How to Optimise a Website for LLMs: Building SEO for AI
Searching for information through LLMs (Large Language Models) has become so widespread that some specialists are predicting the imminent end of traditional search engines. Although a StatsCounter report indicates that Google still controls a solid 90% of the market¹, the sharp rise in AI-powered searches deserves attention – so let us explore how to optimise websites for LLMs, so that they appear in tools such as ChatGPT.
What does LLM SEO involve?
LLM SEO involves publishing information that can subsequently be used by large language models.
Usually, chatbots that use LLMs to generate answers do not carry out research, as they have access to the large amount of information on which their models were trained. This primarily includes widely known facts, definitions and principles.
However, research is activated when:
- the query requires up-to-date information – relating to the weather, sporting events or statistics. For example: “How many births are there in the UK?”;
- the data is niche or detailed – relating to a local market, a specific person or a company, for example: “Where can I find the best kebab in London?”.
Actively searching for information means that a language model is not limited to its training data – it is therefore not a closed system, unlike the initial version of ChatGPT. This opens up completely new possibilities not only for readers, but also for companies that provide data.
Why optimise a website for LLMs?
Supporters cite many arguments for using AI to search for information.
One undeniable benefit is the accuracy and speed of obtaining data – we do not have to wade through dozens of results, accept cookie banners, close adverts and search for the section that interests us.
Instead, we receive a direct answer that matches our query. That is why an interesting metaphor has been circulating online: Google is a monumental library, while ChatGPT is a well-read librarian.
There is no doubt that AI represents an entirely new technology for obtaining data, offering strategic advantages over traditional search engines. This does not mean that SEO is ending; it is simply changing its form. ChatGPT cites its data sources (from GPT-4 onwards), indicating the company name and linking to the source – so we still benefit from the traditional advantages of generating traffic and building brand awareness.
It is therefore worth moving with the times and adapting the information on your website to this new way of searching, so that you do not lose visibility online.
How do you optimise a website for large language models?
Let us use the standard division between activities carried out within the website (on-site) and those carried out beyond it (off-site).
On-site SEO
The basic principle of LLM SEO is the same as with Google search: reliable and valuable information is essential, as it can be shown to audiences.
However, the nature of the content changes. General, widely known information is very likely to already be included in the model’s knowledge base, meaning it will be cited without external sources. We need to publish unique, detailed data that our competitors do not have – for example, detailed descriptions, product reviews, the results of our own surveys or research, and so on.
This undoubtedly raises the bar in the content marketing industry. In traditional SEO, the principle of using popular keywords still applies – in practice, this means describing topics in the same way as many predecessors, only better. LLMs, on the other hand, require data that can fill gaps when responding to a precise query, creating much higher demands in terms of content uniqueness.
To combine LLM SEO with building visibility in Google, we should remember the four areas that consistently characterise good content.
Expertise
This comes from many years of activity in a particular industry. Experience allows us to approach a topic differently, formulate individual insights, predict trends and address more practical aspects. In-depth knowledge brings a particular kind of ease, insight and charisma that separates expert content from AI.
Example article: “What does a content writer do?” (as specialists in content creation, we described the content writer profession in considerably more depth than others).
Uniqueness
Information that is not widely available and represents something new or an alternative approach to a given topic, such as product reviews or tests, the results of a survey, the outcome of an experiment, trend analyses, market research or the publication of statistics.
Example article: “The most popular writing services” (as the creators of the contentwriter.co platform, we published our own data on the popularity of writing services).
Functionality
Dynamic content that allows us to complete a task: buy a product, book an appointment, compare prices and so on. Websites that use this type of information processing rank better than those based on static content. Functionality is a very important aspect of content marketing that prevents search engines from monopolising the process, as they only perform an informational role.
Example: English-speaking writers in the UK (a subpage that enables users to work with copywriters who know English). This is an example of detailed functional information that AI cannot copy.
Authenticity
This type of content is not purely informational, but also contains emotion and subjectivity. It includes, for example, a travel blog in which the authors describe their own feelings, genuine reviews based on personal use of a product, and all types of user-generated content, such as lively discussions and exchanges of ideas.
If content meets the above criteria, it has a greater chance of building visibility in search engines and LLMs. Regardless of this, good content also fulfils other important business functions – according to the Stackla report², for example, as many as 90% of consumers say that authenticity is an important factor when choosing a brand.
Data from CMI³, meanwhile, indicates that content marketing generates more than three times as many leads as traditional marketing methods. We should therefore remember that investing in the quality of information contributes to more than just SEO.
Off-site SEO
Off-site LLM SEO is based on digital public relations activities.
By definition, these activities involve communicating with a company’s surroundings, presenting a range of information about the business.
AI not only analyses the information on our website, but also processes information about us, collecting data from other portals. What can support LLM SEO?
Publishing sponsored and guest articles
The more information there is about a company across the media landscape, the more opportunities LLMs have to refer to it. With a diversified supply of information, we are more likely to appear in response to prompts such as “Which company provides service X?” or “Where can I find suppliers of product Y?”.
Collecting reviews and taking part in rankings
All comparisons, rankings and directories that gather information about a company offer a clear benefit for LLM SEO. They can help us gain visibility for prompts such as “List the best companies in industry X”.
The potential range of queries is unlimited, so it is worth ensuring the broadest possible exposure online. In doing so, we highlight our presence and maximise the chance that an LLM will notice us during its research.
Summary
- Searching for information through AI continues to grow in popularity because of its ability to generate data quickly and accurately.
- LLM SEO focuses on content that the model does not have in its training data. Publishing up-to-date, niche or detailed information therefore maximises the chances of being included in AI results.
- On-site activity requires the creation of unique, high-quality content. Off-site SEO, meanwhile, focuses on building a company’s reputation across the media landscape.
- Valuable content not only increases visibility in LLM tools, but also in traditional search engines. It also delivers other business benefits, such as building a reputation and generating leads. It is well worth the effort.
Leave a Reply