What is Generative AI Optimization (GAIO)?

What is Generative AI Optimization (GAIO)? And why should not only marketers, but also PR professionals take notice?

Generative AI Optimization (GAIO) is a new, largely still hypothetical marketing discipline, similar to Search Engine Optimization (SEO). In our agency, we have so far called this discipline Chatbot Optimization, but moving forward we will refer to it as GAIO ([?ga?jo]), a term proposed by the German SEO expert Philipp Kl?ckner .

GAIO revolves around the question of how companies can ensure that their brand and products feature prominently in widely used Large Language Models (LLMs) such as ChatGPT, Bing Chat and Google Bard. Figuring this out is important, as these models could potentially influence many purchasing decisions in the future – and are likely already doing so to an increasing extent. Some examples: If you ask Bing Chat about the best toaster for a dorm kitchen, the AI recommends a device from Philips. If you pretend to be the CEO of a fast-growing startup and ask for help in selecting a suitable ERP software, it proposes to try SAP Business One. How do these recommendations come about?

Brand mentions replace backlinks

Bing Chat always considers the context that is provided by the user. For example, it will suggest a different laptop to a gamer than to a graphic designer. The AI cites secondary sources such as trade media, blogs or comparison sites that inform its recommendations. Manufacturer pages, on the other hand, are almost never used as a source.

This upends the established rules of search engine optimization, at least to some extent. In traditional web search, backlinks continue to be one of the most important criteria for relevance. Whoever can collect more high-quality links has a clear advantage in the competition for the highest rankings in the search results. In chat-based search, the situation seems to be different: Here, the focus is no longer on backlinks, but on brand mentions. It's no longer crucial who links to my website, but where I'm mentioned and what is being said about me. (Great news for PR consultants!) At the same time, one’s own website loses importance if search engine users are less frequently redirected there.

Putting theory into practice

At the beginning of this text, GAIO was referred to as a "hypothetical" marketing discipline; perhaps it is also justified to speak of an "emerging" discipline. There are two hurdles to overcome: First, a critical mass of users needs to be convinced about the benefits of chat-based search. Second, we need to figure out what the GAIO toolkit will look like.

In order to attract and retain a sufficient number of users, AI models need to significantly improve – especially when it comes to product recommendations and product-related questions. While Bing Chat is already quite good at providing context-based and even comprehensibly justified recommendations, Google Bard currently only provides brief product lists without justifications or source citations, offering no added value to the user. ChatGPT, on the other hand, offers an enormous amount of advice and context, but refuses to recommend specific products or brands. This is because the training data for this LLM currently only extends up to September 2021, unlike Bard and Bing Chat which can access live data from the internet. However, if the pace of technology advancement continues as it has over the past few months, these issues could soon be resolved.

The GAIO toolkit

How will GAIO work in the future? The most immediate problem for companies is certainly the lack of visibility: Currently, they do not know how often and in which context their products and brands are mentioned in the answers of LLMs. What questions are asked and how frequently? Do the answers include outdated or inaccurate information? In which contexts are certain products recommended, and what are the sources of these recommendations? Are competitors mentioned more often??To answer such questions, new tools are needed that ought to function similarly to popular SEO tools, such as Semrush or Sistrix.

Once this visibility is established, aspiring GAIO specialists must then learn to understand which factors influence a company's presence in the LLMs' responses. This refers to the parameters that are known in search engine optimization as ranking factors. So far, Bing Chat has been the primary source of insight into how responses are formulated, as it links to information sources beneath its responses. For ChatGPT, at least the training data underlying the model and their weighting are known – a good starting point for further research.

As soon as we understand these variables, we can try to influence them. That's when the real optimization begins. A relatively simple tactic could involve systematically evaluating which websites the LLMs regularly use as sources for certain product categories. Subsequently, brands could attempt to secure placements of their own products and messages on these pages.

In order to be able to understand whether such measures actually make a difference, it is necessary to measure their success. So far, all the prerequisites for this are missing, because the providers of the LLMs do not yet provide any insights and statistics. However, this is expected to change in the next few weeks: Microsoft wants to take the first step by integrating Bing Chat Reports into its Webmaster Tools. In addition, Bing Chat is to be specifically identified in the referrer data so that website owners can understand how many visitors the chatbot directs to their sites. From this point on, it will be possible to draw initial conclusions and better understand the relevance of Generative AI in the customer journey.

How hypothetical is GAIO?

Is GAIO more than a thought experiment? Yes – provided that we see an increasing number of users employ Generative AI for product research. If this happens, more and more brands will eventually start to wonder how they can increase the likelihood of appearing in the responses of popular LLMs. But as with most fast-growing digital channels, a significant first-mover advantage beckons. Thus, engaging with GAIO at an early stage could prove worthwhile.