• The Web as a public entity is thirty years old since it was launched in the summer of 1991, according to our partner The Conversation.

  • The robot portrait of hypertext browsing in 2021 reveals a session typically started on a search engine, from which you access a multitude of sites.

  • The analysis of this phenomenon was carried out by Victor Charpenay, teacher-researcher at the computer science, modeling and systems optimization laboratory of Mines Saint-Etienne - Institut Mines-Télécom.

On August 5, 1991, a few months before I was born, Tim Berners-Lee made public his invention, called the “World Wide Web”, and encouraged anyone who wanted to discover it to download the world's very first prototype Web “browser”.

The Web as a public entity is therefore thirty years old.

Tim Berners-Lee praises the simplicity with which one can, with World Wide Web, access any information from a single program: his browser.

Thanks to hypertext links (shortened hyperlinks), navigation from one page to another is done in one click.

The principle, which at the time was still a subject of research, however seems to have worn out over time.

Thirty years later, the nature of our web browsing has changed: the number of sites we visit is reduced and the time spent on a site is increasing.

Hypertext in the past: exploration

One of the first scientific studies of our browsing behavior, carried out in 1998, made a strong assumption: hypertext browsing would be used mainly to search for information on a site.

To explore, in short, the tree structure of the site by clicking.

Search engines weren't very efficient yet, and Google Inc. had just entered the business registry.

Again in 2006 (according to another study published the following year), it would seem that Internet users only used a search engine once in six to launch a browsing session, which was then followed on average by a dozen clicks.

In 2006, Internet users only used a search engine once in six to launch a browsing session © B. Dada / Unsplash

Today, like the majority of Internet users, your first instinct is most certainly to “Googling” what you are looking for, bypassing the (sometimes tedious) search from click to click. Often the first result of your search is the correct one. Sometimes Google even displays the information you are looking for directly in the results page. There is then no more click, therefore no more in-game hypertext navigation.

To be able to measure this decline in hypertext from 1998 to today, I too performed a (modest) analysis of browsing behavior, based on the browsing history of eight people over a two-month period ( April-May 2021).

These people have provided me with their history voluntarily (no snippet of code is hidden in their web pages, unlike other browsing analysis algorithms) and the names of the websites visited have been anonymized (www .facebook.com becomes *. com).

The summary of the recurring patterns that can be seen in these browsing histories shows not only the importance of search engines, but also the concentration of our browsing on a small number of sites.

Hypertext today: cruising

Not everyone uses the web with the same intensity.

Some of the histories analyzed come from people spending the vast majority of their time in front of the screen (me, for example).

These logs contain between 200 and 400 clicks per day, or one every 2-3 minutes for a 12 hour day.

In comparison, people who use their browsers for personal use only get an average of 35 clicks per day.

If we take a daily average of 2h30 of navigation, a user makes a click every 4 minutes.

One click every 4 minutes on 2h30 of navigation © M. Spiske / Unsplash

How are these clicks structured during a browsing session?

One figure seems to show the persistence of hypertext in our habits: ¾ Web sites we visit are the result of a click on a hyperlink.

More exactly, only 23% on average of sites are "source" sites, coming from the home page, a bookmark or a suggestion of the browser.

However, when looking at the number of pages visited per website, the dynamics are different.

Indeed, most of the pages visited come from the same sites.

On average, 83% of clicks are made within the same site.

This figure is relatively stable over all of the eight historical data analyzed;

the minimum is 73%, the maximum 89%.

Typically, we move from one Facebook page to another or from one YouTube video to another.

There is therefore a dichotomy between “main” sites, which we focus on, and “secondary” sites, which we occasionally consult.

The main sites are very few: 10 maximum, or barely 2% of all sites visited by a person.

Most of the people in the analysis only have 2 main sites (maybe Google and YouTube, if we look at the statistics of the most visited sites in France).

Google and YouTube are the two most visited sites in France © Geralt / Pixabay

On this basis, we can draw the robot portrait of hypertext navigation, thirty after the democratization of the principle. A browsing session typically begins with a search engine, which leads to a multitude of sites. For most of these sites, you consult them once before starting your search engine again. For the few main sites of its navigation, we always access it via its search engine, but once on the site, we perform many actions before stopping the session.

The diagram below summarizes the robot portrait that I have just made.

The sites which launch a browsing session are in yellow, the others in blue.

By analogy with exploratory browsing in the 1990s, browsing today looks more like a slow cruise on just a few platforms, most likely social platforms like YouTube and Facebook.

The phenomenon that our navigation concentrates on a handful of sites is not unique to the Web.

It is one of the many instances of Pareto's law, which originally stipulated that the majority of the wealth produced was held by a minority of individuals.

We find this statistical law in many cases of socio-economic studies.

A simplified navigation graph;

a node of the graph represents a website (yellow for a site launching a navigation session, blue for the other sites) and an edge represents one or more clicks from one site to another (thickness proportional to the number of clicks) © Victor Charpenay

But what is interesting here is to see that the phenomenon of concentration is intensifying.

The 1998 study gives an average of between 3 and 8 pages visited per site.

That of 2006 mentions 3.4 visits per site.

The average I get in 2021 is 11 visits per site.

Equip your navigator with a "window"

The principle of hypertext navigation is today largely misguided by the major platforms of the Web. The majority of hyperlinks between websites - as opposed to self-referencing (links pointing from a website to itself, in blue in the diagram above) - are no longer used by humans to navigate, but by machines. automatically install snippets of spy code on our browsers. Serge Abiteboul and Pierre Senellart explain in more detail how to guard against this surveillance and go in search of our lost data.

There is a small community among researchers who still see the interest of hypermedia on the Web, especially when Internet users are no longer humans, but bots or "autonomous agents" (who are programmed to explore the web rather than staying on the same site).

Other initiatives attempt to give Internet users (humans or bots) more control over their browsing, such as Solid, the new project by Tim Berners-Lee.

Our "Internet" file

As an individual, you can inspect your own web browsing for patterns (and possibly change them).

The Navigation Window web browser extension serves this purpose, it is available online for Chrome and Firefox - if you wish, you too can help flesh out my analysis by submitting your history (with site name anonymization) to through the extension.

To do so, you just have to follow the right hyperlink.


Tourism: How can virtual reality experiences make you want to travel?


Monitoring emails, keyloggers ... Can an employer with impunity "police" his employees by teleworking?

This analysis was written by Victor Charpenay, teacher-researcher at the computer science, modeling and systems optimization laboratory of Mines Saint-Etienne - Institut Mines-Télécom.

The original article was published on The Conversation website.

Declaration of interests

Victor Charpenay is a member of the World Wide Web Consortium (W3C).

  • Facebook

  • Computer science

  • Internet

  • Google

  • Youtube

  • The Conversation

  • Video

  • High-Tech