AI Traffic

Anyone who runs a website wants to know where their visitors come from. In the past, there were only a few ways to find out: the search engine, a link on Facebook, a newsletter, or the address typed in directly. For a few years now, there has been another way. People ask a chat program their question, and it answers in full sentences, and then they click on one of the sources this program names. These are exactly the visitors called AI traffic. Traffic is simply the English word for the flow of visitors to a page.

What publishers and shops depend on

For many websites, visitors are the basis of the business. News sites earn money from advertising that appears alongside their articles, and from subscriptions that readers take out. Online shops earn from what visitors buy. If the number of visitors falls, revenue falls too. That is why the industry watches very closely how the path to the website is changing.

And this path is currently changing fundamentally. Anyone who once wanted to know how long to boil an egg ended up on a recipe page. Today they get the answer directly in the chat window and don’t click any further at all. Experts call this a zero-click answer, meaning an answer without a click. AI traffic is what remains of this process: the visitors who still click through to the source anyway.

In numbers, this share is small so far. For most larger news sites, only a low single-digit percentage of visitors comes from AI services. However, it is growing quickly, while the flow from classic search engines is shrinking for many sites. This exact divergence is the reason for the nervousness in the industry.

How AI visitors are identified at all

Every click on a link leaves a small note behind. The browser transmits to the visited site the address the user came from. This piece of information is called the referrer. If the address of a chat service is listed there, the analytics software attributes the visit to AI traffic. This is how the percentage figures that appear in reports come about.

However, this measurement is imprecise. Some AI services obscure the origin, others don’t pass it on at all within apps. And anyone who reads an answer in the chat and later types the brand name into the address bar themselves shows up nowhere as an AI visitor. The measured value is therefore more of a lower bound than an exact figure.

It is important to distinguish this from a similarly sounding term. Besides real humans, programs also visit websites, so-called crawlers. They collect text for training AI models or fetch current pages to generate an answer. This machine traffic costs server capacity but brings no readers. In statistics, it should be counted separately.

Where the term appears in the news

AI traffic is most often mentioned in quarterly reports and interviews from media companies. Publishers use it to justify cost-cutting programs or new contracts. Several major publishers have signed licensing agreements with AI providers: they permit the use of their texts and receive money as well as links in return. Others are suing in court instead.

A new buzzword has also become established in marketing. Those who used to try to rank high in Google's results list are now trying to be named as a source in AI answers. Acronyms such as GEO or AEO are circulating for this, meaning optimization for generative answer engines. Whether this can really be planned for is disputed among experts.

One frequently cited advantage: AI visitors come less often but buy more frequently. They have already clarified their question in the chat and only click when they really want to know more. For advertising providers, this shift is nonetheless uncomfortable, since fewer page views mean fewer ad placements. So far, there is no replacement business model for the lost clicks.