# Generative engine optimization for online stores: ChatGPT, AI Mode Generative engine optimization for online stores: what ChatGPT, Gemini and Google AI Mode read, product data, llms.txt, robots.txt and a 12-step checklist. Source: https://cloudcart.com/en/blog/geo-ai-search-online-store · CloudCart · 2026-11-06 Generative engine optimization (GEO) prepares your online store to be quoted and recommended by ChatGPT, Gemini and Google AI Mode. Four things decide it: indexed pages, complete product data (price, stock, delivery, reviews), direct answers with numbers, and mentions on other sites. In our check of 82 ecommerce searches in Bulgarian on Google.bg, 81 showed an AI Overview above the results. ## What generative engine optimization is and why it matters for a store GEO is the work on your content and data that makes generative search engines use your site as a source. Classic SEO competes for a place in the list of ten links. GEO competes for a place inside the answer itself: in the paragraph the AI writes, in the product cards it shows and in the sources it cites under the text. This guide is by the CloudCart team, the Bulgarian ecommerce platform, for stores selling in or from Bulgaria and across the EU, and it shows which parts of GEO the platform handles for you. The term comes from a study by Princeton and IIT Delhi (Aggarwal et al., ["GEO: Generative Engine Optimization"](https://arxiv.org/abs/2311.09735), presented at the KDD conference). The authors ran 10,000 queries through a generative engine and compared which edits to a page raised its visibility in the answer. Adding statistics, quotations and cited sources helped most, with up to 40% more visibility. Keyword stuffing barely helped. For a store this is about sales, because the way people search is changing: in Google Ads data for Bulgaria, "ai mode" alone is searched 9,900 times a month on a yearly average. Shoppers increasingly type a whole question ("coffee machine for a small office under €400 with next-day delivery") and get a ready shortlist from the AI. A store outside that shortlist loses the customer before the click. The next step is an AI agent that also places the order for the customer. ### GEO, AEO, LLMO and AI SEO: the terms in brief Agencies use several names for related things. For a store the work is the same; the difference is where you look for the result: | Term | What it means | Where you see it | | GEO (Generative Engine Optimization) | Optimizing for generative engines that write one answer from several sources | ChatGPT, Gemini, AI Mode, Perplexity | | AEO (Answer Engine Optimization) | Optimizing for direct answers: featured snippets, voice assistants, answer boxes | Snippets in Google, voice search | | LLMO (Large Language Model Optimization) | How large language models describe your brand, including from memory | Answers given without a web search | | AI SEO | An umbrella term for SEO adapted to AI search; often used as a synonym for GEO | Everywhere | | AIO | Short for AI Overviews, sometimes for AI optimization in general | The summary above Google's results | AI Overview is Google's AI-written summary above the search results, and it is one of the first places where GEO work shows. ## SEO vs GEO: what is different for an online store GEO builds on SEO. AI Overviews and AI Mode take their sources from Google's index, and ChatGPT searches the web through its own crawlers and partner indexes. A page that is not indexed, or is weak in classic search, rarely makes it into an AI answer. The difference is in what you measure and which content wins. | | SEO | GEO | | Where the store appears | Google's results list, Google Shopping | Answers in ChatGPT, Gemini, AI Mode, AI Overviews, Perplexity | | What you measure | Rankings, impressions, clicks, organic orders | Mentions, place in the shortlist, cited pages, visits from chatgpt.com and gemini.google.com | | What wins | Keyword in the title, internal links, speed, backlinks | A direct answer in the first 60 words, numbers, tables, FAQ, the same facts everywhere, mentions on independent sites | | Product data | Meta title and description, URL, image | Price, stock, delivery, returns, reviews, attributes the AI can compare | | The query | 2 to 4 words: "office coffee machine" | A full question with conditions: budget, deadline, city, use | | Time to results | Weeks to months | Months; answers change daily, so you measure with a fixed list of questions | For an online store, most of the work is shared. A clean catalog, complete product pages and clear delivery terms help both classic [store SEO](https://cloudcart.com/en/seo) and AI. The full technical list is in our [ecommerce SEO checklist](https://cloudcart.com/en/blog/online-store-seo). ## What Google AI Mode is and how it affects stores AI Mode is a tab in Google Search where Gemini answers the question with text, comparisons and links, and the user carries on with follow-up questions. Bulgarian is on Google's [list of supported AI Mode languages](https://support.google.com/websearch/answer/16011537?hl=en), as is English, so your store can appear for Bulgarian and English queries alike. Technically Google breaks the question into many smaller searches ("query fan-out"), runs each one and builds the answer from different pages. For shopping, AI Mode uses the Shopping Graph: Google's database of more than 50 billion product listings, over 2 billion of which are refreshed every hour, according to Google. That is where the product cards with price, stock, reviews and store come from. Data enters the Shopping Graph from websites and from Google Merchant Center. What this means in practice: - A product without price and stock in Google cannot be compared. AI Mode prefers listings where it knows the price, the delivery time and whether the item is in stock. - There is no special file or markup for AI Mode. Google's guide to AI features in Search says the usual requirements apply: the page must be indexed and eligible to show a snippet. You do not need new machine-readable files, AI text files or special schema.org markup. - AI Mode traffic is reported in Search Console , in the overall Performance report under the "Web" search type. There is no separate AI Mode filter, so watch changes in impressions and clicks per page. How to connect your catalog to Merchant Center is explained step by step in [Google Merchant Center and Google Shopping in Bulgaria](https://cloudcart.com/en/blog/google-shopping-merchant-center-bulgaria). ### AI Overview: the summary above the results AI Overview is the answer box Google shows above the links, with sources next to the text. We checked 82 Bulgarian-language searches about ecommerce: an AI Overview appeared in 81 of them, and the "People also ask" box in only 2. For most commercial questions in Bulgaria, the first thing a shopper sees is an AI answer. The search for the Bulgarian term for GEO shows which pages get in. The AI Overview cited six Bulgarian pages about SEO, mostly from agencies. Most of them open with a definition in the first paragraph and end with a questions-and-answers block. That is the recipe for a category page or a store blog post too: the direct answer at the top, the sub-questions as headings, the FAQ at the end. ## How ChatGPT and Gemini choose which store to recommend: our test We ran 30 questions that merchants ask when choosing a platform (25 of them in Bulgarian) through the API of ChatGPT's model (GPT-5.5) with web search, and through Gemini 2.5 Flash with Google Search. We recorded who was mentioned, in what position and which pages were cited. This is what got cited most, ordered by weight: | Format | Why the AI picks it | What it means for your store | | Pricing page with a table | Answers "how much does it cost" with numbers; our pricing page with its FAQ was cited 9 times, more than any other page of ours | Price, delivery cost and delivery time visible as text, not only in an image | | "X vs Y" comparison | A ready table for "which one should I choose" | Comparison tables in categories and the blog: model, capacity, price, who it is for | | Step-by-step guide | Numbered steps are easy to retell | Buying and care guides: "how to choose a size", "how to maintain it" | | Partner page | AI trusts a third party more than the store itself | Presence on the sites of the brands you sell, suppliers, price comparison sites and marketplaces | | Questions-and-answers block | The question is phrased the way the user asks | FAQ on categories and products with your customers' real questions | The result in numbers: GPT-5.5 mentioned CloudCart in 15 of 30 answers, Gemini 2.5 Flash in 12 of 30. Gemini ran a Google search for 22 of the 30 questions and answered the rest from memory. The most cited external page was a partner page: the promo page of Microinvest, a Bulgarian accounting and warehouse software, about connecting to an online store, with 7 citations. Two lessons from the test apply to every store: - AI repeats old facts. In the test, the models quoted outdated prices and terms taken from old pages and from memory. If you have changed prices or terms, old pages with the old numbers should 301-redirect to the new ones, and the new pages should state the facts clearly. - Where there was no suitable page, there was no recommendation. Where we had a page with a direct answer, numbers and an FAQ, the AI cited it. Where the answer sat only in sales copy or was missing, the AI cited other people's guides and partner pages. That is why redirects matter most for GEO when you [switch ecommerce platforms](https://cloudcart.com/en/blog/switch-ecommerce-platform-checklist): old URLs stay in AI indexes for months after the move. When you [migrate to CloudCart](https://cloudcart.com/en/migration), 301 redirects are added one by one or thousands at a time from a CSV in Marketing → SEO → 301 redirects. ## The product data AI reads When an AI recommends a product, it compares attributes. The store with the most complete and consistent data on the site, in the feed and in reviews has the best chance of making the list. Here is what to check for each product and where to do it in CloudCart: | Data | Why it matters to AI | Where in CloudCart | | Name with brand, model and the main attribute | AI looks for a match with the conditions in the question | Product → main details | | Price and stock | Without them the product cannot enter a comparison by budget | Product → price and stock; to Google: Apps → Google Shopping | | Attributes (size, material, power, volume) | They answer "under 40 cm", "for a small room", "no lactose" | Product → properties | | Star reviews | Social proof the AI quotes ("rated 4.7 from 120 reviews") | Apps → Product Reviews | | Questions and answers | Ready passages for long questions | Apps → Product Reviews, "Questions and answers" (switched on in the app settings) | | Meta title and description | The snippet the AI often starts from | Product → SEO settings | | Delivery and returns | Questions like "by tomorrow", "easy returns" | "Delivery" and "Returns" pages, linked from the product | Structured data (schema.org Product). It tells Google in machine language the name, price, currency, stock and rating. In CloudCart the markup comes from the store's theme and depends on it. Test one product page in Google's Rich Results Test (search.google.com/test/rich-results) and look for these fields: - name , image , description and brand : what the product is; - offers with price , priceCurrency (EUR) and availability : the price and stock the AI compares; - aggregateRating and review : the stars and the number of reviews; - gtin or mpn if you sell branded goods: they let Google recognize that your listing is the same product as in other stores. If the test does not show "Product", or price and stock are missing, email us at support@cloudcart.com with the page URL. The barcode entered in the product goes to the Google feed as the GTIN. The Google feed. The Google Shopping app sends your catalog to Merchant Center. From there products reach the Shopping Graph, and through it AI Mode and Gemini, including the unpaid product listings. With auto-update switched on (in the app settings, with price and stock ticked), a new price saved in the admin goes to Google when you save the product. How many products the feed carries depends on your plan; see [prices and limits by plan](https://cloudcart.com/en/pricing). The same catalog can go to price comparison sites with an XML feed. They are among the sources an AI checks for price and stock. Important: the price on the site, in the feed and in reviews must be the same. A mismatch is a common reason for Merchant Center to disapprove a product and for the AI to show a competitor's offer. ## What llms.txt is and how to write one for your store llms.txt is a text file at /llms.txt where you tell AI assistants, in your own words, what you sell and which pages to read. The standard was proposed by Jeremy Howard ([llmstxt.org](https://llmstxt.org/)). It is voluntary: not every assistant reads it, and Google says AI Overviews and AI Mode need no new AI text files. The file is cheap to maintain and gives an accurate description to the assistants that do read it, so it is worth 15 minutes. In CloudCart the file is built in. Here is how to write it: - Open the card. Marketing → SEO → the "Llms.txt file" card (admin URL: /admin/marketing/seo). - Write a title and one sentence. The store name after # , and under it, after > , one sentence on what you sell, to whom and how you differ. - Add 8 to 15 links. Your main categories, the "Delivery", "Returns" and "About us" pages, your blog guides. One sentence for each link. - Add the facts AI gets wrong. Delivery time and cost, minimum order, payment methods, cities with a shop, opening hours. - Save. The public file updates within 5 minutes. Open yourdomain.com/llms.txt and check it. An example for a coffee store: # yourstore.bg > Specialty coffee beans, roasted every week in Sofia. Delivery within 24 hours across Bulgaria. ## Categories - [Coffee beans](https://yourstore.bg/category/coffee-beans): single-origin coffees with flavor profile and roast date - [Coffee machines](https://yourstore.bg/category/coffee-machines): manual and automatic, with a comparison table ## Terms - [Delivery](https://yourstore.bg/page/delivery): Econt and Speedy, within 24 hours; free shipping over €40 - [Returns](https://yourstore.bg/page/returns): 14 days ## Guides - [How to choose an office coffee machine](https://yourstore.bg/blog/office-coffee-machine) Econt and Speedy are the two largest Bulgarian couriers; naming them helps an assistant answer "how will it be delivered". What to know about the file in CloudCart, from the platform documentation: - It is written in Markdown, up to 65,536 characters, and published exactly as written. Do not put internal notes or customer-specific prices in it. - If you leave the field empty, the store publishes a short summary itself: the store name, the home page SEO description, and links to the home page and contacts. Your own text is better. - One text is served on every domain the store answers on. If you run separate storefronts on separate domains, write the links for the domain that matters most. - The file does not update itself with the catalog. When you change categories, update it too. With an AI assistant connected to the store, the new text takes one sentence. - On a trial or expired plan the URL returns 404 and the store is hidden from search engine crawlers. The file and indexing start once the plan is paid. > A store run by an AI assistant Your llms.txt and missing meta descriptions in one sentence Connect Claude Code, ChatGPT Codex, Gemini CLI or Cursor to your store with a PAT token and type: "Read my categories and draft an llms.txt" or "Show me products without a meta description and suggest texts". The assistant has only the access you gave the token: read-only, full access or selected permissions. How the AI assistant connects → ## How to get your store recommended in ChatGPT ChatGPT searches the web in real time when a question needs current data, and for shopping questions it also shows products. These are the steps that decide whether it finds you: - Allow OpenAI's search crawler. OAI-SearchBot indexes sites for ChatGPT search. GPTBot collects data for model training. According to OpenAI's crawler documentation , each setting is independent: you can block GPTBot and stay in ChatGPT search. Check robots.txt in Marketing → SEO → Robots.txt: there must be no Disallow for OAI-SearchBot. The platform adds the block on cart and checkout itself, so do not write it yourself. - Register the site in Bing Webmaster Tools. Bing's index powers Copilot. It takes minutes: submit your store's sitemap.xml there (the URL is in Marketing → SEO → the "Sitemap" card). - Write the answers people look for. A "Delivery" page with prices and times, comparisons in categories, buying guides. ChatGPT quotes a specific passage, so the fact must be in the text, close to its heading. - Get mentioned elsewhere. The site of the brand you sell, price comparison sites such as Pazaruvaj, marketplaces such as eMAG, media articles and forums. The AI checks whether independent sources say the same as the store. - Keep facts consistent. The same delivery price and the same delivery times on your site, in your Google Business Profile and on marketplaces. The channels that take your catalog to marketplaces and price comparison sites from one admin are on the [sales channels](https://cloudcart.com/en/sales-channels) page. ## Which AI crawlers to allow in robots.txt Each AI engine has its own crawlers, and most have two: one for answering users and one for training the model. For visibility in answers you need the first kind; the second is your choice. | Crawler | Owner | What it does | For a store | | Googlebot | Google | Google's index: results, AI Overviews, AI Mode | Must be allowed | | Google-Extended | Google | Gemini training; does not affect AI Overviews or AI Mode | Your choice | | Bingbot | Microsoft | Bing's index and Copilot; some other AI engines use it too | Allowed | | OAI-SearchBot | OpenAI | Search and product results in ChatGPT | Allowed | | ChatGPT-User | OpenAI | Opens a page when a user asks about it in the chat | Allowed | | GPTBot | OpenAI | Model training | Your choice | | PerplexityBot | Perplexity | The index behind Perplexity's answers | Allowed | | Claude-SearchBot | Anthropic | Search in Claude | Allowed | | ClaudeBot | Anthropic | Model training | Your choice | If you decide to block a training crawler, name it explicitly, for example User-agent: GPTBot with Disallow: / under it. A general rule for all crawlers blocks the search ones too. The crawler roles in the table follow the documentation of OpenAI, Google and [Anthropic](https://support.claude.com/en/articles/8896518-does-anthropic-crawl-data-from-the-web-and-how-can-site-owners-block-the-crawler). How to set up robots.txt, sitemap.xml, canonical tags and 301s for the whole store is in the [ecommerce SEO checklist](https://cloudcart.com/en/blog/online-store-seo). ## How to write categories, products and blog posts that AI quotes An AI takes a chunk of 2 to 4 sentences from a page and retells it. Write so that every such chunk carries an answer on its own: - The answer first. The first 40 to 60 words of a category or article give the direct answer with a number: "Office coffee machine for up to 10 people: automatic, with a water tank over 1.5 l, from €350 to €600." - Headings as questions. An H2 "Which coffee machine for a small office" matches the way people ask an AI. - Numbers with a source. The GEO study found that statistics and quotations raise visibility the most. "We roast every Tuesday" is more useful than "always fresh". - A comparison table in categories with many similar products: model, main attribute, price, who it is for. - An FAQ with real questions from your emails, chat and reviews, with 2 to 4 sentences per answer. - A description written for a person on every product. Copied supplier text is the same in ten stores, and the AI has no reason to choose yours. An example of the first paragraph of a category in a coffee store: | Before | After | | "Browse our wide range of high-quality coffee machines at great prices." | "Coffee machines for home and office: 24 models from €89 to €1,150. For 1 or 2 people a manual or capsule machine is enough; for an office of up to 10, choose an automatic one with a tank over 1.5 l. Delivery with Econt and Speedy within 24 hours, returns within 14 days." | The first text answers nothing. The second answers three questions at once (how much, which one, when will it arrive) and every sentence can be quoted on its own. If you sell in Bulgarian and English, write each language version this way; an English buyer in Sofia asks the AI in English and gets English sources. The fastest source of questions is your store's own search box. Customers type into it exactly what they type into an AI: "something quiet for the bedroom", "gift for my dad under €50". Searches with no results are topics for categories, FAQs and articles. > AURA Aura Search: what customers ask, in their own words Aura Search searches by meaning and by words, understands Bulgarian typed in Latin letters and up to two typos in an eight-letter word, and shows which searches find nothing and which bring no order. That is your ready list of questions for GEO. Aura Studio writes descriptions and SEO texts for whole lists of products. See how Aura Search works → More on the AI tools in the admin and the storefront is in [AI for your online store: 12 tasks AI does for you](https://cloudcart.com/en/blog/ai-online-store), and the full list is on the [AURA](https://cloudcart.com/en/aura) page. ## Can a customer buy directly in ChatGPT? This is agentic commerce: the AI agent recommends the product and places the order for the customer. Where things stand: - ChatGPT shopping. For a purchase question, the chat shows product cards with price, image and reviews and links to the store. For your store to appear in these results, OAI-SearchBot must be able to read your pages. - Instant Checkout in ChatGPT and the open Agentic Commerce Protocol, built by OpenAI with Stripe, let the agent pay for the order inside the chat. It started with buyers and merchants in the US. - Universal Commerce Protocol (UCP) from Google is an open standard that lets agents in AI Mode and Gemini find a product, ask about terms and pay. Shopify is among its co-authors. Checkout in Google also started with US merchants. For a store selling in or from Bulgaria, the real path today is: the AI recommends the product, the customer clicks and buys in your store. So the last step decides. A customer coming from an AI answer has already chosen and expects payment to take seconds, by card, Apple Pay or Google Pay. Every extra field or redirect to another site loses orders. The in-chat purchase protocols rest on the same things too: machine-readable product data, clear terms and card payment. > CloudCart Pay 0% transaction fee through Dec 31, 2026 A customer who arrives from ChatGPT or AI Mode with a decision made pays by Visa, Mastercard, Apple Pay or Google Pay at checkout, and with Apple Pay and Google Pay straight from the product page. You need no separate bank contract. The order also carries 0% commission from the platform. How CloudCart Pay works → Every CloudCart plan comes with [0% commission on sales](https://cloudcart.com/en/offer/zero-commission-ecommerce), whether the order came from Google, an AI answer or an ad. ## How to tell whether AI recommends your store AI answers change, so you measure them with a protocol: the same list of questions, the same engines, every month. - Write 20 to 30 questions a customer asks before buying: "where to buy…", "which is better…", "… with next-day delivery". Take them from your store search, Search Console and your emails, in each language you sell in. - Run them in ChatGPT, Gemini and Google AI Mode in a new conversation, without being logged in to a profile that knows your store. - Record in a table: whether the store is mentioned, in what position, which of your pages is cited, which competitors come before you and with what pages. - Track AI visits in Google Analytics. Sources chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com. For a single "AI" row, create a custom channel group with a source condition: chatgpt\.com|perplexity|gemini\.google|copilot\.microsoft|claude\.ai . Setup is in the GA4 and Tag Manager guide . - Watch Search Console. Impressions and clicks from AI Mode and AI Overviews are in the overall report. You connect it from Apps → Google Search Console; the walkthrough is in the Search Console guide . - Repeat after a month and compare. Make one change a month so you know which one helped. ## GEO in CloudCart: what is built in and what is up to you Part of GEO is technical, and the platform does it for you. The rest is content and data that only you know. All settings below are in Marketing → SEO unless stated otherwise. | Task | What CloudCart does | What you do | | sitemap.xml | Generates and updates it; products are included with images | Submit the URL in Search Console and Bing Webmaster Tools; add a Sitemap: line to robots.txt, because the platform does not add it | | Canonical | Every page points to itself, without URL parameters | Keep the "Canonical tag" switch on | | Filtered and sorted pages | The "Deindex Filtered and Sorted pages" card sets noindex on filtered categories above a threshold you choose (0 to 5 parameters) | Switch it on so hundreds of versions of one category stay out of the index | | robots.txt | Adds the block on cart, checkout and wishlist itself | Allow the search crawlers from the table above | | llms.txt | Publishes a short default summary | Write your own text with categories and terms | | 301 redirects | Accepts them one by one or thousands from a CSV | Redirect every old URL when you change platform or structure | | Product structured data | Comes from the store theme | Test one product page in the Rich Results Test | | Google feed | The Google Shopping app sends the catalog to Merchant Center | Switch on auto-update for price and stock | | Descriptions and SEO texts | Aura Studio writes them for whole lists of products | Review them and add your experience: who the product is for, how it differs | | Measuring ChatGPT, Gemini and AI Mode | Connects Google Analytics and Search Console from Apps | Run the monthly protocol with 20 to 30 questions | ## Generative engine optimization checklist: 12 steps for an online store | # | Step | Done when | | 1 | The store is on a paid plan and indexed | Search Console shows indexed products and categories | | 2 | robots.txt allows Googlebot, Bingbot and OAI-SearchBot | No Disallow: / for them | | 3 | sitemap.xml is submitted to Google and Bing | Both tools read it without errors | | 4 | Product pages have structured data | The Rich Results Test shows "Product" | | 5 | The catalog is in Merchant Center | No products disapproved over price or stock | | 6 | Every product has price, stock, attributes and its own description | No product in your top 100 with copied supplier text | | 7 | Reviews show on the product | At least 5 reviews on your best sellers | | 8 | The "Delivery" and "Returns" pages state the numbers | Cost, time, threshold, couriers, days to return, as text | | 9 | Categories open with a direct answer and have an FAQ | The first 60 words answer the category question | | 10 | llms.txt is written by you | yourdomain.com/llms.txt shows your text | | 11 | Old URLs redirect with 301 | No old page returns 404 or shows old prices | | 12 | A monthly protocol with 20 to 30 questions is running | A table of results for ChatGPT, Gemini and AI Mode | > €1 for the first 3 months A store ready for Google and AI for €1/month sitemap.xml, canonical tags, noindex on filters, robots.txt and llms.txt are built into every plan. A new store pays €1/month for the first 3 months on any plan, so you can work through the checklist while the catalog gets indexed. Create a store for €1 → ## Mistakes that keep a store out of AI answers - Prices and stock only in an image or a script. The crawler sees an empty page. - The same supplier text on hundreds of products. The AI picks the source that says something new. - Mismatched facts. Delivery "within 2 days" on the site and "3 to 5 days" on the marketplace: the AI cannot tell which to trust and prefers a competitor. - AI crawlers blocked by habit. A rule blocking every crawler with "bot" in its name also blocks the search crawlers, along with the training ones. - Old pages deleted without a 301. The AI keeps citing the old URL and the customer lands on an error. - Expecting llms.txt to do everything. The file helps the assistants that read it. Indexing, data and mentions remain the foundation. - Text chopped into pieces "for AI". Short, artificial paragraphs with no link between them do not raise citations and make the page worse for the customer. Write for a person, with clear headings and a direct answer under each. - Bought "citation packages". Nobody can guarantee a mention in an AI answer. Measure it yourself with the protocol. ## Frequently asked questions **What is generative engine optimization?** Generative engine optimization (GEO) prepares a website so that ChatGPT, Gemini, Perplexity and Google AI Mode find it, understand it and cite it in their answers. For an online store that means indexed pages, complete product data, direct answers with numbers and mentions on other sites. GEO builds on SEO and extends it. **How do I get my store to show up in ChatGPT?** Allow OAI-SearchBot in robots.txt, submit your sitemap.xml to Bing Webmaster Tools and write pages with direct answers: delivery with prices, comparisons, buying guides. Add reviews and make sure brands, marketplaces and price comparison sites mention you with the same facts. Measure the result monthly with the same list of questions. **What is Google AI Mode and how does it affect online stores?** AI Mode is a tab in Google where Gemini answers with text, comparisons and product cards, and the user continues with follow-up questions. It works in Bulgarian and English. Products come from the Shopping Graph, so a store with its catalog in Merchant Center and accurate prices and stock has a chance to be shown. The traffic is reported in Search Console. **What is llms.txt and do I need it?** llms.txt is a text file at /llms.txt where you describe in Markdown what you sell and which pages an AI assistant should read. The standard is voluntary: Google says AI Overviews and AI Mode do not require it, but other assistants read it. In CloudCart you write it in Marketing → SEO → "Llms.txt file", up to 65,536 characters, and it updates within 5 minutes. **What is the difference between SEO and GEO?** SEO aims for a place in Google's results list and is measured by rankings and clicks. GEO aims for a place in the AI's answer and is measured by mentions, cited pages and visits from chatgpt.com or gemini.google.com. The foundation is shared: indexed pages with useful content. **Can customers buy directly in ChatGPT?** In-chat purchase exists: Instant Checkout in ChatGPT and Google's Universal Commerce Protocol for AI Mode and Gemini. Both started with merchants in the US. For a store selling in or from Bulgaria, the path today is a recommendation in the AI, a click and a fast card payment in your own store. **Should I block GPTBot and other AI crawlers?** It depends on what you block. GPTBot collects training data, and blocking it does not remove you from ChatGPT search. OAI-SearchBot indexes for search and must be allowed if you want to be recommended. Google-Extended controls Gemini training and does not affect AI Overviews or AI Mode. **How long does GEO take to show results?** The technical steps (robots.txt, sitemap, llms.txt, Merchant Center) take a day. New pages must be indexed first, so citations usually come after months. Measure every month with the same list of questions and change one thing at a time.