SEO 2026: How to Get Your Products Into ChatGPT and Gemini

The world of search is changing fundamentally. While classic SEO relied on keywords and rankings in Google, AI systems like ChatGPT, Gemini, or Perplexity increasingly act as gatekeepers between customers and companies. They compare products, recommend providers, and give purchase recommendations – without the user ever seeing a classic search results page.

The driving force behind it? Convenience. Users no longer want to gather answers themselves. They want the solution served on a silver platter – personalized, contextual, instant. And that's exactly what AI systems deliver.

The decisive question is therefore no longer "How do I rank number 1?" but: "How do I become the answer that the AI gives?"

This guide shows you concretely and step by step how to achieve this.

First things first: No, SEO is not dead.
This debate persists stubbornly – but remains misleading. What changes is not whether people search, but how they search. Whether Google, ChatGPT, or Gemini – as long as people search, companies need search optimization. The basic mechanics stay the same, only the playing field shifts.

The Foundation: 80% Basic SEO, 20% AI-Specific

Before you dive into AI optimization, a reality check: Around 80% of your success in AI search still depends on solid basic SEO – onpage and offpage fundamentals, technical cleanliness, good content. AI systems read your website, and if the fundamentals aren't right, you won't show up in AI answers either. Google's Danny Sullivan put it in a nutshell: "Good SEO is good GEO" (Generative Engine Optimization).

This is also logical: If an AI is to recommend someone, it needs trust. And trust is built – online as offline – on two pillars:

Both pillars must be strong. According to a Yext study with 6.8 million AI citations, 86% of all AI citations come from sources that brands can control themselves – above all their own website (44%) and industry directories (42%). Your own content is therefore the most important lever.

The remaining 20% are AI-specific factors – and these are exactly what the following is about.

1. Prompt Research: The New Keyword Research

FAQ: What exactly is prompt research – and why is keyword research no longer enough?

Classic keyword research analyzes which individual words or short phrases users enter into Google: "buy running shoes", "coffee grinder review". Prompt research, on the other hand, analyzes which complete need situations users describe in AI tools.

The difference is decisive: No real person enters just "coffee grinder" into ChatGPT. Instead, they write something like: "I like to drink filter coffee in the morning and want to grind fresh. My budget is 150 euros. Which grinder with a conical burr mill would you recommend?"

Users put their problems, concerns, and wishes into a concrete context. They expect a personalized answer – not a list of links. Analyses by OpenAI show that product-related queries break down into clear patterns: users compare products, seek advice based on use case, or ask where they can buy.

Key takeaway: A prompt is always an expressed need situation. A keyword is the basis (it signals interest), but not the goal. Prompt research is fundamentally different from keyword research, because people don't enter keywords into chatbots but place problems and wishes into a context.

FAQ: How do I carry out prompt research in concrete terms?

Step 1: Answer the core question – "What are we the best solution for?"

Before you formulate a single prompt, you must answer this question. Classic SEO made this comfortable here: people oriented themselves toward search volume figures from tools instead of asking themselves which concrete pain point their own product solves. In the AI world, you have to do fundamental marketing again:

• In which concrete situations does someone need your product?
• Which problem does it solve better than the alternatives?
• For which target group and which use case are you the first choice?

Step 2: Collect the need situations of your target group

• Talk to your sales and customer service: Which questions come up again and again?
• Search review portals and forums: What is praised, what is criticized?
• Read Reddit threads, YouTube comments, and reviews in your niche.
• Use tools like AnswerThePublic or AlsoAsked to identify question patterns.

Step 3: Create prompts according to five categories

Create 5 to 10 prompts per product category, oriented toward these five types:

Categorize your prompts cleanly: brand vs. non-brand, by product category, and by funnel stage. This way you can later evaluate in a granular manner which areas of your shop are growing in AI search.

Step 4: Test and document prompts in AI tools

Enter each prompt into ChatGPT, Gemini, and Perplexity and document:

1. Are you mentioned as a provider? In what context and with what narrative?
2. Which competitors are recommended – and with what arguments?
3. Which sources does the AI cite? (Directly visible in Perplexity; for ChatGPT, ask: "What sources is your answer based on?")
4. Is the sentiment positive, neutral, or negative?

Practical tip: Create a tracking sheet
Create a table with the columns: Prompt | Category | ChatGPT Result | Gemini Result | Perplexity Result | Own Brand Mentioned (Yes/No) | Sources Cited | Sentiment | Next Step. Repeat the analysis monthly.

FAQ: What do I do with the results of the prompt research?

The results show you three types of gaps:

• Gaps on your website: If the AI doesn't mention your brand, content that addresses exactly this need situation is probably missing. Create a dedicated subpage or an advisory article for it.
• Gaps on third-party sites: If the AI cites sources on which you are not represented, you need to become present there (see the Search Reputation section).
• Argumentation gaps: If the AI recommends competitors because they communicate clear advantages, check whether you state similar facts just as clearly on your website.

Handle top-funnel selectively
General advisory questions like "What is the difference between Robusta and Arabica?" are answered by the AI directly itself – without linking to your website. Current analyses confirm: Top-funnel content (what-is, how-to, guides) records massive traffic declines. Prioritize prompts where the AI must compare providers or give purchase-related recommendations – only there does revenue flow.

FAQ: Is there a central AI index in which I can register my domain?

No. There is no central index in which you register your domain and then see for which prompts you are visible. Visibility is always prompt-based, not domain-based. This is exactly why you work with concrete prompt sets that map your most important need situations.

FAQ: Which KPIs are really relevant for AI visibility?

Don't be dazzled by made-up metrics. Meaningful KPIs:

• Mention Rate: In what percentage of relevant prompts do you appear?
• Sentiment: How are you mentioned (positive, neutral, negative)?
• Referral Traffic: How is traffic from AI systems developing? (Measurable via GA4 referral data – filter by referrers like chat.openai.com, gemini.google.com, perplexity.ai)

Metrics like "position in the AI answer" are of little significance – AI answers are never identical and have no fixed ranking order.

Note the "dark search" effect
Many users get advice in ChatGPT but then enter your brand name directly into the browser. This indirect traffic cannot be precisely attributed. Therefore, also observe your direct traffic and brand search development as complementary indicators.

2. Optimizing Content for AI Visibility

Principle 1: Facts instead of platitudes – AI loves clarity

For AI citations and recommendations, clear, verifiable facts work significantly better than metaphors or purely emotional wording. Especially in e-commerce, many shops try to let the emotionality of their brand resonate. That has its justification for human readers – but AI systems preferentially cite concrete, verifiable statements.

A Surfer SEO analysis of 46 million AI citations confirms: Pages cited in AI Overviews contain on average 62% more verifiable facts than non-cited pages.

Write your most important facts explicitly into the body text: year of founding, customer number, certifications, awards, concrete results. In one case example, such a clear formulation was already picked up in AI answers the very next day. This shows: When facts are clearly stated in the text, the chance that AI systems adopt them increases.

Principle 2: Incorporate evidence and sources

Content that refers to external studies or specialist sources is rated as more credible by AI systems. The Princeton GEO study (published at KDD 2024) showed in tests with 10,000 search queries that adding statistics and verifiable data points can increase visibility in generative answers by up to 22%.

• Link to at least one relevant external source per page (study, industry report, official statistics).
• Use formulations like "According to a study by [institution] ...".
• Avoid unsubstantiated superlatives ("the best", "unique") – they lower credibility.
• Whether follow or nofollow is usually not the main lever – more important are relevance, the authority of the source, and the context of the mention.

Principle 3: FAQ sections on product pages

FAQ areas work particularly well because they precisely map the question-and-answer format in which AI systems think. The prompts from your prompt research are at the same time the perfect FAQ questions for your pages.

Shopware tip: Implement FAQ section technically
If you use Shopware, you can embed FAQ areas directly on category and product pages – e.g. with a FAQ plugin like FAQ Professional. Important here: The FAQ content must be rendered as static HTML (not loaded only via JavaScript) so that AI crawlers can read it. Additionally, pay attention to correct FAQ schema markup (FAQPage structured data), so that the content is optimally prepared for Google and AI Overviews as well.

3. Technical Readability: The JavaScript Problem

FAQ: Why doesn't the AI see my content, even though it's on the website?

Many AI crawlers cannot – unlike the Googlebot – render JavaScript. An analysis by Vercel and MERJ, which evaluated over 500 million GPTBot requests, found no evidence of JavaScript execution at all by the major AI crawlers (GPTBot, ClaudeBot, PerplexityBot). SearchViu quantifies that 69% of all AI crawlers cannot execute JavaScript. The only notable exception: Google's Gemini uses the Googlebot infrastructure and can therefore render JavaScript.

This means: Everything that only arises in the browser through JavaScript is simply not captured by most AI systems.

From practice: From impressive figures to "0"
A shop had an animated counter on the homepage: When loading, the numbers counted up – "Over 35 years of experience", "Over 800 customers served". Looked impressive to visitors. But since the counter ran via JavaScript, the AI saw: "Over 0 years of experience. 0 customers served." After switching to static presentation in the HTML, mentions in AI answers almost tripled in this case.

Typical problem cases that AI crawlers don't see:

• Animated counters and number widgets
• Product review widgets (stars, reviews) – affects many online shops
• Dynamically loaded testimonials and customer quotes
• Accordion and tab content that only becomes visible on click
• Single-page applications (SPAs) whose content is rendered entirely client-side

FAQ: How do I check whether my content is visible to AI crawlers?

The simple JavaScript check (2 minutes)

1. Open your website in Chrome.
2. Open the developer tools (F12 or right-click → "Inspect").
3. Press Ctrl+Shift+P (Mac: Cmd+Shift+P) and type "Disable JavaScript".
4. Reload the page.
5. Check: Is all important information still visible? Customer numbers, reviews, trust signals, core content?

Everything that disappears after disabling JavaScript is invisible to most AI crawlers.

Solution: Output trust signals additionally as static text
You don't have to remove all widgets. The simplest way: Write the most important trust signals additionally as static text in the HTML code. So e.g. "Over 1,300 customers trust us – 4.8 out of 5 stars on Google" directly in the body text of your category page. The widget can remain – but the AI needs the information as readable text. For JavaScript frameworks: Rely on server-side rendering (SSR) or static site generation (SSG).

Crawling vs. Training: An important difference

Not all AI bots are the same. There is a difference between crawlers that read your page for search queries in real time (e.g. OAI-SearchBot, ChatGPT-User), and crawlers that collect content for model training (e.g. GPTBot). Via the robots.txt you can control these separately:

• GPTBot (OpenAI training) – Blocking prevents future training use, not real-time search.
• OAI-SearchBot (OpenAI search) – Blocking prevents ChatGPT from reading your page live for search queries.
• Google-Extended (Google training/Gemini) – Blocking prevents use by Gemini without affecting Google search.

Shopware tip: Optimize sitemap and crawlability
Make sure your Shopware sitemap is up to date and complete – it is the basis for both Google and AI crawlers to find all relevant pages. A sitemap plugin like Google Sitemap Professional can help control the sitemap granularly and keep it automatically up to date.

4. Search Reputation: What Others Say About You Decides

FAQ: What is search reputation and why is it so important?

AI models form their opinion about your brand not only based on your website. When someone asks "Which product for XY can you recommend?", the AI researches independently on the internet. An analysis by Surfer SEO (46 million citations) shows: YouTube (~23%), Wikipedia (~18%), and Google pages (~16%) dominate the sources. Depending on the industry, comparison portals, specialist media, Reddit, and retailer reviews come into play.

This means: The narrative about your brand is largely shaped outside your own website.

FAQ: How do I find out which sources the AI uses for my industry?

Step by step: Source analysis

1. Enter 5 industry-relevant prompts into Perplexity – there you see the cited sources directly.
2. Note all cited domains and sort by frequency.
3. Repeat this with ChatGPT: After each answer, ask "What sources is this recommendation based on?"
4. Create a priority list: On which platforms are you already present? Where are you missing?
5. Develop a concrete strategy for the top 5 sources (product listing, guest article, active response to reviews).

Scalpel instead of watering can
The generic tip "just be present everywhere" sounds good but is not practical. Identify via your prompt research the concrete sources that the AI actually uses for your industry, and concentrate your limited resources on them.

FAQ: Can I hand my content directly to the AI?

Directly "optimizing into training data" is not possible. What you can influence is what the AI finds when it researches independently on the internet. Via your robots.txt you can control whether your content may be used for training and/or real-time search (see the section above).

5. Structured Data: The Machine-Readable Layer

In addition to readable text, AI systems benefit from structured data (Schema.org markup). This helps the AI contextually classify your content – and increases the likelihood of being recognized as a citable source.

Important schema types for online shops:

• Product – product name, price, availability, reviews
• FAQPage – questions and answers on category/product pages
• Organization – company name, year of founding, location, contact
• AggregateRating – overall ratings and number of reviews

Shopware tip: Google Shopping & feeds
Structured product data is important not only for AI visibility, but also for Google Shopping. A Google Shopping plugin like Google Shopping Professional ensures that your product data is handed over to Google correctly and completely – which at the same time improves the data quality that AI systems access.

6. Outlook: What Happens When the AI Becomes the Customer?

Currently, the AI is an intermediate layer – an advisor that gives the user recommendations. But development continues: In the area of agentic commerce, the AI could in the future not only recommend but buy directly – on behalf of the user.

In this scenario, the AI itself becomes the customer. And it will use the same methodology: Can I trust this provider? What is on the page? What do others say? Those who lay the foundations today are also prepared for this scenario.

Quick Win Checklist: Measures You Can Implement Immediately

Conclusion

AI answer systems will displace classic search in many use cases – the question is less whether, but in which intents and how quickly. The direction is clear: First users had to research themselves, then came featured snippets, then AI Overviews, and many search interfaces are visibly moving toward AI-supported answers.

This does not mean that SEO is dying. It means that SEO is transforming – from keyword optimization to context and need optimization. Those who start today optimizing their content for concrete need situations, delivering clear facts instead of marketing speak, and actively managing their reputation on third-party platforms secure a decisive head start.

Because in the end: The AI doesn't recommend the one who shouts the loudest – but the one it trusts the most.

Sources and further studies:

• Vercel/MERJ: The Rise of the AI Crawler – JavaScript rendering analysis of 500+ million bot requests
• SearchViu: AI Crawlers & JavaScript Rendering – 69% of AI crawlers do not execute JS
• Princeton GEO study (KDD 2024) – Generative Engine Optimization, analysis of 10,000 search queries
• Surfer SEO: AI Citation Report 2025 – 36 million AI Overviews, 46 million citations analyzed
• Yext: 86% of AI citations from brand-controlled sources – 6.8 million citations analyzed
• Position Digital: 90+ AI SEO Statistics 2025