The Evolution of SEO with Generative AI

ai generative seo

Let’s start by setting things straight: artificial intelligence has and will change the way we search for information on online search engines, Google first. In fact, in the coming years, increasingly online search will evolve in hybrid solutions, the result of the union of the search engine (which selects the best results based on relevance, quality and the authoritativeness and reliability of the source) and a language model (which amalgamates information into answers for the user).

Therefore, there are those who say that Generative AI and in particular tools like ChatGPT represent the death of online searches as we have known them to date. This could also mean the end of web page visits to news outlets, but will this really be the fate that awaits us?

Concerns grew when, in May, Google unveiled its Supercharging Search, i.e. a version of its search engine enhanced with generative AI that will provide ‘conversational’ answers and no longer (not only) a list of pages sorted according to SEO criteria and the PageRank algorithm.

With new generative AI capabilities in Search, we’re now taking more of the work out of searching, so you’ll be able to understand a topic faster, uncover new viewpoints and insights, and get things done more easily”, writes Google describing this new research model. 

Let’s take for example a question like: “What are the best music festivals in Europe for folk music lovers?“. Normally we would get a list of pages to visit to search for the answer by dividing the various information into those more or less relevant to our question. Or we could read the article from our favourite newspaper, which may have overlooked certain details for the sake of brevity. With the new search engine, all this discovery and reading activity will be processed directly by the search engine’s AI, which will provide a discursive answer based on a pre-selection of answers. It is not yet clear how the pre-selection will take place. Some words will be highlighted so that users can click on them and go deeper, but if they wish, the answer could be limited to the result suggested by Google.

Of course, it is not just Google. Microsoft has also integrated Bing, which works in the same way. The same direction is taken by other conversational search engines that are gaining some visibility these days, such as Perplexity.aiYou.com and Neeva AI, which have in fact presented hybrid solutions (search engine and language model).

How will this impact on publishers? 

Publishers are expressing concern about this way of browsing results, mainly because they fear that users will have less and less need to continue their in-depth study by clicking on a result. Incidentally, this concern about a change in SEO has been alive for some years now.

In fact, the same thoughts have already emerged in the past regarding the amount of information directly present in SERPs when performing a search, and it has reached its peak with the “Pay Per Position” service that allows websites to be placed in the top spots of the search engine regardless of page structure and content.

Again, successful content production, which in this case will be measured by the number of times that content is used to generate a ‘discursive’ response to the user’s question, will depend solely on the quality of the information. And it is the same difference that in recent years has favoured relevant and information-rich content over hundreds of ‘little articles’ written only by complying with some SEO rules and that have contributed to polluting the proliferation of content with irrelevant information written ‘for The Search Engines’ and not for the end users and readers. Google itself in drafting its guidelines on the use of artificial intelligence to produce text for the web, puts the focus on an extremely specific concept: ‘Rewarding quality content, regardless of how it is made.

Hallucination problems

A recurring theme in the field of Generative Artificial Intelligence are the hallucinations that characterise current language models. The risk here too is that the answers provided by the search engine may be correct in form but based on concepts that deviate from reality. And it is not clear how this information is selected.

“Scaling neural network models – making them bigger – has made their faux writing more and more authoritative-sounding, but not more and more truthful,” says Gary Marcus, American psychologist, cognitive scientist, and AI expert. 

Subscribe to Newsletter

Lorem ipsum dolor sit amet consectetur. Aliquam malesuada tempor ipsum a.

Follow Us:
Recent Articles