How Long Do Backlinks Last? We Checked 16,012 of Ours — Twice

We re-fetched every placement in thirteen delivered orders — 16,012 links — and checked which ones still carried the client's link. Eleven of the thirteen came back between 85% and 100% live, including orders six and seven weeks old. Two came back near 40%.

The interesting part is that age does not explain which. Two orders delivered in the same week scored 95% and 39%.

We are publishing this because we appear to be the only people measuring it, and because we got it wrong the first time. That second part is below too.

The correction, first

Our first pass produced a tidy decay curve: roughly 100% live in week one, falling to 27% by week eight. It was wrong, and the fault was ours rather than the links'.

The checker fetched pages ten at a time on a twelve-second timeout, and counted every timeout as a dead link. Hosts were rate-limiting us, and we were recording our own failures as their removals. The tell was a re-run: the same forty URLs scored 33% live in one pass and 87% in the next. Random sampling cannot do that.

Re-run at five concurrent requests on a twenty-second timeout, with network errors reported separately instead of counted against the link, the same URLs return 95% live and are stable — zero of forty verdicts changed between passes.

Everything below uses the corrected method. If you are checking your own links, this is the part to get right before you trust any number you produce.

How long do backlinks last? The actual data

Across 759 individually checked placements, 88.8% were live.

Two orders are the outliers, and they are not the oldest ones. A six-week-old order scored 95% while another six-week-old order scored 39%. That is not a curve. That is two batches that were worse than the rest, and the age of a batch tells you very little about it.

Do backlinks disappear? Some do — but not the way it is usually described

The failures have a consistent shape. Across all 759 checks there was exactly one hard 404. Every other failure was a page that loaded perfectly and simply no longer contained the client's link.

That matters more than it sounds. It means:

• The pages were real and stayed real.
• Something removed the link specifically — a host cleaning up user-generated content, a moderator, an automated sweep.
• A checker that only reads status codes would have scored every one of those orders at effectively 100%.

If a supplier shows you a "live link report" with green ticks, the only question worth asking is what the tick measures. If it is the HTTP status, it is measuring whether a web server is switched on.

Why do backlinks disappear on these platforms?

Backlink decay on this tier has one cause. Foundation links live on web 2.0 blogs, wikis, profile pages and bookmarking sites — places anyone can publish without approval. That is exactly why they cost a rupee or two: no editor reads the submission, so nothing limits the volume.

The same absence of a gate is why some vanish. Nobody at the host approved your link, so nobody has a stake in keeping it. When a platform runs a clean-up of dormant or low-quality accounts, your link goes with the account.

What our data adds is that this is not inevitable and not uniform. Most batches on these same platforms were still standing weeks later. The two that were not suggest the batch itself — which hosts were used, how the accounts were made — matters far more than the calendar.

Do backlinks expire?

No. There is no clock and no expiry date. A foundation link can sit for years or disappear in a fortnight, depending on the host it landed on and how that host is run. That is why we quote a range from real orders rather than a lifespan.

What actually holds

Here we are reasoning rather than measuring, and it is worth saying so — we do not have a comparable study on editorial placements, so we will not quote you a number for them.

The mechanism is different, though. An editorial link sits inside an article a site owner chose to publish, on a site they are invested in. There is no dormant account for a sweep to catch. That difference is the honest explanation for the price gap between a ₹50 link and a ₹3,000 one: you are largely paying for the absence of this failure mode.

What this should change about how you buy

Check, do not assume — in either direction. The pessimistic story ("cheap links all vanish") and the optimistic one ("our links are permanent") are both wrong. Most held; some did not; only measurement told us which.

Ask for live URLs and open five yourself. Search each page for your own domain. Two minutes, and it is the one claim in this industry that is trivially verifiable.

Judge a supplier by a bad batch, not an average. Our own average is 88.8%, which sounds fine and hides two orders where a customer got well under half of what they paid for. The average is the least useful number here.

Treat foundation links as a layer to maintain, not an asset to own. Whether the loss is 5% or 60%, it is not zero, and a batch bought once and never repeated is a layer that thins.

Frequently asked questions

How long do backlinks last on average? There is no useful average. In our data, most orders were 85–100% live weeks after delivery and two were near 40%, with age a poor predictor. Ask about the specific batch, not the category.

Do backlinks disappear on their own? Not on their own — a person or a script at the host removes them, usually while clearing out inactive or low-quality accounts. Your link is collateral in someone else's cleanup.

Do lost backlinks hurt my rankings? Lost backlinks remove whatever they were contributing. Foundation links contribute diversity rather than ranking power, so the loss is usually smaller than the count suggests — but it is not nothing, which is why the layer needs maintaining.

How do I check my own backlinks are still live? Open the page and search it for your domain. Do five at random from any report you have been given. And if you automate it, keep the concurrency low and treat timeouts as unknown rather than dead — that mistake cost us a whole set of conclusions.

What do you do about it? Every link in a report we send is fetched and checked before the report goes out, so the count you receive is what resolved that day rather than what was ordered. Anything that drops afterwards is replaced free for twelve months. And when our own measurement turns out to be wrong, we republish it — this page is the second version.

Method: 16,012 report rows across 13 delivered orders. 60 placements sampled at random per order, fetched at 5 concurrent requests with a 20-second timeout. A link counts as live only if the page returns a success response and its HTML contains the client's domain; 404 and 410 count as removed; timeouts and 403/429 blocks are recorded as unknown and excluded from the rate rather than counted as failures. First measured 2026-09-02, corrected and re-measured 2026-09-03.