How to Optimise Your Ecommerce Store for ChatGPT Search (Practical Guide)

Shoppers used to search "best bed sheets" on Google and scroll through the blue links. Now they ask [ChatGPT](/blog/chatgpt-already-real-search-engine-brand-2025) directly: *"I sleep hot and have sensitive skin - can you recommend breathable bed sheets that won't irritate me?"* The shift sounds incremental. The shape of the work changes completely. Traditional SEO targets keyword rankings. ChatGPT targets answers. If you're not in the answer, you're invisible - and your traffic decline shows up months before your sales decline does. This guide is the practical version. Three sections: how to be technically visible to ChatGPT's crawlers, how to structure your data so the model can use it and how to write content that gets cited. ## 1. Make your store visible to ChatGPT and AI bots Before any optimisation work matters, the model has to be able to read your site. Most ecommerce stores have at least one technical barrier that makes them partly invisible. **Update your robots.txt.** Allow the AI bots that should be reading you. The crawlers that matter for ChatGPT and its commerce experiences are GPTBot (used by OpenAI for crawling) and ChatGPT-User (used when ChatGPT browses live). Add explicit allow rules: ``` User-agent: GPTBot Allow: / User-agent: ChatGPT-User Allow: / ``` If your robots.txt is silent on these, you're relying on default behaviour - which varies. Make it explicit. **Solve the JavaScript problem.** This is the one most brands miss. AI crawlers differ from Googlebot in one critical way: most of them don't fully execute JavaScript. So if your product titles, descriptions, prices and availability are rendered client-side via React, Vue, or any modern framework, the AI sees a near-empty shell. The fix: server-side rendering (SSR) or hybrid rendering for product pages. If you're on Next.js, App Router pages by default render server-side - verify your PDP routes do too. If you're on Shopify, the storefront is server-rendered already, but custom apps and headless setups need explicit handling. Tools like Prerender.io serve pre-rendered HTML to crawlers if SSR isn't an option. We've covered the diagnostic side in [what LLMs actually read on your website](/blog/what-llms-actually-read-your-website-2025). Run the disable-JavaScript test on your top five pages before doing any optimisation work - if the page is empty without JS, that's your priority job. **Maintain technical fundamentals.** Broken links, slow page speeds, weak Core Web Vitals - all of these matter for AI crawlers too. The technical hygiene that helped Google for the last decade still helps the AI surfaces now. The bar isn't higher. It's just non-negotiable. ## 2. Use structured data and product feeds the AI can read Structured data is the bridge between your store and the model. Brands that ship clean schema get featured 3–5× more often in AI-generated shopping recommendations than brands that don't. The model treats structured data as high-confidence: it can extract attributes from JSON-LD without guessing, so it does. (Our [foundational guide to Schema.org for LLM optimisation](/blog/what-schemaorg-foundational-guide-llm-optimisation-2026) walks through which types matter most.) The two layers that matter for ecommerce specifically. **On-site schema markup.** Product schema is the minimum. Include `name`, `brand`, `sku`, `gtin`, `price`, `priceCurrency`, `availability`, `image`, `description`, `aggregateRating` and `review`. The fields most often skipped - `gtin` and `aggregateRating` - are the ones that matter most for trust signals. Add FAQPage schema on PDPs and category pages. Add Organization schema on the homepage with `sameAs` links to your social and retail profiles. > **Update - May 2026:** Google deprecated FAQ rich results on 7 May 2026. `FAQPage` schema no longer earns a visible feature in Google Search. It still has value for AI assistants - ChatGPT, Claude and Perplexity use it to parse Q&A content - but you do not need it for a Google rich result. **ChatGPT product feeds.** Some assistants now ingest structured product feeds directly. The format varies by platform (CSV, TSV, XML, JSON), but the fields are consistent: unique product IDs, detailed titles and descriptions, links and images, pricing and inventory. Match each platform's spec and submit through the supported channels. The principle behind all of this: AI systems don't guess. They prefer explicit. The brand whose data is structured wins the citation; the brand whose data is buried in JavaScript or scattered across mismatched feeds doesn't. ## 3. Shift from keywords to prompts and build entity presence The search approach has evolved beyond keywords to natural language prompts. Traditional SEO focused on ranking for phrases like *"best coffee maker under $100."* Now shoppers ask AI: *"I'm training for a marathon in rainy weather - what's a lightweight shoe with good grip?"* This is a different optimisation target. Your content needs to mirror the constraint-stack of real prompts. **Build prompt-led content.** Map content to: - **Customer personas.** Hot sleepers, allergy sufferers, marathon trainees, busy parents. - **Specific use cases.** Travel, gifts, replenishment, gifting, peak occasions. - **Problems your products solve.** Sensitivity, durability, compatibility, sustainability. Each combination is a potential prompt. The brand that pre-answers the most relevant prompts gets cited. For example, [TRIP Drinks](/blog/trip-drinks-ai-search-visibility-lift-2026) won prompt presence in evening-relaxation and wind-down categories by writing FAQ and category content specifically around constraint-led queries: *"non-alcoholic drinks for evening that aren't just water"*, *"CBD drinks that taste good"*, *"functional drinks for sleep without caffeine."* That's the language pattern. Generic "best functional drinks" doesn't get cited the same way. **Build entity authority across multiple platforms.** ChatGPT references search engine results plus high-trust domains. Mentions in publications like TechCrunch, Forbes, BBC, or category-specific authority sites (Runner's World for running, Allure for beauty, etc.) substantially improve the chance of model citation. The work here is editorial - being a brand that genuinely earns coverage in the places shoppers and models trust. **Monitor your AI visibility regularly.** Tools like [Share of Model](https://shareofmodel.ai/) track your content's appearance in AI-generated answers. They reveal which content formats earn the most citations, which prompts you're winning and which competitors keep showing up in your category. Without this, you're optimising blind. ## Why this matters now ChatGPT is already operating at 4–12% of Google's search volume by usage and traffic from it converts at up to 9× organic search rates. We covered the data in detail in [ChatGPT is already a real search engine](/blog/chatgpt-already-real-search-engine-brand-2025). The brands building AI visibility now are establishing positions that compound. AI systems get more confident in brands they've encountered repeatedly with consistent, structured information. Late adopters don't just have to do the work - they have to overcome the model's existing preference for the brands that got there first. ## What to do this week Three concrete moves you can ship before next Monday. **Run the disable-JavaScript test on your top five pages.** Note what disappears. If your PDPs are empty without JS, you have a server-rendering problem and that's the priority fix. **Audit your robots.txt.** Make GPTBot and ChatGPT-User explicit allow rules. Two lines of config and you've removed a barrier that may be silently blocking you. **Pick three priority prompts and check them.** Run *"best [your category] for [your top use case]"* through ChatGPT, Gemini, Claude and Perplexity. Log what each model says. If you're absent or misrepresented in any of them, you've got your first content priority. If you want help executing the full programme, [Lmo7 builds AI search visibility for consumer brands every week](/contact). The audit costs nothing. The work that follows is what protects your category position over the next 12–24 months. That is the work.