How GEO differs from SEO, how AI answer engines choose sources and products, and which measures pay off in Shopware 6: crawl access, answer paragraphs, JSON-LD, feeds, reviews, measurement, and costs.
GEO (generative engine optimization) makes sure ChatGPT, Gemini, Perplexity, and Google AI Mode use your store as a source and recommend your products. It doesn’t replace SEO but builds on it: answer engines pull their sources from search indexes, product feeds, and reviews.
For online stores, seven signals count: crawl access (permission for AI bots to fetch your pages), clear answer paragraphs on category and product pages, structured data, product feeds (Google Merchant Center, OpenAI), reviews, brand mentions on third-party sites, and up-to-date data.
In our experience, costs run €2,000 to €8,000 for the technical foundation and €800 to €3,000 a month for content and measurement. Almost every one of these measures also improves your visibility on Google.
Customers increasingly research products in ChatGPT, Gemini, and Google AI Mode, and a store that’s missing from these answers doesn’t show up in their research. This article is for store owners and ecommerce leaders who want to decide whether, and in which order, to invest in GEO, regardless of their ecommerce platform.
After reading it, you’ll be able to check whether AI systems are allowed to read your store, rank the seven signals by effort, and set budget and KPIs for the first year.
Before founding solution25, we built our own online store, and since 2017 we’ve been delivering Shopware, Shopify, and WooCommerce projects for merchants. Our own website allows GPTBot, OAI-SearchBot, ClaudeBot, and PerplexityBot via robots.txt, and in Search Console we now see queries phrased like prompts: whole questions instead of keywords.
Two separate articles go deeper into implementation in Shopware: GEO for Shopware with the technical settings, and AEO for Shopware with answer paragraphs and FAQs. All prices as of October 2026, excluding VAT.
According to a Bitkom survey from November 2025, half of German internet users at least sometimes use an AI chat instead of classic search, and among 16- to 29-year-olds it’s two thirds. According to OpenAI, ChatGPT had around 900 million weekly active users in February 2026; Google’s AI Overviews (the AI summaries above the search results) reach more than 2.5 billion people a month according to Google, and AI Mode (the conversational mode of Google Search) over a billion. AI Mode has been available in Germany since October 2025.
What matters for merchants is that this traffic buys. During the 2025 holiday season, Adobe Analytics measured 693% more visits from generative AI sources to US retail sites than the year before. These visitors converted 31% better than visitors from other channels.
At the same time, clicks from classic search are declining. In 2025, Pew Research found that Google users click on a classic result only 8% of the time when there’s an AI summary, compared to 15% without one. Ahrefs puts the drop in click-through rate for position 1 on queries with an AI Overview at 34.5%.
The table shows where the four systems get their answers from and which crawlers (the vendors’ fetching programs) let you control access.
| System | Where the answers come from | Shopping feature | Control |
|---|---|---|---|
| ChatGPT (OpenAI) | Own search index (OAI-SearchBot), third-party search data, partner content | Product cards since April 2025, merchant feeds via the Agentic Commerce Protocol, US only so far | robots.txt: OAI-SearchBot (search), GPTBot (training) |
| Google AI Overviews, AI Mode, Gemini | Google search index, Shopping Graph with more than 50 billion offers | Product panel in AI Mode, checkout via the Universal Commerce Protocol for US merchants | Googlebot, “Search generative AI” setting in Search Console, Google-Extended for Gemini training |
| Perplexity | Own index (PerplexityBot) | Product cards, free Merchant Program, in-app purchase only in the US | robots.txt: PerplexityBot |
| Claude (Anthropic) | Web search via Claude-SearchBot, page fetching via Claude-User | No documented merchant integration | robots.txt: Claude-SearchBot (search), ClaudeBot (training) |
Figure: The chart shows how much AI summaries reduce clicks: for stores, being mentioned in the answer counts, not just the position in the results list. Source: Pew Research Center, July 2025
The term comes from a study by Princeton University and IIT Delhi from November 2023 (“GEO: Generative Engine Optimization”). Across 10,000 queries, the researchers tested nine methods for preparing content for AI answers. Quotations, statistics, and source citations increased a source’s visibility by up to 40%, while keyword stuffing (piling up search terms) reduced it, by around 10% on Perplexity.
More important for merchants: sources that ranked only fifth in classic search gained 115% visibility through source citations. That makes GEO an opportunity for stores that sit behind the big marketplaces on Google.
In German-speaking countries, the term “GEO-Optimierung” (GEO optimization) has caught on, even though the O already stands for optimization. Alongside it, ChatGPT SEO, LLM SEO, and AEO (answer engine optimization) are used with the same goal; AEO usually refers to work on the individual answer and its data, which our article AEO for Shopware covers in depth.
| Criterion | SEO | GEO |
|---|---|---|
| Goal | Position in the results list, click | Mention, citation, or product recommendation in the answer |
| Unit | Page and keyword | Paragraph, brand, product, record in the feed |
| Success metric | Rankings, clicks, impressions | Mentions, citations, share of answers, referrals from AI tools |
| Key signals | Relevance, links, user signals, technical setup | Crawlability, clear evidence-based statements, brand mentions, product feeds, reviews |
| Measurement | Search Console, rank tracker | Search Console (impressions only), prompt tests, specialized tools |
In its guide to generative AI features (as of July 2026), Google writes that the familiar SEO fundamentals still apply and that there’s no special markup for AI Overviews and AI Mode. A page has to be indexed, allowed to be shown with a text excerpt (snippet), and enabled for generative AI features in Search Console. Our guide to Shopware SEO describes the foundation; GEO starts with the question of whether a paragraph on your page works as an answer.
For the cosmetics brand Banana Beauty, we rebuilt the frontend of the Shopware store and reworked metadata, structured data, and the page hierarchy in the process. According to our case study, conversions rose by 28%, the SEO measures contributed to 35% more organic traffic, and load times dropped by 40%.
AI visibility wasn’t measured in this project. But the case shows the part of GEO that reliably pays off: structured data and a clean page structure are the SEO fundamentals that, according to its guide, Google also requires for AI Overviews and AI Mode.
Answer engines don’t answer current questions from their training data, but fetch documents at runtime (retrieval). Google describes the process for AI Mode like this: the model breaks the question down into several parallel sub-queries (“query fan-out”), fetches current pages from the search index for them, and assembles the answer with links. According to OpenAI, ChatGPT uses third-party search providers, partner content, and its own crawler, OAI-SearchBot; Perplexity works with its own index.
It follows that anything blocked, not indexed, or only loaded via JavaScript can’t be cited. And anything that can be fetched has to deliver a clear statement in a few sentences, because the model takes over passages, not entire pages.
All of these systems link to their sources, but the links are rarely clicked: according to Pew Research, only 1% of users click a link within Google’s AI summary. The value lies first in the mention. If you appear in the answer as a recommended provider, people will find you later through a branded search.
For product questions, the systems rely on structured product data. Since April 2025, ChatGPT has shown product cards with image, price, and reviews; according to OpenAI, the results are organic, unpaid, and sorted by relevance, and when there are several sellers, availability, price, and quality count. Merchants supply their data as a product feed via the Agentic Commerce Protocol (ACP).
Since March 2026, customers complete the purchase in the merchant’s store. According to OpenAI, the merchant program is limited to the US so far, has no fees, and is supposed to get a self-service portal later in 2026.
Google feeds AI Mode from the Shopping Graph, which according to Google contains more than 50 billion offers and updates 2 billion of them every hour; the data comes from Merchant Center. Via the Universal Commerce Protocol (UCP) introduced in January 2026, agents complete purchases in AI Mode and the Gemini app, initially with US merchants.
Perplexity has offered a free Merchant Program since November 2024. What this means for checkout is covered in our articles on agentic commerce and AI agents in ecommerce.
From the vendors’ documentation and the studies, seven signals can be derived that determine a store’s visibility in AI answers.
The vendors distinguish between crawlers for training, crawlers for search, and fetches on user request. If you block all of them across the board, you disappear from the answers; if you only block training, you stay visible.
| Bot | Operator | Purpose | Effect of blocking |
|---|---|---|---|
| OAI-SearchBot | OpenAI | Index for ChatGPT search and citations | According to OpenAI, the page no longer appears in ChatGPT search answers, at most as a navigational link |
| GPTBot | OpenAI | Model training | No training on your content, search is unaffected |
| PerplexityBot | Perplexity | Index for Perplexity answers, no training | Store doesn’t appear in Perplexity, change takes effect after up to 24 hours |
| Claude-SearchBot | Anthropic | Web search in Claude | Visibility in Claude answers drops |
| ClaudeBot | Anthropic | Training | No training, search is unaffected |
| Google-Extended | Gemini training and grounding in the Gemini app, no separate user agent | No effect on Google Search, AI Overviews, and AI Mode |
The information comes from the documentation of OpenAI, Perplexity, Anthropic, and Google. On top of that, there are ChatGPT-User and Perplexity-User, which fetch pages on user request and don’t necessarily respect robots.txt. For AI Overviews and AI Mode, what counts at Google is the regular Googlebot plus the “Search generative AI” setting in Search Console, which has been available worldwide since August 31, 2026.
Since version 6.7.1, Shopware generates robots.txt itself and lets you maintain it per domain in the basic settings. The default rules block parameter URLs, not AI bots. Still, three things regularly go wrong:
Whether you allow the training crawlers GPTBot and ClaudeBot is a business decision. On solution25.com, we allow all of the bots listed with a dedicated group:
User-agent: OAI-SearchBot
User-agent: GPTBot
User-agent: PerplexityBot
User-agent: Claude-SearchBot
User-agent: ClaudeBot
User-agent: Google-Extended
Allow: /
The proof is in the server logs, your web server’s access records: if OAI-SearchBot shows up there, ChatGPT is reading your store. If the entry is missing for weeks, there’s a block that isn’t visible in robots.txt.
Our technical guide GEO for Shopware describes all the settings in Shopware, from per-domain rules to feeds to JSON-LD.
| Priority | Measure | Effort | Affects |
|---|---|---|---|
| 1 | Check crawl access in robots.txt, CDN, and firewall, analyze server logs | 1 to 2 hours | All systems |
| 2 | Complete product data (GTIN, brand, availability, shipping, returns), enable JSON-LD | 1 to 3 days | Google, all feeds |
| 3 | Set up or clean up the Merchant Center feed | 1 to 2 days | AI Overviews, AI Mode, Gemini |
| 4 | Answer paragraphs for the 20 most important categories and 50 most important products | 2 to 4 weeks | All systems |
| 5 | FAQ blocks on category and service pages | 1 to 2 weeks | All systems |
| 6 | Collect reviews, display them on product pages and in the feed | Ongoing | ChatGPT, Google, Perplexity |
| 7 | Prepare the OpenAI product feed via the Agentic Commerce sales channel | 0.5 to 1 day | ChatGPT (US, Europe announced) |
| 8 | Brand mentions in trade media, directories, and retailer lists | Ongoing | All systems |
| 9 | Create llms.txt | 1 hour | Not proven |
A category text that begins with “Welcome to our rain jacket category” gives an answer engine nothing. A text that begins with “Rain jackets with a waterproof rating of 10,000 mm or more withstand a full day of steady rain, and 20,000 mm or more even with a backpack” answers the question users are asking. The same applies to product pages: dimensions, materials, compatibility, and the product’s limitations in the first three sentences.
FAQ blocks with five to eight genuine customer questions per category deliver the same paragraph format, as long as the text is in the HTML and isn’t loaded only on click. Our article AEO for Shopware shows how to create answer paragraphs, FAQ blocks, and properties in Shopware.
Since May 7, 2026, FAQPage markup no longer produces rich results; according to its changelog, Google has discontinued the feature. The FAQ block itself remains valuable; the markup is optional.
Since 6.7.9 (April 2026), Shopware also outputs structured data as JSON-LD, via a feature flag (a switch for new features) that’s disabled by default; the older Microdata format will be dropped with 6.8. After switching, check price, availability, and GTIN in Google’s Rich Results Test; details are in our article on rich snippets in Shopware.
Merchant Center is the shortest route into Google’s Shopping Graph and therefore into AI Mode, and for UCP checkout Google requires matching data in feed and website. Our guide to Google Shopping and Merchant Center with Shopware 6 shows the setup in Shopware.
For ChatGPT, the OpenAI specification requires nine mandatory fields, including brand, merchant name, availability, and price. In Shopware, this feed is created via the Agentic Commerce channel, whose core functionality has been moving into the free Agentic Commerce (Beta) extension since 6.7.12; as a beta, it belongs on a test instance.
Trade media in your industry, manufacturers’ retailer lists, industry directories, Trusted Shops, and Google reviews count for more in AI answers than classic link metrics. Company name, product range, and location have to be identical everywhere; Google explicitly warns against artificially generated mentions. Our article on Google reviews in Shopware shows how to integrate reviews.
The cheapest measurement is a fixed prompt set: 20 to 30 questions your customers would ask, asked once a month in ChatGPT, Gemini, Perplexity, and AI Mode, without a logged-in account and without saved memory. For each prompt, note whether your store is mentioned, which URL is cited, and which competitors show up. The results fluctuate from day to day, so the trend over months counts, not the individual test.
Since August 31, 2026, Search Console has shown a “Search generative AI” report worldwide with impressions in AI Overviews, AI Mode, and Discover, broken down by page, country, and device. The report doesn’t include clicks or queries yet; Google has announced more metrics. The same area contains the setting that lets you exclude your website from AI features; according to Google Help, “Include” is the default.
You can see referrals from AI tools in Google Analytics 4 under the referral sources chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com, and claude.ai; create a dedicated channel group for them, as described in our GA4 guide for Shopware 6. Purchases that an agent completes via UCP or ACP don’t generate page views; you can only identify these orders server-side or via the affiliate ID of the Agentic Commerce channel.
| Tool | Price (as of October 2026) | Scope |
|---|---|---|
| Ahrefs Brand Radar | Lite $129 a month with 5 tracked prompts, Standard $249 with 10, Advanced $449 with 20; Brand Radar AI add-on from $199 | Mentions and citations in AI Overviews, ChatGPT, Perplexity, Gemini, and Copilot, daily |
| Otterly.ai | Lite $29 with 15 prompts, Standard $189 with 100, Premium $489 with 400; AI Mode, Gemini, and Claude as add-ons | Prompt tracking in ChatGPT, AI Overviews, Perplexity, and Copilot |
| Peec AI | Starter 50 prompts, Pro 150, Advanced 350; prices on request, trial available | Three models per plan from ChatGPT, AI Mode, AI Overviews, Copilot, and Gemini |
| Semrush AI Toolkit | Add-on to the Semrush subscription, around $99 a month per domain according to third-party sources; check with Semrush | Visibility, share of answers, and sentiment in ChatGPT, Gemini, Perplexity, and AI Overviews |
| Manual prompt set | €0, about 2 hours a month | 20 to 30 of your own questions in four systems |
The prices come from the vendors’ pricing pages (Ahrefs, Otterly, Peec AI), net in each case. According to Google, no third party has access to its internal ranking or AI systems; the tools run prompts and count answers. For stores with less than €1 million in revenue, the manual prompt set plus Search Console is usually enough.
| KPI | Tool | Benchmark |
|---|---|---|
| Fetches by AI crawlers | Server logs | OAI-SearchBot and PerplexityBot every week (from experience) |
| Mention rate | Prompt set with 20 to 30 questions | upward trend over three months compared to the documented baseline |
| Cited pages | Prompt set or monitoring tool | Category and product pages, not just the homepage |
| Impressions in AI features | Search Console, generative AI report | upward trend per page group |
| AI referrals and revenue | GA4 with a dedicated channel group | Conversion rate at least at the store average (from experience) |
Collect the numbers monthly. You should adjust course if crawler fetches stop, the mention rate stagnates after three months, or competitors get cited with pages that answer your customers’ questions better.
Then there are the limits of the approach: no vendor discloses how it weights sources, and answers fluctuate from day to day. GEO increases the likelihood of a mention, but it can’t guarantee one.
The following ranges are based on experience from our projects, not list prices. They assume your SEO foundation is in place; what a Shopware project costs overall is covered in our article on Shopware pricing.
| Item | Range | Included |
|---|---|---|
| GEO audit | €1,500 to €4,000, one-time | Crawl access, server logs, product data, feeds, prompt baseline with 30 questions, action list |
| Technical foundation | €2,000 to €8,000, one-time | robots.txt and CDN, JSON-LD, Merchant Center feed, Agentic Commerce channel, tracking in GA4 |
| Answer paragraphs and FAQ | €150 to €400 per category or product page | Customer question research, copy, properties, comparison table |
| Ongoing support | €800 to €3,000 a month | Monthly prompt tests, Search Console analysis, content updates, PR and directories |
| Tools | €0 to €500 a month | Manual prompt set up to Ahrefs or Otterly |
A small store with an audit, technical foundation, and 20 revised category pages comes to €6,000 to €15,000 in the first year, and a mid-sized store with ongoing support to €20,000 to €45,000. Most of this also pays off in Google rankings and product data quality.
Not every store needs to invest in GEO in 2026. Three profiles help you assess where you stand:
The prerequisite for GEO is a store without indexing problems, with complete product data and some initial reviews. If that’s missing, this groundwork comes first, because it pays off on both fronts.
Our take: For most stores, GEO pays off first as basic hygiene: crawl access, product data, a Merchant Center feed, and a prompt set cost little and work on Google and AI answers at the same time. We only recommend ongoing budgets for content and PR once the prompt baseline shows that competitors get mentioned and you don’t.
GEO stands for generative engine optimization: measures that get a website mentioned, cited, and linked in the answers of ChatGPT, Gemini, Perplexity, or Google AI Mode. The term comes from a 2023 Princeton University study. For online stores, it’s mainly about crawl access, answer paragraphs, product feeds, and reviews.
SEO aims for a position in the results list and the click, GEO for a mention in a generated answer. SEO evaluates pages and keywords, GEO paragraphs, brands, and data records. The technical side overlaps heavily: without indexing and clean product data, neither works.
First come crawl access, product data, and feeds, then answer paragraphs, reviews, and mentions in trade media. You measure with a fixed prompt set and the Search Console report; the eight steps above show the order in detail.
Yes. Google pulls the sources for AI Overviews and AI Mode from the regular search index and names the SEO fundamentals as a prerequisite. SEO remains your ticket in, and GEO decides whether you show up in the answer.
In our experience, a small store comes to €6,000 to €15,000 in the first year, and a mid-sized store with ongoing support to €20,000 to €45,000, excluding VAT in each case. The cost table above shows the individual items.
No: llms.txt is a Markdown file in the root directory that’s meant to list a website’s most important content for AI systems. According to its guide, Google doesn’t read such files, and SE Ranking found no link to citations in AI answers across around 300,000 domains. The file takes an hour and does no harm, but it isn’t a priority.
Crawl permissions and feed changes take effect within days; OpenAI and Perplexity cite up to 24 hours for robots.txt changes. In our experience, new answer paragraphs show up in answers after two to eight weeks, once the pages have been recrawled. Brand mentions on third-party sites take months, but they have the most lasting effect.
Want to know whether OAI-SearchBot and PerplexityBot are reading your store, whether your product data is good enough for feeds, and how often ChatGPT mentions your competitors? Get in touch: we’ll run the GEO audit with a prompt baseline and name the three measures with the biggest impact.
Already running a store? Our free store audit checks indexing, crawl access, structured data, and product data quality, the foundation for any visibility, on Google and in AI answers.
Learn more about how we work, our certifications, and our locations on the solution25 Shopware agency page.
Besar Bilalli is co-founder and Co-CEO of solution25, the Shopware agency based in Bremen. Before founding the agency in 2017, he and Besnik Bilalli built their own online store without any startup capital and grew it to more than 500 orders a day. Since then, he has guided merchants through Shopware projects, from platform selection to day-to-day operations. He’s happy to answer questions about this article personally: get in touch.
Responsible for content and updates: Besar Bilalli. Last updated: October 2026. Next content review: January 2027.
