Nobody Actually Knows How to Win AI Search—But We're Sure Going to Try
Let’s be honest: we’re all figuring this out in real-time. Much like the early days of SEO—when we were telling clients about spiders crawling their sites (while secretly learning what that even meant ourselves)—AI Search is the new digital Wild West.
As Large Language Models (LLMs) like ChatGPT, Gemini, and Perplexity take center stage in how people find information, businesses everywhere are asking: “How do I show up?” While there’s no official rulebook (yet), there are clear strategies that increase your chances of being noticed—and even cited—by AI tools.
So buckle up, because in this post, we’ll walk you through what we know, what we’re testing, and what we recommend you start doing right now to increase your visibility in the age of AI-driven search.
How do LLMs Work?
Remember back in 2010 when we had to explain to our grandma (or basically anyone!) how SEO worked? We took a complex concept about spiders crawling web pages, indexation, and secret algorithms, and were able to explain Search Engine Optimization to the layperson. That is what we need to be able to do now. We need to understand how LLMs retrieve and synthesize content, as well as the process that runs in the background from the person querying to receiving an output. We also need to explain how and why the LLMs choose the output content, including the factors involved.
This is How LLMs Work
A LLM (Large Language Model) like ChatGPT, Gemini, and Perplexity is a giant library of information - books, websites, articles, podcasts, etc. It isn’t a brain; it doesn’t understand things the way a human brain does - it understands and remembers patterns and how they’re used together. When you query ChatGPT, for example, it uses the words you typed in and searches its “library of patterns” to provide what it believes to be the best output, and it does this word for word. The reason the output sounds so much like a human is that LLMs rely on probabilities, laying out each word in the output until they think the answer is complete.
Factors used by LLMs that affect the Output:
- Training Data - Everything in the LLM “library”
- Prompt - The specific query, question, request, etc.
- Context - All the words before the current one
- Probability Scores - How likely each word is to determine what it will say next
- Temperature - The creativity setting on the output determines how safe or wild the output is
- System Rules - Any special instructions given in the prompt, like length, tone, etc.
- Extra sources - Web search, database, APIs, documents, etc., to keep the output current
But how do I get the content I publish on my website to show up and be cited in LLMs?
This is the big, bold, underlined question for SEOs these days. Currently, there isn’t a foolproof solution, but online content creators should consider SEO best practices and new GEO tactics to ensure opportunities for citation in LLMs.
Ensure your website and web pages are public and crawlable.
- Allow reputable AI crawlers (like OpenAI, GoogleAI, Anthropic, etc.) via your robots.txt and avoid blocking these bots unless you specifically do not want them to train on your content. In addition, consider creating an llms.txt file - a new emerging standard designed to give website owners more transparent control over how Large Language Models (LLMs) can access and use their content. This extra layer of guidance helps ensure your preferences are communicated directly to AI systems.
Make sure your content is easy to understand.
- Just like our old SEO days when we were taught to structure our page content like an essay, using headers, sub headers, bulleted information, and easy digestible content, the same is true for today’s world of search. “Semantic chunking” breaks large chunks of content within the same topic into digestible bits of content, ensuring each part of the chunk makes sense on its own. Don’t forget to keep up with your descriptive metadata and structured data - AI models love this because it helps to quickly match questions.
Authoritative, Linkable, and Popular Content Wins
- Think E-E-A-T is so early 2020s? Think again, the more your content is referenced by other websites, the more likely it is to show up in both search engines and AI training datasets. LLMs tend to use these signals and trusted sources to influence their output.
Don’t rely on old, dusty content.
- LLM-connected search tools prioritize recent information, so publish fresh content frequently and revitalize older content on your website that you would want cited in an LLM query.
Trust matters.
- LLMs rely heavily on trust, like credible sources and consistent information, to influence their output. It doesn’t understand trust the way a human does; it has to look for signals in the data, such as a brand's visibility, the quality of its links (not just quantity), and online feedback about the brand. Having a strong online reputation makes it more likely that your business gets included in the sources the LLMs draw from and potentially cited.
AI Search is Evolving and So Should Your Strategy
Winning at AI Search isn’t about hacking the algorithm; it’s about showing up with high-quality, trustworthy content that’s easy for both humans and machines to understand. While we can’t guarantee LLMs will cite your brand tomorrow, we do know that forward-thinking, test-and-learn approaches are what separate leaders from laggards.
So, is there a magic bullet? Nope. But there’s power in experimentation—and we’re here for it. Try the tactics above, measure what matters, and keep pushing your content to evolve. Page One Web Solutions will be right there with you, figuring this out and sharing what we learn. Let’s get (ex)CITED!!! [GEO jokes, get it!?!]