Visibility doesn’t come from Reddit
First, it’s important to understand that generative AI systems like ChatGPT are based on training data collected at a specific point in time, known as the cut-off date. Anything published after this date is not initially part of the model’s knowledge base. For OpenAI’s latest version, GPT-5, this cut-off date is October 1, 2024, for example. The training data itself comes from various sources, primarily the internet, but also from books and studies.
When clever SEO agencies currently claim they can increase visibility in LLMs within a few weeks through Reddit posts, purchased PR links, or Wikipedia entries, that’s simply not technically possible. The reason lies in the cut-off date of the training data mentioned above. Content published after this date cannot immediately flow into the responses. So where does this assumption come from? A study by SEMrush and Statista from last year showed that many sources cited by AI systems come from Reddit, YouTube, or Wikipedia. While this may be true globally, particularly in the American market, it cannot be broadly applied to all industries and markets.
I recently analyzed for a provider of energy solutions in the photovoltaic and storage sector how visible their partner businesses are in AI systems. This involved local trade businesses and typical inquiries like “best installation company in Bamberg.” The result: not a single Reddit post in sight. Instead, platforms like MyHammer or even the providers’ own websites appeared as sources in ChatGPT and Perplexity.
What does this mean? The larger and more competitive a market is, the more training data there is. In niche markets, local markets, or with disruptive offerings, however, there is no large training dataset. Here, individually prepared information can often be the only relevant source and thus secure the coveted entry in the responses.
But what happens if you’re not part of the training data? In many cases, AI systems resort to current web searches, for example on topical issues or specific offers. This is where Retrieval-Augmented Generation (RAG) comes in: content is generated from the web in real-time. This way, you can also become visible retrospectively, though not as strongly as through the training data itself.
For startups and niche providers, it’s therefore worthwhile to carefully analyze the sources actually cited for relevant queries. Often these are industry listings or well-structured landing pages from competitors rather than Reddit or purchased PR links. Matching or even surpassing the competition is surprisingly often easier than you might think.
SEO tools no longer work without data
Online marketing was for decades primarily one thing: measurable. Clicks, keywords, and search volumes made the success of measures tangible and understandable. This exact logic is being shaken by the use of AI systems. While Google provides clearly measurable search queries and rankings, ChatGPT and Co. function fundamentally differently. Classic KPIs like search volume or search terms are simply no longer provided, and users no longer use standardized keywords but increasingly individual queries. The AI boom is thus changing the entire system of digital visibility.
With the introduction of AI Mode and AI Overviews, increasingly fewer Google searches land on websites but are answered directly. While various providers are named or listed as sources in Google’s AI responses and in ChatGPT, measurability is far more complex. The reason: prompts are much more diverse than classic keywords. They often consist of entire sentences or concrete questions. A clear assignment of which query leads to which result is thus practically impossible.
But the use case is also completely different. Google is often used for navigation. Users search specifically for a particular offering. A study by Harvard University shows that ChatGPT, on the other hand, is primarily used for “practical guidance,” i.e., as a decision aid. Whether when purchasing a product, formulating a business plan, or creating a training schedule. Writing and editing texts is also among the most common applications. LLMs are thus not classic search systems but assistants in production and decision-making.
Accordingly, it’s difficult to measure the concrete benefit of a mention. It’s more reminiscent of the market dynamics of TV advertising, whose effectiveness can generally be demonstrated at the supermarket checkout, but certainly does not allow attribution to individual sales. While there are approaches like prompt decoding, clickstream data, or AI visibility surveys, all of these approaches are more “qualified guessing” than reliable truths. Smaller providers in particular should not fall into expensive tool traps here.
Sebastian Schulze is an expert in artificial intelligence, data-driven marketing, and digital transformation and part of our Munich Startup Experts. He works as a keynote speaker and consultant and helps companies strategically deploy new technologies—particularly generative AI like ChatGPT—to increase visibility, marketing efficiency, and growth. With over a decade of experience in the digital economy, he advises both mid-sized companies and larger organizations. He focuses particularly on topics such as AI-powered marketing, new forms of online visibility, and how large language models are changing traditional search and marketing logic. (Photo: Sebastian Schulze)
Relevance no longer arises only on the website
Currently, numerous new terms like GEO, LLMO, or GAIO are flooding the web. They all share the same foundation: deliberately influencing discoverability in AI assistant systems like ChatGPT, Perplexity, or Claude, similar to how it’s possible with Google through SEO.
The problem: LLMs function fundamentally differently. They don’t have a fixed index where websites are assigned to specific keywords. Instead, responses are generated based on statistical probabilities. While Google must necessarily link to websites, this isn’t required in AI systems. Visibility therefore no longer arises primarily through your own website.
This can be seen, for example, in the fact that influencers like Sandra Hunke are often mentioned in ChatGPT and Co. for their field, even though they only have a moderately optimized website. What matters is their thematic relevance, built through other channels, particularly TikTok and Instagram. For AI systems, it’s secondary where this authority comes from, as long as it’s clearly recognizable. Accordingly, content from major media brands is also frequently used as sources.
Instead of investing your marketing budget in supposed GEO measures like Reddit or backlink purchases, you should instead actually generate real attention in your niche. Well-thought-out social media activities and credible PR often have a more sustainable effect here, including for citability in LLMs.
Sources, PR, and social media instead of expensive GEO activities
Especially startups that have neither unlimited time nor large budgets should not make the mistake of betting on supposedly quick solutions for visibility in ChatGPT, Claude, or Gemini. Most of these offerings are, to put it mildly, questionable.
It’s more sustainable to analyze the sources actually used for relevant queries and be present there strategically or build equivalent contributions. In rare cases, however, is it Reddit.
Instead, real attention and relevance in your own niche pay off in the long term, for example through credible PR, well-structured content, or thoughtful social media activities.
Because in the world of AI responses, the winner is not whoever optimizes best, but whoever is most relevant to their topic.