AI SEO for ecommerce: how to get your store recommended by ChatGPT

Shoppers used to type a few words into Google and click through a page of links. Now many ask an AI assistant a full question, like “what's the best trail running shoe for wide feet under $150?”, and get a short answer that names three or four brands. If your store isn't one of them, you don't get a chance to make your case.

Getting named in those answers is what people mean by AI SEO. You'll also see it called AEO (answer engine optimization) or GEO (generative engine optimization), which are other names for the same job. This guide covers the basics first, the two things every store should sort out before anything else. Then it gets into the special sauce: nine steps for owning one category's answers, from a database of real shopper questions to a category page that assistants want to quote.

It matters because the answer is now where much of the choosing happens. When Pew Research tracked 900 US adults' Google searches in March 2025, people clicked a regular result far less often when an AI summary appeared, and they almost never clicked the links inside the summary.

How AI assistants choose the brands they name

Most AI answers about products come together in three moves. The assistant turns the question into several searches, reads what comes back, then writes an answer and cites a few of the pages it used. Google calls the first move query fan-out. Its AI Overviews and AI Mode may run several related searches across the parts of a question, and they can show a wider set of links than the classic results do.

Fan-out changes who gets cited. In July 2025, Ahrefs found that 76% of the pages cited in AI Overviews also ranked in the top 10 for that search. By March 2026 the figure was 38%, and roughly a third of cited pages ranked outside the top 100. Ahrefs puts the change down to fan-out, a newer Gemini model behind AI Overviews, and partly its own better measurement. Other assistants stray even further from Google's rankings. In an August 2025 study, only about 12% of the pages that ChatGPT, Gemini and Copilot cited were in Google's top 10 for the same question.

Classic SEO is still the entry ticket, since nothing gets cited from a page the assistant never found. Ranking first just matters less than it did. An assistant that searches for each small part of a question can pick up a page that answers one part well, even if that page doesn't rank for the main search. It also leans on what other sites say about the brands involved.

Part 1: the basics

Two steps come before any writing. Pick the category you're going to win, then make sure AI assistants can read your store at all. Skip either one and the rest of the work goes to waste.

Step 1: Pick one category

Most shopping questions are about a category, not a brand: “best dog food for itchy skin”, “linen sheets that don't crease”, “nail polish that's safe for kids”. You won't win every category at once, so pick one and win it properly before you move on.

Choose the category where three things overlap:

• It makes you money. Good margins, and products people buy again or buy together.
• You have a real reason to win. A material, an ingredient, a fit or a price that nobody else matches.
• People research it. Shoppers ask questions before they buy, rather than grabbing the first one they see.

Every step after this one runs on that single category. It keeps the work small enough to finish and the results clear enough to read.

Step 2: Run a technical audit

Before you write anything new, check that AI assistants can actually reach and read your store. Start with your robots.txt file, because each AI company now runs separate bots for separate jobs. OpenAI's GPTBot collects training data. OAI-SearchBot is the one that matters for shoppers, because OpenAI says sites that block it won't be shown in ChatGPT's search answers. Anthropic splits its bots the same way, with ClaudeBot for training and Claude-SearchBot for search. So you can say no to training and still be found.

Google works differently. Blocking Google-Extended keeps your content out of Gemini's training and out of the live searches the Gemini app uses to answer, but it doesn't change your place in Google Search. AI Overviews and AI Mode have their own switch. Since mid-2026, Search Console lets a site opt out of them, and Google says sites that do get no traffic or impressions from those features. Make sure nobody has turned it on by mistake.

Then check what the bots can see. When Vercel and MERJ studied AI crawlers in 2024, none of the major ones ran JavaScript, OpenAI's, Anthropic's and Perplexity's included. Google's and Apple's crawlers were the exceptions. Shopify themes send most product details in the page itself, but some apps for reviews, FAQs or size guides load their content with JavaScript after the page opens. An AI crawler may never see those words. If your store is headless, check that pages are built on the server.

Part 2: the special sauce

With the basics done, assistants can read your store, but that alone won't get you named. The next nine steps are where most stores never go. They take you from a pile of shopper questions to a category page that assistants cite, with a benchmark that shows whether it's working.

Step 3: Build a keyword database

Classic keyword research hunts for short phrases with lots of searches. Questions to AI assistants are longer and more specific, and many never show up in a keyword tool. So build a database of the real questions shoppers ask about your category, in their own words.

Pull them from these places:

• Google Search Console. Filter your queries to long ones, or ones that start with how, what, which, best or vs. These are the closest thing you have to AI prompts.
• Google's “People also ask” boxes, and the related searches at the bottom of the results page.
• Reddit and forums. Search your category name plus “reddit” and note the questions people ask, and which brands others recommend in reply.
• Your reviews and support inbox. The questions people ask before they buy, and the reasons they give for choosing you.
• Competitors' reviews on Amazon. Their complaints point to questions nobody has answered well yet.
• A keyword tool, to see which of the shorter versions get searched most.

Give each question its own row, with these columns:

Don't wait for the list to be perfect. A first pass of a hundred or so questions is plenty to start with, and it grows every time you run a benchmark.

Step 4: Assess your topic coverage, quick and deep

Group the questions into topics, the handful of themes your category keeps coming back to. For a dog food brand those might be allergies, puppies, switching food, ingredients and cost per day. Then check how well your site covers each topic, in two passes.

The quick pass takes an afternoon. For each topic, is there a page whose main job is to answer it? Mark each one covered, thin (only mentioned on a page about something else) or missing.

The deep pass is for the topics that matter most. Take the page you think covers a topic and test it against every question in that topic. Does it answer each one directly, in a sentence someone could quote? Does it answer the follow-ups an assistant would search for during fan-out? Then open the pages assistants cite for that topic today, and list what they cover that yours doesn't. It's usually something concrete: a sizing chart, a comparison table, a cost per serve, a straight answer to “is it safe for…”.

Step 5: Sort questions by the five stages of awareness

In his 1966 book Breakthrough Advertising, the copywriter Eugene Schwartz described five stages of awareness a buyer moves through. They're a handy way to sort your database, because shoppers ask assistants very different things at each stage, and each stage needs a different kind of page.

Most stores only have pages for the last two stages. That's where the sale happens, but the shortlist forms earlier. When someone asks for the best food for an itchy dog and the assistant names three brands, those three are likely to be the ones the shopper compares next. Tag every question with its stage, and the gaps usually show up straight away.

Step 6: Write a topic domination plan

You now have topics down one side, stages across the top, and a mark in each box for covered, thin or missing. That grid is your plan, and the aim is to have the clearest page on the web for every box in it that makes you money.

Work through it in this order:

1. 1.The category page first. It should answer the solution-aware questions, the ones that start with “best”. Step 10 covers what goes on it.
2. 2.The missing solution-aware and product-aware pages next: buying guides, “best X for Y” pages, and fair comparisons with the brands you keep losing to.
3. 3.Then problem-aware guides. They pay off more slowly, but they get your brand into the conversation earlier and earn links and mentions from other sites.
4. 4.Unaware content last, unless it fits naturally with what you already publish.

Link every page in the topic to the category page, and link the category page back to each of them. Put a real date on every page, and update it when the facts change: prices, ranges, stock, the year in a title. Freshness counts more here than in classic search. When Ahrefs studied 17 million cited pages, the ones AI assistants cited were about 26% newer on average than the ones in regular search results, and ChatGPT leaned hardest toward recent pages.

Two rules for the new pages. First, write each one for a real question in your database. Google warns that pages mass-produced to target small variations of a search can break its spam rules. Second, be fair on comparison pages. Say where a competitor is the better pick. An assistant reads the rest of the web too, and a page that admits a trade-off is easier to trust and to quote.

Step 7: Create a benchmark report

The platforms only show you part of the picture. Since June 2026, Google Search Console has a generative AI report that shows impressions from AI Overviews and AI Mode by page, with no clicks or queries. Bing Webmaster Tools has an AI Performance report that shows your citations in Copilot. Neither tells you what ChatGPT or Perplexity say about you, or which brands they name instead. So build your own benchmark:

1. 1.Pick 30 to 50 prompts from your database. Spread them across the stages, with the most at solution aware.
2. 2.Run each one in the assistants your customers use: ChatGPT, Google (AI Mode and AI Overviews), Perplexity, Gemini and Copilot. Use a logged-out or fresh session so past chats don't color the answer.
3. 3.Run each prompt a few times. Answers change from one run to the next, so a single answer is only a sample.
4. 4.For each answer, note whether your brand is named and where in the list, whether your site is cited and which page, which other brands and pages appear, and whether what it says about you is right.
5. 5.Work out your share of voice: the share of answers that name you, next to the same number for each competitor.

Repeat it every month with the same core prompts. Add new ones as your database grows, but keep the core set fixed, or the trend line means nothing. Paid tracking tools can run this at scale. For one category, a spreadsheet does the job.

Check your analytics too. Look for visits from chatgpt.com, perplexity.ai, gemini.google.com and copilot.microsoft.com, and for links tagged utm_source=chatgpt.com, which OpenAI says ChatGPT adds to the links it sends. Visits from Google's AI Overviews and AI Mode are counted inside your normal Google search traffic, so you can't split them out there.

Step 8: Identify your top competitors, then compare category by category

The benchmark shows who you're really up against, and it may not be who you'd guess. Your search rivals are the stores ranking next to you on Google. Your AI rivals are whoever the answers name, and that can include marketplaces, review sites, and brands much bigger or smaller than yours.

Take the five brands named most often for your category and compare them with yours, side by side:

Keep asking one question as you go: what does an assistant get from their pages that it can't get from yours?

Step 9: Extract citation opportunities

Every answer in your benchmark cites sources. Collect every cited URL across all your prompts and you have a map of the pages that shape your category's answers. Most of them won't be yours, and each one is a chance to be included.

Those outside pages carry real weight. When Ahrefs looked at 75,000 brands, the number of times a brand was mentioned across the web was more closely linked to its visibility in AI Overviews than anything else it measured, backlinks included. That's a correlation, so it doesn't prove mentions cause citations, but it fits the way assistants build answers from what other sites say. Ahrefs' tracking of 3 million US searches also finds YouTube and Reddit are the two sites AI Overviews cite most.

Sort the URLs into groups and work each group:

Step 10: Audit your category page

On most Shopify stores the collection page is a title, some filters and a grid of products. That works for someone who already knows what they want. It gives an assistant almost nothing to quote when someone asks which one is best.

Research backs up what to add. In a study published at the KDD conference in 2024, researchers from Princeton and elsewhere tested changes to pages and measured how visible they became in AI answers. Adding quotations, statistics and cited sources all helped. Keyword stuffing made pages less visible. Pages lower down the results gained the most, which is good news if you don't rank first. The tests ran on a simulated AI search engine, and the top method also helped on Perplexity.

Check your category page against this list:

Step 11: Finish with technical optimization

The last step locks the work in. Go back over the audit from step 2 with your new pages, then tidy up the data that machines read about your products.

Start with product feeds, because they matter more for shopping answers than anything on the page. Shopify's Agentic Storefronts sends your catalog to ChatGPT, Google's AI Mode and Gemini, Copilot and Meta, and Shopify switches it on by default for eligible stores. Google Merchant Center feeds Google's shopping results, and in January 2026 Google added new product attributes meant for conversational shopping. OpenAI says ChatGPT ranks the merchants it shows partly on availability, price and whether they're the maker or main seller. So keep titles, prices, stock and variants complete, and identical everywhere they appear.

Structured data is the most oversold part of AI SEO. Google says it isn't required for its AI features and that there's no special AI markup. When Ahrefs tested 1,885 pages that added it, it found no major lift in AI citations. Keep accurate Product markup anyway, with price, stock and reviews, because it supports your product details in Google's results, and Microsoft has said it uses schema to help its models understand pages. It won't rescue a thin page, though.

A few smaller jobs finish it off:

• Send IndexNow pings when pages change. Bing, which feeds Copilot, uses them. Google doesn't.
• Keep URLs stable. A product moved without a redirect loses whatever citations it had.
• Treat llms.txt as optional. It's a plain-text guide to your site for AI tools, proposed in 2024. Google says its search doesn't use it, and an Ahrefs study in June 2026 found 97% of the llms.txt files it checked got no requests at all in a month.
• Watch WebMCP, a proposal in Chrome that lets a site hand AI agents tools like search or add to cart. No answer engine has said it uses it to pick sources.

Checkout is changing too. OpenAI launched Instant Checkout in ChatGPT in September 2025. In March 2026 it was reported to be scaling that back, and Shopify's help pages now say ChatGPT sends buyers to the store's own checkout. Google launched Universal Cart in the US on 15 October 2026, built on its Universal Commerce Protocol, with the retailer still the seller of record. Whichever way checkout goes, the assistant has to pick your product before anyone can buy it.

Putting it together

Then run the benchmark again next month and see what moved. When the category is yours, pick the next one and start again at step 3.

Sources

• Google Search Central, “AI features and your website” (updated 10 December 2025) and “Google's guide to optimizing for generative AI features on Google Search” (updated 10 July 2026).
• Google Search Central blog, “Top ways to ensure your content performs well in Google's AI experiences on Search” (21 May 2025) and the generative AI performance report announcement (June 2026).
• Google, documentation of its common crawlers (Google-Extended), and the Search Console opt-out for AI features (announced 3 June 2026).
• OpenAI, crawler documentation. Anthropic, help article on its crawlers (7 April 2026). Perplexity, crawler documentation.
• Vercel and MERJ, “The rise of the AI crawler” (17 December 2024).
• Pew Research Center, “Google users are less likely to click on links when an AI summary appears in the results” (22 July 2025).
• Ahrefs: AI Overview citations and the top 10 (21 July 2025 and 2 March 2026), overlap between AI assistants and Google (11 August 2025), brand mentions and AI Overviews (26 May 2025), freshness of cited content (28 July 2025), most-cited domains in AI Overviews (updated 2 September 2026), schema and AI citations (11 May 2026), llms.txt study (15 June 2026).
• Aggarwal and others, “GEO: Generative Engine Optimization”, KDD 2024.
• Eugene Schwartz, Breakthrough Advertising (1966).
• Shopify, Agentic Storefronts help page and announcements (10 December 2025, 24 March 2026). Google, Universal Commerce Protocol (11 January 2026) and Universal Cart (15 October 2026). Bing Webmaster blog, AI visibility insights (16 June 2026). Cloudflare blog (1 July 2025). IndexNow participating search engines.