Digital Economy

From Google to agents: how AI in search will change again in 2026

by Gianni Rusconi

18 December 2025

4' min read

Translated by AI

Versione italiana

4' min read

Translated by AI

Versione italiana

The world of online search experienced substantial changes during 2025, in the wake of the previous year and in the aftermath of the global availability of ChatGPT and other generative AI models. Artificial intelligence has progressed steadily in every direction, and every platform, starting with Google's Gemini and obviously ending with OpenAI's (and consequently also Microsoft's Copilot ecosystem), has made several steps forward. Speaking of 'search', the transformation that companies and professionals have had to face can be summed up in this concept: no longer a list of results, but an increasingly intelligent, customised and contextual experience, marked by the (AI's) ability to understand the context, the user's need and the circumstances in which the query arises. Search, this is the summary on which experts agree, has become conversational surprisingly quickly and users, for their part, have progressively reduced the use of 'dry' keywords and increased complex and specific queries, relying on assistants capable of interpreting intent and preferences. But with this in mind, what can we expect for the

2026? Ale Agostini founder of Avantgrade.com, an SEO, SEM & AI Search agency, tried to answer this question by confirming how we will generally see a further acceleration of the impacts of LLM technology on marketing, e-commerce, content and customer journey and how the rules of online visibility will change.

Google changes skin

In the two-year period 2023-2024, the Mountain View giant remained relatively cautious on the subject of artificial intelligence. But this was probably only an apparent calm: in the last twelve months, Google has in fact pushed on the accelerator by launching several new Gemini models and entering the world of AI processors, the unchallenged realm (until now) of Nvidia, in a big way. By 2026, Alphabet is expected to control the entire technology stack, from silicon computing architectures to end applications. The performance of Gemini 3 will be applied to the services of the BigG galaxy, combined with the efficiency of its TPUs (Tensor Processing Units, the specialised computing units for processing AI models and optimised for machine learning operations), while Google Ads will undergo a significant evolution, with the interface of the traditional organic results page displayed by the search engine destined to further (progressively) lose centrality.

Advertising costs and multimodal responses

The integration of AI and the Gemini 3 model in Google Ads will lead to improved conversion ability but will also have the effect of increasing the cost per acquisition, i.e. both the fee required to obtain the expected result from a paid campaign. A similar trend, according to Agostini, is also expected in Meta. Several companies will therefore be called upon to reconsider the distribution of advertising budgets, re-evaluating the role of SEO and the growing GEO (Generative Engine Optimisation) AI search, where competition is not yet exasperated. The answers provided by AI engines, this another trend on the horizon, will become increasingly synthetic, multimodal and aimed at understanding the user's real intent. The space for text will diminish in favour of results with multimedia elements and those operating with these tools will be encouraged to understand and exploit the potential of more intuitive and immediate paths, making greater use of voice and image search.

The role of AI agents

As in other areas, AI agents will find more room for intervention and application, starting to perform concrete actions on websites on behalf of users, searching, comparing and filling in forms and purchasing products or services. This scenario will require B2B and e-commerce companies to rethink their online visibility and the entire user experience: it will no longer be enough to be convincing to the human user, explains Avantgrade's founder, but it will also be necessary to be easily interpretable and technically usable by browsers and intelligent assistants that adopt more structural and rational criteria. In order to be 'chosen' by intelligent agents, simply put, marketing will have to have a new mindset and consider these tools as new 'target personas' to be satisfied.

Web analytics in crisis, moving towards an AI driven world

Clicks, sessions and site visits have fuelled web analytics systems for years. In the near future, traditional tracking is destined to lose ground and information value, and as early as 2026 we will see a redefinition of performance metrics: crawl budgets for AI, citations, presence in responses, influence on decisions and quality of structured data will become central. For years, web analytics systems have been based on clicks, sessions and site visits. With answers provided directly by AI, which can strongly influence purchasing processes, traditional tracking is set to lose ground and informative value. In 2026 we will see a redefinition of performance metrics and elements such as AI crawl budgets (i.e. the resources artificial intelligence systems decide to invest in exploring and analysing a site's pages in a predefined time), citations, presence in responses, influence on decisions and quality of structured data will become central. Companies, Agostini points out in this regard, will first of all have to understand that new media cannot be measured by the old indicators and that with AI search everything changes.

Real and dynamic customisation of responses

Search activity will progressively become 'situational', it will not merely return results but will shape decisions, purchase choices and information paths through intelligent systems capable of filtering and selecting content before it is even seen by users. What does this mean? That companies and brands will have to gain visibility within these new logics, because it is on this level that the game will be played between those who will know how to make themselves recognised as relevant by machines (exploiting generative models to create dynamic and customised content for each user based on behavioural data and preferences in real time) and those who will have to adapt to a less predictable scenario than in the past. And from a practical point of view, what will change? Given the same question, the answers given by the engine will vary according to context, history, goals, preferences and stage of the buying process, opening up the challenge of having to optimise sites and campaigns by clusters of intent (types of searches) and overcoming the rigidity of traditional SEO models.