The spread of AI tools (like ChatGPT, Gemini, Claude, Perplexity and others) is revolutionising the way people search for information and make decisions online. Consequently, SEO, traditionally focused on positioning in "classic" search results, is now facing a paradigm shift.
From customer journey to the AI journey
Once upon a time, the online consumer journey followed an almost liturgical ritual: it started with a Google search, followed by visits to various websites, slaloming between advertisements, reading reviews, making comparisons, etc. Today, this pattern is evolving into an "AI journey": users increasingly rely on AI assistants and chatbots to obtain advice, ideas and recommendations: what to buy, where to go, which solution to choose.
From digital liturgy to faith in artificial intelligence.
A recent study found that 35% of consumers who have used AI assistants replaced search engines to get answers to questions. In other words, more than a third of active chatbot users prefer to query an AI rather than search on Google (link).
Imagine your customer negotiating directly with ChatGPT instead of browsing the product catalogue on your website. Today it may seem dystopian, tomorrow it will be reality: the funnel will evolve into a continuous and direct dialogue, part of an infinite conversation with a digital oracle.
For companies, this means that the main entry point to their brand might no longer be the homepage, the search engine or the social platform, but a voice or text response given by an AI model.
Welcome to the AI journey.
Brand awareness in AI models
If AI establishes itself as the new gatekeeper of information, the presence – or rather, the correct representation – of the brand becomes crucial. Chatbots are no longer simple Q&A tools; they are assuming the role of "influencers" for consumers. When a user asks "what is the best [product X] for [need Y]?", the AI draws on its knowledge (training datasets and possible integrations from search) to provide an answer that, for various reasons, might not include your brand.
And this is where it gets interesting: if your brand is present in that knowledge and is described adequately, your image will be enhanced.
It therefore becomes necessary to translate PR and branding concepts into the "weights" of language models. However, there is a subtle risk: misrepresentation of positioning.
AI is not infallible; it might present your pasta brand as "gourmet haute cuisine", or, conversely, classify your premium fashion atelier as "everyday casual clothing".
This phenomenon is particularly insidious, as it is difficult to control and risks creating a misalignment between consumer perception and your offering.
Moreover, tests show that ChatGPT's search tools are vulnerable to manipulation - generating misleading or even harmful results when web pages contain hidden or manipulative text (link). This raises concerns about information security, as brands could suffer reputational damage not only due to AI's unintentionally erroneous interpretations, but also due to deliberate sabotage attempts by unscrupulous competitors.
The good news is that, over time, AI reliability will tend to improve; it should be noted that the "weights" in neural networks are not random, but generally reflect the authority and relevance of the brand in the training dataset. The more positive and contextual the mention of the brand in authoritative sources, the greater the probability of being recommended. This creates an interesting parallel with traditional SEO, where authority translates into quality backlinks – today it manifests in favourable representations in the training corpora of LLMs.
Let's do a test. What would ChatGPT answer if I asked it which is the best digital agency in Reggio Emilia? (Sorry, I kept it in Italian).
Today it went well.
The response can vary based on a thousand factors, among other things depending on the model and whether it uses online search. In the first case, your brand will need to be present in the LLM's memory—embedded in the neural network’s weights. In the second case, most LLMs use search engines, so we return to the origin.
Exercise: try asking questions with web search enabled/disabled.
How to be present in LLMs
The winning strategy is to dominate the online sources from which AIs draw: create authoritative content on your site, take care of markup in structured data, participate in Wikipedia, avoid being too opaque with product sheets, obtain citations and positive reviews. All this increases the likelihood that a chatbot will recognise you as a reliable brand.
Brand awareness today requires a double effort: convincing both people and neural network algorithms of your relevance.
The AI’s seemingly impartial recommendation for your products can directly influence the purchasing decision. There is abundant research confirming LLMs' ability to influence and persuade people (and this should lead us to broader reflections).
Google AIO and AI summaries: traffic decreases... or transforms?
Search engines are also equipping themselves with generative AI responses. Google, for example, has introduced "Google AI Overview" (in Europe it will arrive late, as usual), AI-generated summaries that appear at the top of search results for certain queries, synthesising information from various sites and immediately showing the answer to the user.
Example: Google AI Overview in US
The presence of an AI Overview causes a drastic drop in organic CTR. A recent study reveals that, when an AI summary appears, only 0.64% of searches result in clicks on organic results, halving the CTR compared to the pre-AIO period (link).
In fact, Google's generative AI "devours" clicks that could have led to websites, although less casual traffic could mean more targeted and high-quality interactions (long tail keywords).
This raises questions about the value of traffic in the AI era. In my view, we will witness a bifurcation:
- Decline in informational traffic: Purely informational traffic will decrease dramatically, as users get answers without visiting sites.
- Growth in high-intent traffic: Traffic with high intent will increase, composed of users who have already passed the informational phase thanks to AI and are ready for action.
It's a bit like having fewer customers entering the store, but more motivated to purchase. This scenario requires a rethinking of SEO ROI measurement, where traffic volume might become less relevant compared to quality metrics such as conversion rate and value per visit.
Generative AI is reshuffling the deck: it is useful to be present both in the AI summaries of search engines and in chatbot responses.
Soon any SEO strategy will have to include "presence in AI results" and "visibility on LLMs" as new KPIs (the case of Semrush is significant, link).
AI-friendly sites: accessibility and optimisation beyond traditional SEO
With AI becoming an intermediary between user and brand, it is essential to make websites easily accessible and comprehensible for artificial intelligence models. This goes beyond traditional on-page SEO: pleasing Googlebot is no longer enough.
The good news is that a well-made, usable and accessible site is generally so for both humans and crawlers, whether "traditional" or "AI".
A concise, non-exhaustive list of good practices:
- Technically optimised sites: Speed, mobile compatibility, absence of broken links, clear structure, updated sitemap and useful meta-tags. If an AI finds it difficult to extract information from your site, it will probably move on. Conversely, a well-structured site helps crawlers scan and efficiently interpret content, improving your visibility in AI responses.
- Optimised content (quality and "conversational"): AIs reason in terms of natural language: they understand the meaning of phrases more than individual keywords. This entails two things: first, aim for original and useful content. Second, include conversational key questions and phrases that users might pose to AIs in your content. For example, instead of targeting only "running shoes", it's advisable to cover longer queries like "running shoes for long-distances in the rain" – these are more colloquial phrases that users actually use with chatbots.
- Structured data and semantic clarity: Another aspect is facilitating semantic understanding by AIs. AI models use techniques like Entity Recognition to identify entities (brands, places, products) in text. Mention brand, product and category names consistently and precisely, also using structured markup. Help the AI connect the dots. Consistency and structure in data improve information retrieval.
AI as a Universal Intermediary: A Paradigm Shift
Projecting into the future, it is already possible to glimpse the destination of change: AI becomes the main mediator between users and machines. If today a user interacts with a website or an app, tomorrow they will simply give a command to the AI that will execute on their behalf: a sort of transfer of their will through an avatar.
Jakob Nielsen describes AI as "the first new interface paradigm in the last 60 years": if traditionally the user gave instructions via commands or clicks, now they simply communicate the result they want to obtain, leaving it to artificial intelligence to determine autonomously how to achieve it.
Imagine a user in 2035: "Hey Bot, book a flight to Istanbul next weekend, find a central hotel with parking and send me the opening hours of the Blue Mosque". The AI will consult flight sites, use booking APIs, interact with interfaces and return the final result. The user will never have visited a site or filled out a form – they’ve delegated everything to the assistant. We are facing a paradigmatic change: from "clicking links" to giving instructions in natural language.
It's as if every user had a personal assistant exploring the internet on their behalf, an omnipresent digital proxy.
For companies and marketers, this scenario imposes a total rethinking. The traditional website and user interface will remain important, but will be consulted increasingly often by AI agents instead of human users directly.
With AI as a mediator, the challenge for brands will be to make themselves understood by the digital assistant: this will push towards open APIs, universal formats and direct integrations. Those who embrace this paradigm first will become the assistant's preferred source, a valuable competitive advantage in a world where AI will act as a semiotic layer between us and reality, transforming both our perception and the very substance of what we consider "real".
Note:
- in the essay I use the terms AI assistants and chatbots interchangeably;
- my vision generally tends to overvalue technological determinism;
- I have not addressed the ethical issue because it would require deeper exploration on its own.
