for school | Product page copy that names the use in text | Is it in the description, or only in a photo or a badge? |
| Size: an 11-year-old | Age range and dimensions in text, plus the size and age group feed attributes | Is the size chart text or an image? Are the feed fields filled? |
| Price: under $60 | Offer price in Product structured data, matching the feed and the page | Does the schema validate, and do all three show the same price? |
| Shipping: arrives within a week | Delivery times on the product page and shippingDetails in the schema | Is delivery time written per product, or only at checkout? |
| Returns: free returns | A return policy page and hasMerchantReturnPolicy in the schema | Does the product page link to the policy, or is the page orphaned? |
| Reviews: a reason to trust it | Review text and the rating in the page HTML | Do reviews appear with scripts turned off, or only inside a widget? |
Illustration with placeholder names. The prompt is the school backpack example from step 2.
In text, the six checks from the table:
- Use (waterproof, for school): is it in the description text, or only in a photo or a badge?
- Size (an 11-year-old): is the size chart text or an image, and are the size and age group feed fields filled?
- Price (under $60): does the structured data validate, and do the page, the schema and the feed show the same price?
- Shipping (within a week): is the delivery time written on the product page, or only at checkout?
- Returns (free returns): does the product page link to the policy, or is the policy an orphan page?
- Reviews: do they appear with scripts turned off, or only inside a widget?
Two free tools cover the schema checks. Paste the product URL into the schema validator to see what is missing or invalid. Use the Product schema generator to build the corrected markup you will show as the fix.
Rendering deserves its own check. Google recommends putting Product structured data in the initial HTML, and says markup generated with JavaScript can make Shopping crawls less frequent and less reliable. Also read the store's robots.txt. OpenAI says sites opted out of OAI-SearchBot will not be shown in ChatGPT search answers.
On a store with thousands of products, crawl it. CrawlRaven reports missing or invalid Product schema, product pages with no internal links, blocked AI crawlers and content that only renders with scripts, so you leave with a specific cause.
Step 5: Send one finding and ask one question
Keep the message short enough to read on a phone. To find the right person and their address, see my Apollo outreach walkthrough. The message has four parts:
- The question you asked. Quote the prompt so they see it is a shopper's question.
- What was recommended, and why. Attach the capture and repeat the reasons in one line.
- One likely cause. For example: the age range is only inside a size-chart image, so the assistant could not match the backpack to an 11-year-old.
- One question. Ask which products and shoppers matter most to them, and whether this is a question those shoppers would ask.
Word the cause as likely. You can see what the store fails to state, and the assistant's own reasoning stays hidden from you. If the brand wants more than one finding, follow up with a three-finding store teardown.
You may be the SEO expert, but they know the business. I would treat the first prompt as a guess and let the brand correct it. Their answer about which products and shoppers matter gives you the next prompt, and the second test is the one they will care about.
What loses the client
This pitch fails in two predictable ways.
- Being too sure of your recommendation. If you arrive with the answer and do not listen to their specific requirement, you lose them. Do not try to teach business to them.
- Bashing their site too much. Some owners are offended when you go on about everything that is wrong with the store. Stick to one finding and one fix.
The capture already carries the bad news. Your job in the message is to be the person who noticed and knows a fix.
When the test does not go to plan
What if the prospect's product is already recommended?
Tell them. It is good news and it earns trust. Then test a second product or a narrower shopper. If the brand appears every time, open with a different finding, such as one from the ecommerce SEO audit.
What if the answer changes every time I run it?
Expect that, which is why step 3 asks for a second run. Report what you saw: “recommended in one of four runs” is still a finding. Present each capture as one dated answer. OpenAI itself notes that only some of the available products are shown.
What if I cannot find a cause on the store?
Say that too. Send the capture and the question without a cause, and offer to look further once you can see their data. A cause you made up to fill the gap is how you end up being too sure.
Tools that make this faster
Steps 1 to 3 need only the assistants themselves. Step 4 is where tools save time.
- CrawlRaven. Our tool. Before the brand signs, the crawl finds missing or invalid Product schema, product pages with no internal links, blocked AI crawlers and content that only renders with scripts. After they grant access, it joins the crawl with Search Console and GA4 in a workspace with content priorities and separate audits. See the AI search readiness audit.
- Free checks. The schema validator, the Product schema generator and the robots.txt tester, which shows which AI crawlers a store blocks.
- Prompt trackers. CrawlRaven does not run prompts or track AI mentions today, so the prompts here are run by hand. To track them across many prospects, see my roundup of LLM SEO tools.
We are also launching an AI SEO tool soon. Join the waitlist to hear when it opens.
Sources used in this post
- OpenAI Help Center – Shopping with ChatGPT Search: how product results and merchants are selected, read 11 October 2026.
- OpenAI – Power product discovery in ChatGPT: product feeds, Shopify and Etsy catalogs, and the crawl question.
- OpenAI Help Center – Memory in ChatGPT and OpenAI Help Center – Temporary chat in ChatGPT: personalization and the Unpersonalized option.
- OpenAI – Overview of OpenAI crawlers: what opting out of OAI-SearchBot does.
- Perplexity – Shop like a Pro: product cards and the merchant program, published 18 November 2024.
- Google Search Central – Merchant listing structured data: required and recommended properties, and the initial HTML recommendation.
- Google Merchant Center Help – Product data specification: required feed attributes, size and age group.
- Google results for “how to get products recommended by chatgpt” and Google Ads search volumes, United States, checked 11 October 2026.
Related reading on CrawlRaven
- How to get local SEO clients with one ChatGPT screenshot, the same test for local businesses.
- Shopify and Google Merchant Center, for feed and schema fixes once the brand signs.
- Shopify robots.txt, for which AI crawlers a store allows by default.
Frequently asked questions
How do I use ChatGPT to pitch an ecommerce SEO client?
Ask ChatGPT the question one of the store's shoppers would ask, with a product type, a use, a size and a budget. Record which brands and products it recommends and the reason it gives for each. If the prospect's product is absent, find what the store fails to state about that product and send that one finding with one fix and one question.
How does ChatGPT choose which products to recommend?
OpenAI's help center says a product appears when ChatGPT perceives it to be relevant to the shopper's intent, and that it considers structured metadata from first-party and third-party providers, such as price and product description, along with other third-party content. It also says product results are selected independently and are not ads.
Why is an ecommerce brand missing from AI product recommendations?
In my experience a common reason is incomplete information the assistant cannot reach. The store sells the right product, but the size is inside an image, the return policy sits on a page nothing links to, or the price and shipping details are absent from the Product structured data and the feed.
Why do I get different product recommendations than a real shopper would?
ChatGPT can personalize answers. OpenAI's shopping help page says it considers your query and context such as Memory or Custom instructions. Your account history is full of SEO work, so run the test in a temporary chat set to Unpersonalized and phrase the prompt as a purchase for another person.
Does a Shopify store need a product feed to appear in ChatGPT shopping?
OpenAI says product data for Shopify merchants is already integrated into ChatGPT through Shopify Catalog, with no additional work required from individual merchants. Its merchant page also says a feed is optional when ChatGPT already crawls the site. The product data still has to contain the details a shopper asks about.
Does CrawlRaven track which products ChatGPT recommends?
No. CrawlRaven does not run prompts or track AI mentions today, so you run this test by hand. Its crawl explains causes: missing or invalid Product schema, product pages with no internal links, blocked AI crawlers and content that only renders with scripts. After the client grants access, it joins the crawl with Search Console and GA4 data.
What should the first message to an ecommerce brand say?
Four things, briefly: the shopper question you asked, which products the assistant recommended and why, the one piece of information you think it could not find for their product, and a question about which products and shoppers matter most to them. Leave the full audit for later.
15+ years of growing SaaS websites through SEO | Author, 200-Point Audit Checklist
Aditi has spent 15+ years helping SaaS companies scale organic traffic through technical SEO and content strategy. She is the author of the CrawlRaven 200-Point Audit checklist used by agencies and in-house teams to systematically improve search performance.