When a shopper asks ChatGPT for a product, the brands it names come from two places: what the model learned in training and what it finds when it searches live. Here's how ChatGPT, Claude, Gemini and Perplexity each do it, from their own documentation, what we saw when we asked 3 assistants the same shoe question, and how to get your most profitable products recommended.
On this page
- How does answer engine optimization work?
- Which search index does each assistant use?
- What is query fan-out?
- How are citations and product results chosen?
- We asked 3 assistants the same shopping question
- Why agreement across sources moves brand mentions in ChatGPT, Gemini and Perplexity
- How to get recommended by ChatGPT: 6 steps
- Get recommended for the products that make you money
How do you get recommended by ChatGPT? First you need to know how it picks brands, the part most advice skips. When a shopper asks an AI assistant for a product, the answer comes from two places: what the model learned in training, and what it finds when it searches the web live. Which search index it uses, how it splits the shopper's question into smaller searches and which pages it trusts all decide whether your products make the list.
That's the mechanism behind answer engine optimisation, and it works a little differently in ChatGPT, Claude, Gemini and Perplexity. Below is how each works, from their own documentation, plus a test we ran on 20 October 2026. If the term is new to you, start with our guide to AEO for ecommerce.
How does answer engine optimization work?
Answer engine optimization works by making your brand easy for an AI assistant to find, trust and repeat. The assistant builds each answer from its training data and from live web results, then names the products that credible sources agree on.
That gives you two levers. The slow one is what's written about your brand across the web, which ends up in future training data. The fast one is the set of pages an assistant retrieves when it searches today, which shapes today's answer. Generative engine optimisation (GEO) is the same idea, and our guide to AEO vs SEO shows how both overlap with classic search.
What the model already knows
Every model has a training cutoff and can answer from memory for categories it saw discussed often. Anthropic's web search documentation says Claude searches when a request depends on information that is current, changing or outside its training data (prices and products that might have changed are on the list) and answers directly when the request draws on stable knowledge. Shopping questions usually land in the first group, because prices and stock move.
When it searches the live web
OpenAI's help page on searching the web with ChatGPT says ChatGPT may search automatically when a question would benefit from current information. Its shopping help page adds a detail that matters for brands: when choosing products, ChatGPT considers the answer it generated before it looked at any new search results.
Which search index does each assistant use?
An assistant can only cite pages it can reach, and they don't all search the same index. A crawler blocked in robots.txt or a firewall rule means your pages can't be cited, so this is the cheapest check in AEO.
- ChatGPT uses partner search providers, and its help page points users to Microsoft's and Shopify's privacy policies. Sites that opt out of OAI-SearchBot won't be shown in ChatGPT search answers, according to OpenAI's crawler docs.
- Claude added Brave Search as a subprocessor in March 2025. Anthropic also runs Claude-SearchBot to index pages and Claude-User to fetch them when someone asks, and its crawler page says blocking either may reduce your visibility.
- Gemini and Google AI Mode search Google's own index. Google's guide to AI features says a page only needs to be indexed and eligible for a snippet, with no extra technical requirements or special schema.
- Perplexity searches the web in real time for each question and attaches numbered citations to every answer. Its crawler docs ask sites to allow PerplexityBot, including in firewall rules.
What is query fan-out?
Query fan-out is when an assistant turns one question into several narrower searches, reads the results, then writes one answer. Google says AI Overviews and AI Mode may issue multiple related searches across subtopics and data sources. OpenAI gives a worked example: ChatGPT rewrites a question into a targeted query for a search partner, reads the results, then sends more specific follow-up queries.
For a Shopify brand, that changes a product page's job. It won't be matched to "best running shoes for flat feet" as a whole, it gets matched to the pieces: who the product is for, how it compares, what reviewers say, what it costs and whether it's in stock. A page that answers those pieces in plain words gives the assistant something to quote, while a page that only says "engineered for performance" gives it nothing.
How are citations and product results chosen?
Google says its models find more supporting pages while the answer is being written, so AI Overviews show a wider spread of links than normal results. For products, every assistant leans on structured data more than prose.
- ChatGPT shows a product when it judges it relevant to the shopper's intent, weighing structured metadata such as price and description from first-party and third-party providers, other third-party content and model-written summaries of reviews from public websites. OpenAI says product results aren't ads. When several merchants sell the same item, they're ranked on availability, price, quality and whether the seller is the maker or primary seller.
- Shopify stores are already in ChatGPT through Shopify Catalog with no extra work, OpenAI says, and Shopify's Agentic Storefronts also send products to Google AI Mode, Gemini and Microsoft Copilot. Brands can also apply to send a product feed through OpenAI's Agentic Commerce Protocol.
- Google AI Mode and Gemini draw on the Shopping Graph, which Google says holds more than 50 billion listings, with 2 billion refreshed every hour (Google). The way in is Google Merchant Center.
So the maker, with an accurate feed and stock at a fair price, wins the merchant list. Getting chosen at all still depends on what the web says about you.
We asked 3 assistants the same shopping question
On 20 October 2026 we typed "best running shoes for flat feet" into ChatGPT (a temporary chat with personalisation off), Claude (an incognito chat) and Google AI Mode, once each, from the US. That's one run each, not a study, and answers change daily, but the pattern is worth seeing.
ChatGPT and Claude both put the Brooks Adrenaline GTS 25 first, while AI Mode picked an adidas shoe. Two shoe lines appeared in all 3 answers: the ASICS Gel-Kayano and the New Balance 860. The sources barely overlapped. ChatGPT cited two Runner's World guides and the running app Runna, Claude cited specialist retailers such as Running Warehouse plus brand sites, and AI Mode cited RunRepeat, Doctors of Running, a podiatrist's YouTube video and a thread on Reddit.
Two details for brand owners. ChatGPT showed product cards with prices and labels such as "Best overall" and "Best value". And its link to Runner's World carried utm_source=chatgpt.com, so those referrals can show up as their own source in your analytics.
Why agreement across sources moves brand mentions in ChatGPT, Gemini and Perplexity
The shoes that made all 3 lists are the ones that different kinds of source (reviewers, retailers, forums) kept agreeing on. Each assistant got there by a different path, and you can't predict which pages it will read, so you want the same story told on all of them.
The best public data points the same way. An Ahrefs study of 75,000 brands found branded web mentions had the strongest correlation with visibility in Google's AI Overviews (0.664), well ahead of backlinks (0.218). Correlation isn't causation, and it covers AI Overviews only, but it fits fan-out: more independent pages naming you means more chances to be retrieved. For an ecommerce brand, the sources that count are:
- Reviews on your own product pages and on retailers' pages, since ChatGPT summarises reviews from public websites.
- Reddit and forum threads where real customers compare you with the alternatives.
- Editorial roundups and comparison sites in your category, the Runner's World and RunRepeat of your niche.
- Your own product, collection and comparison pages, written to answer the fan-out questions.
- Retailers and marketplaces that stock you, with the same product names and specs everywhere.
How to get recommended by ChatGPT: 6 steps
- Let the bots in. Allow OAI-SearchBot, Claude-SearchBot, Claude-User, PerplexityBot and Googlebot in robots.txt and in any firewall, and submit your sitemap to Bing Webmaster Tools, since Microsoft is one of ChatGPT's search partners.
- Get your catalogue into the feeds. Check Agentic Storefronts in your Shopify admin, keep Merchant Center clean, and make titles, prices and stock match on every retailer that sells you.
- Write for the sub-questions. Say who the product is for and who it isn't for, how it compares with the 2 alternatives shoppers mention most, and what it costs. Our AEO for ecommerce playbook goes through it page by page.
- Earn agreement off-site. Ask buyers for reviews, get into category roundups, and join the Reddit threads about your category honestly.
- Test monthly, logged out. Run your top 10 buyer prompts in a temporary or incognito chat in each assistant, and note who gets named and what gets cited.
- Start with the products that make money. More on that below.
Paid placement is a separate track: ads in ChatGPT sit apart from product results, and our breakdown of ChatGPT ads runs the profit maths before you test.
Get recommended for the products that make you money
Most AI visibility reports count how often your brand is mentioned. They don't ask which product was mentioned. Say ChatGPT keeps recommending your $29 starter product at a 12% contribution margin and never names your $90 bundle at 45%. That's $3.48 of contribution margin per order against $40.50, and your share of voice can look great while your P&L doesn't move.
So choose the products you want recommended by contribution margin in dollars per order, after COGS, shipping, fees and returns, and point your AEO work at those first. StoreHero already calculates contribution margin for every product from your Shopify, ad and cost data, and the product profitability report in our AI Skills Library ranks them for you in Claude or ChatGPT. To ask Claude about your store directly, see our guide to connecting Claude to Shopify.
If you'd rather not run all of this yourself, StoreHero AI Search can do it for you on autopilot. It asks ChatGPT, Gemini and Google AI the questions your buyers ask, shows who they name instead of you and which sources they cite, then writes researched, SEO-backed content for your store and publishes it to Shopify once you approve it. It keeps tracking those questions, so you can see what moved, get in early and lean less on paid ads.
Book a demo and we'll run a free AI Search report on your store: how ChatGPT, Gemini and Google AI answer your buyers' questions, who they name instead of you, and the content we'd write to get you named. Then StoreHero AI Search can keep it running on autopilot.
Key takeaways
- AI assistants answer shopping questions from two sources: what the model learned in training and what it finds in a live web search.
- They search different indexes: ChatGPT uses partner search providers including Microsoft, Claude uses Brave Search, Gemini and AI Mode use Google, and Perplexity searches the web in real time.
- Query fan-out splits one shopper question into several narrower searches, so product pages need to answer the sub-questions: who it's for, how it compares, what it costs.
- For products, assistants lean on structured data: Shopify Catalog, product feeds, Merchant Center and the Shopping Graph, plus summaries of public reviews.
- In our 20 October 2026 test, top picks and cited sources differed by assistant, but the same 2 shoe lines appeared in all 3 answers.
- Choose which products to push in AI answers by contribution margin in dollars, not by share of voice.
Frequently asked questions
How do you get ChatGPT to recommend your brand?
Make sure ChatGPT can reach and trust your pages: allow OAI-SearchBot, keep your product data accurate through Shopify Catalog or a product feed, write product pages that answer shoppers' sub-questions, and earn consistent reviews and mentions on retailers, forums and comparison sites. Then prioritise the products with the highest contribution margin, which StoreHero calculates for every product.
Can you pay ChatGPT to recommend your business?
Not in its product results. OpenAI says ChatGPT's product results are selected independently, aren't ads and aren't influenced by partnerships. ChatGPT does sell ads, but they're shown separately from the answer, so paying for an ad doesn't make the assistant recommend you.
How does ChatGPT decide which brands to recommend?
It weighs what it already knows from training, then searches the web and product data when current information helps. For shopping it considers your query and context, structured product metadata such as price and description, third-party content and review summaries, and it names products that relevant sources agree on.
Does ChatGPT use Google or Bing?
ChatGPT uses partner search providers rather than Google, and OpenAI's help pages name Microsoft and Shopify among the providers it shares queries with. Claude uses Brave Search, while Gemini and Google AI Mode use Google's own index.
How do you get AI to recommend your products?
Do the same work across every assistant: let their crawlers in, get your catalogue into Shopify Catalog and Google Merchant Center, answer comparison questions on your own pages, and build agreement across reviews, Reddit and press. Test your top buyer prompts monthly in a logged-out or incognito chat, and use StoreHero's product-level contribution margin to decide which products to push first.
Sources
- OpenAI Help Center: Searching the web with ChatGPT
- OpenAI Help Center: Shopping with ChatGPT Search
- OpenAI: Overview of OpenAI crawlers
- OpenAI: Agentic Commerce Protocol, get started
- Anthropic: Web search tool documentation
- Anthropic: Does Anthropic crawl data from the web?
- Simon Willison: Brave Search added as an Anthropic subprocessor (March 2025)
- Google Search Central: AI features and your website
- Google: Shopping in AI Mode and the Shopping Graph
- Perplexity: Crawlers documentation
- Shopify Help Center: Agentic storefronts
- Ahrefs: brand mentions and AI Overview visibility, 75,000 brands (May 2025)
