This is the essay that was Palo's home page until October 2026, re-issued as a report. Its figures are as they stood at the last analysis under the census's earlier counting rules (11 October 2026), from crawls up to 12 September 2026, and they will not change. The counting rules changed on 11 October 2026; see How the counting changed.
The wording was corrected on re-issue, without changing what the essay says: claims stated as certainties are now stated as likelihoods, the description of PageRank says only what it measures, and the note on the data no longer describes topic labels, which the census has since removed.
We found 176,719 South African websites. Most of them are barely connected to one another.
We built a crawler, went looking, and mapped what we found. This is what the data says.
Part One: Two computers
Start with the simplest possible picture. Two computers. One has information. The other wants it. How do they share it? You connect them with a cable. Simple. It works. Information flows from one to the other.
That is the internet, in its most basic form. Not magic. Not a cloud somewhere in the sky. Computers talking to each other through cables, fibre, and radio.
But what happens when a third computer wants in? A fourth? A thousand? You cannot connect everyone to everyone with direct cables — the number of cables required would be impossible to manage. So instead, you put something in the middle. A box that receives a message and forwards it to the right destination. We call that a router.
As more computers join, more routers are added. Routers connect to other routers. A message hops from one to the next, passing through several hands before it reaches its destination. That network of networks, billions of computers connected through countless routers, is the internet.
Part Two: The web
The internet moves information between computers. That is what it is built to do.
The web is something different. It is a collection of pages that live on computers connected to the internet. What makes the web special is that its pages link to each other.
A link is a connection from one page to another. You click it and you move from one computer to another, somewhere else in the world. That system of pages connected by links is the web. Not the internet. The web runs on the internet, the way phone calls run on electricity, one is the infrastructure, the other is what people do with it.
This distinction matters because of what links do beyond helping you navigate. When a page links to another page, it is pointing at it. Saying: this exists, this is relevant, this is worth visiting. The more pages point at something, the more visible it becomes. The fewer pages point at something, the harder it becomes to discover, no matter how good it is.
Part Three: Social media and the broader web
Social media is a private estate on the internet. Facebook, Instagram, TikTok; each one is a very large, very busy estate with its own roads, its own rules, and its own way of connecting people. When you build a presence on Instagram, you are building inside that estate.
The estate works well. Millions of South Africans are on it. The connections built there are real followers, conversations, and recommendations spreading from one person to another. For reaching people inside the estate, nothing works better.
But private estates have walls. People outside can often see that the estate exists. They may even be able to look through the gate. What they cannot do is participate in it the same way they can on the open web. The platform decides what can be seen, linked to, indexed or reused. The roads belong to the estate.
Outside the estates is the broader web. This is where websites live, pages with their own addresses, their own presence, not owned by any platform.
When you have your own website, anything else on the web can point to it. Another website can link to it. A search engine can discover it. A researcher can reference it. It becomes part of the shared record.
Most of South Africa's digital life happens inside the estates. This census asks a different question: what exists on the broader web, where anyone can build roads to anyone else?
Part Four: How the web gets read
The broader web is visited by people, and it is read by machines.
Search engines send out programs called crawlers. A crawler visits a page, reads what is on it, follows the links it finds, and keeps going. Over time it builds a map of what it has discovered. When you search Google, you are searching that map rather than the live web itself.
Much of the publicly accessible web has also become part of the data used to train modern AI systems. ChatGPT, Gemini, Claude and many others learned from enormous collections of text gathered from many sources, including the open web. Pages that are public, connected and discoverable are far more likely to become part of those collections than pages that are isolated or difficult to reach.
This means the web is more than a place to publish information. It is one of the ways knowledge becomes discoverable. What is connected is easier for people, search engines and AI systems to find, reference and learn from. What is disconnected is far less likely to become part of that shared record.
Once you understand that links shape what can be discovered, an obvious question follows.
What does South Africa's web actually look like?
Part Five: What we did
We built a web crawler — the same kind of program search engines use to build their maps. We gave it a starting point: a public record of South African websites that had configured secure connections. From there, the crawler visited every address it could find, followed every link it discovered, and mapped the connections between websites.
The result is more than a list of websites: it is a map of how South Africa's web is connected. The crawler is still running. The census below is a snapshot of what it has found so far.
Part Six: What we found
So far, we have found 176,719 South African web addresses, and we are still counting.
When our crawler visited them, 56,004 were unresponsive. Some have disappeared. Some were temporarily unavailable. Some are configured in ways our crawler cannot yet understand. The true number of abandoned websites lies somewhere inside that figure, but we cannot separate them precisely.
176,719 SA websites found
The first thing we looked for was connections. Think of a link as a road between two houses. We found 268,138, such roads which works out to roughly 2.3 roads per address, and many addresses have none at all.
Now consider a website with no roads at all. No links in. No links out. It exists. It has an address. But nothing connects to it, and it connects to nothing. A crawler that only follows links is unlikely to find it. Search engines have fewer ways to find it. It is much less likely to become part of the connected web. We call that an island.
44,567 SA websites that are islands — no roads in, no roads out
There is a second, quieter group. 21,142 South African websites have no roads in, but they do link out to others. They participate in the web. They point to other South African websites. They simply have not been pointed to yet. Not islands. Unrecognised.
Together, islands and unrecognised websites account for more than half of South Africa's 117,282 active websites. More than half.
One question naturally follows. If our crawler follows links, how did it find islands? Because we did not rely on links alone. We began with independent records of South African websites and then crawled outward from there. That allowed us to measure websites that a crawler starting only from existing links might never reach.
Being connected is not the same as being connected to everyone else. Even among websites with links, many exist in separate clusters. The largest group of South African websites where every member can ultimately reach every other contains just 8,155 websites — 7% of active websites. The rest live in smaller neighbourhoods that remain largely disconnected from one another.
Not every link carries the same weight. A link from a small, little-known website counts for something. A link from a widely referenced website counts for much more. Google's PageRank algorithm measures this by treating each link as a vote, with votes from more widely linked websites counting for more.
In SA's web, the sites with the highest PageRank were:
- sacoronavirus.co.za
- domains.co.za
- google.co.za
- tripadvisor.co.za
- justice.gov.za
- inforegulator.org.za
- payfast.co.za
- sars.gov.za
- popia.co.za
- ispa.org.za
Part Seven: What it means
The numbers above are not abstract. Behind every island is a real business, a real person, a real piece of work.
A link is more than a road for finding things. It is a channel, a way for information to travel from one place to another. When information travels, it does something: it gives people ideas. When enough people share enough of the same information, they begin to share ideas too. Shared ideas, held by enough people over enough time, become shared knowledge. Shared knowledge is what a community is built from.
This is what a connected web actually does. Beyond helping search engines, it is the mechanism by which separate people and separate websites become something more, a shared record, a shared understanding, a community with common ground.
A site with no links is harder to find, and it is also largely cut off from that mechanism. Little information flows in, and little flows out. Whatever that site knows, whatever it could contribute, has few channels through which it can travel. It is unlikely to become part of anyone's shared understanding, because little of it reaches anyone to be understood.
This is not absolute, and it does not happen instantly. A researcher can still find an unlinked site if they know where to look. A single new link can open a channel that did not exist yesterday. But without connection, a site sits outside the process through which a community's shared knowledge actually forms. No one decided to exclude it. It is simply absent from the mechanism.
The unrecognised site is not absent through neglect. Someone built it. Someone linked out to others. They showed up. The web just has not pointed back yet. One link from another South African website would open that channel.
The web itself works the same way for an entire country. When enough of South Africa's websites link to one another, they begin to function as shared infrastructure, a common ground that search engines, AI systems, and researchers can all draw from, much like a language or a currency becomes common ground for the people who use it. When too much of the web remains disconnected, that infrastructure does not fully form. The parts are there. The shared ground is not.
If the mechanism that builds a shared, connected record does not function across large parts of South Africa's web, the country's own record of itself stays incomplete.
The record is being written now. Every link is a small piece of shared infrastructure — built once, but there for anyone who comes after.
Part Eight: What you can do about it
The mechanism behind this problem is also the mechanism behind its solution.
Links are how the web becomes connected. Every link from one website to another is a signal; this exists, this matters, this is part of the record. Every link creates another path for people, search engines and AI systems to find something that might otherwise be harder to reach.
You do not need to build a directory. You do not need a campaign. You need to link to one South African website you think deserves to be found.
That is a road. Built by you, today, that did not exist yesterday.
If you have a website, link to other South African websites. If you know someone with a website, encourage them to do the same. The record South Africa builds about itself is made one link at a time. That is how this infrastructure gets built.
A note on the data
What this census can and cannot see
Our crawler visits websites and reads what is on each page. Some websites only display their content after running code in your browser. Our crawler cannot always interpret those pages, which means some websites may appear empty in our data when they are not.
When we mark a website as unresponsive, it does not necessarily mean it has been abandoned. It may simply have been unavailable when we visited. We revisit websites over time and update our records.
The rankings shown are calculated only from the South African web we have mapped so far. They are not measures of global popularity.
The census is still growing. We have not found every South African website yet.