It’s one of the questions that’s been coming up most often lately in any meeting where content or search engine optimisation is discussed. The short answer is yes, to some extent, for certain types of search. The long answer is more interesting: it’s not that traffic is disappearing, but that it’s changing form, and anyone who continues to measure only clicks in the old-fashioned way will be left with an incomplete picture of what’s happening.
It is worth starting by separating fear from the data. Fear suggests that Google is going to disappear and that nobody will ever click on anything again. The data paints a more nuanced picture: certain informational searches — those that previously led to a click on an article to find a specific piece of information — are increasingly being resolved directly within the AI’s response, without the need to leave it. That is true and it is happening. But not all searches are the same, and therein lies the key to it all.
What sort of traffic is really being lost?
Not all web traffic reacts in the same way to ChatGPT or Gemini. There is a fairly clear difference between two types of search:
Simple information searches — “what is GEO?”, “how many employees does such-and-such a company have?”, “how is such-and-such calculated?” — are the ones losing the most ground. If the answer fits into two sentences, the AI provides it straight away and the user doesn’t need to visit any website. This type of traffic—which previously relied on very basic content that was replicated across hundreds of different websites—is the one that is truly at risk.
Searches relating to decision-making, comparison or transactional intent — “which communications agency should I hire?”, “what’s the best option for my situation?”, “I want to speak to someone who knows about this” — work differently. In these cases, AI does not replace the final decision: the person still wants to verify, compare and look at real-life examples, and in many cases ends up visiting the websites of the options that the AI has suggested. That traffic does not disappear; it simply changes source. This is consistent with the fact that an AI system does not interpret individual web pages: it interprets entities, something we discuss in detail in our article on GEO.
Traffic isn’t lost; it’s redistributed
This is where it is important to be precise with our language, because ‘losing traffic’ and ‘traffic coming in a different way’ are two different things, and they are constantly being conflated in public debate.
What is happening is a redistribution. Previously, almost all traffic came via a list of ten blue results. Now that traffic is split between traditional search results, answers generated directly by AI (which sometimes include a link, sometimes not) and mentions made by generative AI when someone asks it a question related to your sector. The total number of people searching for information hasn’t disappeared. What has changed is how many entry points there are and what each one looks like.
For an organisation, this has a practical implication: measuring only organic visits in Google Analytics as if it were the sole indicator begins to paint an incomplete picture. If a generative AI mentions you as a reference when someone asks about your sector, that has value — in terms of visibility, credibility and reaching someone at the very moment they are making a decision — even if that person never clicks through to your website that same day.
Why widespread panic doesn’t help anyone
There is a pattern that repeats itself every time a disruptive technology emerges: first comes panic (“this changes everything and we’ll be out of business”), then overreaction (“we have to dive straight into this right now, without a plan”), and only later does a more measured analysis follow. Exactly the same thing is happening with the emergence of AI-powered search engines.
Panic leads to unhelpful decisions: compulsively publishing content in the belief that ‘more volume’ makes up for the loss of traffic, or simply abandoning SEO because ‘it’s no use anymore’. Neither reaction is an appropriate response to what is actually changing. The indiscriminate overproduction of content was already a problem before ChatGPT existed, and now it is even more so: a generative AI needs to find coherence and recognisable criteria in order to cite someone, not volume.
What really matters right now
Rather than wondering whether it will lose traffic, an organisation would be better off asking itself something else: if someone asks a generative AI about its sector, is it mentioned as a reference? That question is more uncomfortable, but also more honest, because it does not depend on a metric that can easily be inflated, but rather on whether there is a body of content that is coherent and specialised enough for an AI system to recognise it as a reliable source.
What you really need to keep an eye on is not a drop in traffic figures, but three more fundamental factors: whether basic informational searches relating to your sector are still driving traffic (it’s normal for these to fall; this isn’t automatically a cause for alarm); whether decision-making and comparison searches are maintaining their volume (you should certainly pay attention if these fall); and whether, when you ask a generative AI about your sector, your organisation is mentioned in any way.
None of these three issues can be resolved by panicking or by producing more content. They are resolved using the same tried-and-tested formula, applied with greater discipline: content organised into clear categories, consistency of approach across pieces, and patience, because this — just like traditional SEO — isn’t built overnight.