Between 2007 and 2019 we ran IndiBlogger, India's largest blogging community. 95,743 people joined it. Every one of them handed over a web address on the day they signed up, and we still have the list. In September 2026 we went and knocked on all those doors.
Of the addresses those people paid to register, 52.5% no longer resolve at all. Of the free addresses on somebody else's platform, registered by the same people in the same years, 0.5% are gone.
That is a hundredfold difference between two groups who are otherwise identical. Same community, same years, same country, and in many cases the same person holding both kinds of address.
Why nobody has this number.
Because you cannot get it by crawling the web.
Every way of finding websites finds the ones that are still findable. Search needs a site to rank. Common Crawl needs it to have kept its inbound links. A directory needs somebody to have listed it. All of them look at the present, and abandonment is the thing that removes a site from the present. You end up measuring survivors and calling it a population.
A membership list written in 2009 has no such problem. The row exists because somebody joined in 2009, not because their site is findable today. The dead are still on it, and you can go and count them.
The Pew Research Center used the same method for its 2024 link-rot study, re-fetching a sample of pages captured years earlier instead of looking for pages that still work. The difference here is that our list is a real, named, dated Indian cohort, and it records what each person actually bought.
What we found.
Every figure on this page is taken from the study's own summary output. The three rates in the table carry 95% confidence intervals; the counts and medians further down are stated as counts.
| In 2026 | An address they paid for | A free platform address |
|---|---|---|
| No longer resolves | 52.5% (50.7 to 54.3) | 0.5% (0.2 to 0.8) |
| Decidably gone | 56.7% (54.9 to 58.5) | 19.9% (18.3 to 21.5) |
| Still serves a page | 40.7% (38.9 to 42.5) | 80.1% (78.5 to 81.7) |
"Decidably gone" adds the addresses that resolve but answer nothing, and the ones that return a flat 404. On the free side that is almost all blogs deleted from the platform rather than addresses that stopped existing.
The death rate does not climb with age.
The 2007 intake sits at 50.6%. The 2018 intake sits at 44.4%. An eight-year-old paid address is about as likely to be gone as a nineteen-year-old one, which is what you would expect if the thing that kills them is a renewal somebody stopped paying rather than a slow decline.
Nor do the survivors rot quietly.
Fewer than half the surviving paid addresses publish a date at all, and only one in ten of the surviving free blogs do. Among the ones that do, the median copyright year is 2026 on the paid side and 2014 on the free side, and 85% of the dated free blogs have not been touched in five years.
The rendering pass says the same thing. Of the surviving sites it could actually judge, 122 of 179 paid domains read as a modern site, against 2 of 53 free blogs. It returned no verdict on the rest. A paid address is either tended or gone, with very little in between.
The free platform preserves the corpse. It costs nobody anything to leave a dead blog on the internet, so it stays. A paid address has somebody deciding, once a year, whether to keep paying, and about half of them stop.
Some of what looks alive is not.
A lapsed domain bought for resale answers a fast, healthy 200 and passes any automated check. We drew 100 of the surviving paid domains that showed parking signals and looked at every one by hand. 75 were dead: parking landers, resale listings, expired registrations, and 11 domains now serving gambling or piracy brands. One of those 11 was arunachaldiary.com, which used to be a travel blog about Arunachal Pradesh and today sells online slots.
The modelled figure, which is a different kind of claim.
If the flagged survivors we did not examine are like the ones we did, then roughly 70% of the paid addresses are gone, parked, or serving somebody else's content. The point estimate is 70.4%.
That is modelling, not measurement, and we are labelling it as such. The number to hold us to is 52.5%, because anybody with a DNS client can check it.
What this is not.
These were bloggers, not businesses. IndiBlogger was a community of writers, photographers, cooks and hobbyists, so what we have measured is the fate of Indian personal and creator websites. Anyone stretching it to Indian small business is modelling and should say so. Our own guess is that a business does better, because a business has a stronger reason to keep its address alive than a hobbyist does. Treat this as a ceiling on small-business loss rather than an estimate of it.
The cohort year is when the person joined, not when the address was registered. Somebody who joined in 2009 and added a second address in 2016 puts that address in the 2009 column. That leaves the pooled headline alone and affects the shape of the year-by-year curve.
And people who join a blogging community are not a random sample of people who put up a website. If that biases anything it biases toward survival, which makes 52.5% a floor.
How we measured it.
We asked every drawn address for its home page exactly as the registry recorded it. A domain that does not resolve is the least arguable measurement on the web, and we re-asked a sample of the deaths of Google's and Cloudflare's public resolvers: 25 of 25 came back confirmed gone.
An obstacle is not a death. A live server that refuses our declared research bot, a server error, a private blog behind a login wall: all counted and set aside, never scored as dead. That is 246 of the 6,329. A site whose security certificate has expired is not set aside either, and not counted dead: we ask again without checking the certificate, because a broken certificate is one of the commonest signs of the neglect we came to measure.
Every death went back on a 45-second clock, in case our own patience was the thing that had killed it. 24 changed their answer: six came back alive, and another 17 moved to states that prove a live server without proving a working site. 23 of the 24 left the death count.
One check most surveys never get: IndiBlogger ran its own link checker and marked dead blogs before it closed in 2019. Asked again in 2026 with no knowledge of that column, the blogs it had already found dead came back 72.2% dead, and the blogs it last saw alive came back 28.6% dead. Our instrument and a seven-year-old observer agree.
We kept the record of our own mistakes too. An earlier run of this study had six defects that moved the figures, five of them in the direction that flattered us, and one hand verdict we got plainly wrong: a live site that renders in the browser, recorded as a parked domain. We found it by re-checking every verdict that had been reached without looking, rather than the ones that looked doubtful.
Why we care.
We run websites for a living, so this is the number underneath our own business, and it is not flattering to the industry we are in.
A website is not something you buy. It is something somebody has to keep running, and the evidence here is that when that somebody is the owner alone, about half of them stop within a few years. The site rarely got bad. A renewal notice arrived at a busy time, or the person who set it up stopped answering, and nothing made it urgent until the address was gone and somebody in another country was selling it back.
Use this.
If you want to quote this, quote the measured number, name the population as Indian personal and creator websites, and link back here so your reader can see the caveats.
The method is above in full. We are not publishing the list itself, because it is 6,329 real people's web addresses and most of those people are still using them. We will share it with any researcher or journalist who asks: write to [email protected].
If anything here is wrong we would like to know, because we have to live with the number too.
We are the somebody who keeps it running.
Paste your web address and we'll email you a fresh redesign in 48 hours. Free, no card. If you like it, it's $29 a month to host it, change it and keep it found, and your domain and code stay yours to take whenever you like.
No card. No commitment. 48 hours.