1,301,839 links. 22,260 referring domains. Not one of them bought. Here is what the data says about how long a backlink actually lasts — and what it says about my own site.
I have been publishing on this domain since around 2006. Over that time it has accumulated 1,301,839 backlinks from 22,260 referring domains, and I have never bought a single one of them. Every link is natural, with the exception of my own properties linking here.
Last week I downloaded the complete Majestic export — both the Fresh and the Historic index — and analyzed all 1.3 million rows. I did it to build teaching material for a link building course I am putting together. What came out was more interesting than I expected, and some of it was uncomfortable.
Here is what twenty years of unbought links actually looks like.
Half of my referring domains stopped linking within three years
The Historic export records, for every link, the date it was first indexed and the date it was lost. That is enough data to build a proper survival curve rather than an average, and the distinction matters: averaging the age of links that died ignores every link still working, which understates lifespan badly.
So I ran a Kaplan–Meier analysis. One observation per referring domain, with links still alive censored at their last-seen date rather than counted as either survivors or losses.
Median survival: 1,080 days. Just under three years. After one year, 80.9% of referring domains are still linking. After three, 49.3%. After ten, 9.5%.
Across the whole profile, 49.6% of every link ever recorded is now flagged lost. That is not a failure of anything. It is what happens to links.
The practical version of that number is a replacement rate. If half your referring domains are gone in three years, standing still means acquiring roughly a sixth of your profile every year before you have grown by anything at all. A link building program that stops is not a program that plateaus. It reverses.
Links from strong domains last about four times longer
This is the finding I had not seen published anywhere, and it is the one I would most like other people to test on their own data.
Segmenting the survival curve by the Trust Flow of the referring domain produces a clean, monotonic relationship:
| Referring domain | Domains | Median life | Still linking at 5 yrs |
| Trust Flow 0 | 15,421 | 859 days | 22.2% |
| Trust Flow 1–10 | 4,031 | 1,531 days | 40.1% |
| Trust Flow 11–20 | 1,318 | 1,911 days | 52.6% |
| Trust Flow 21–30 | 647 | 2,223 days | 59.5% |
| Trust Flow 31–40 | 397 | 2,375 days | 59.9% |
| Trust Flow 41–60 | 323 | 2,555 days | 60.8% |
| Trust Flow 61+ | 123 | 3,353 days | 69.9% |
A link from a Trust Flow 61+ domain lasts a median of 9.2 years. A link from a Trust Flow 0 domain lasts 2.4. That is close to a four-fold difference in how long the asset exists.
We already compare links on the authority they pass today. This says a strong link is also a longer-lived asset, which means the gap in lifetime value is wider than the gap in Trust Flow alone suggests. It also means cheap links are cheap twice over: weaker while they last, and lasting less than half as long.
A methodological trap I nearly fell into
Run the same analysis per link instead of per domain and the relationship inverts. Trust Flow 61+ comes out with a median of 851 days — the worst band on the board.
That result is an artifact, and a very believable one. A handful of large sites contributed tens of thousands of sitewide links each. When one of them re-templates, tens of thousands of observations die on the same day and drag the whole curve down with them. The link-weighted view is measuring a few CMS migrations, not link durability.
Weighting by domain fixes it. Same data, computed honestly twice, opposite answers. If you take one thing from this post, take that one.
Links do not die because pages die
Ask anyone in this industry why backlinks disappear and you will hear that pages die — sites shut down, posts get deleted, 404s accumulate. I would have said the same thing before I looked.
Of my 645,202 lost links, here is what the source page was doing at the moment the link was lost:
| What the source page did | Links | Share |
| Downloaded fine — link simply no longer on it | 503,390 | 78.0% |
| Now canonicalizes somewhere else | 124,861 | 19.4% |
| 301 permanent redirect | 11,739 | 1.8% |
| 302 / 307 temporary redirect | 4,814 | 0.7% |
| 404 Not Found | 4 | 0.0006% |
| Dead host, DNS failure or timeout | 24 | 0.004% |
97.4% of my lost links were lost while their source page was still perfectly reachable. Four links out of 645,202 were lost to a 404.
Links die because somebody edited the page, or a redesign dropped the link, or a migration changed the canonical. The page carries on living without you in it.
Three things follow from that, and all three are actionable:
- Link reclamation is far more viable than most people assume. The page still exists, it still ranks, and there is a live human maintaining it. That is a warm prospect, and a completely different conversation from broken link building.
- Watching for 404s to catch link loss catches essentially nothing. In my case it would have caught four links in twenty years.
- Your biggest exposure is other people’s redesigns, which you cannot predict and cannot control. Which is the argument for a replacement rate rather than a one-time campaign.
Seven in ten of my referring domains are Trust Flow 0
I want to be precise about this, because it is the number I think will be most useful to other people.
Of the 22,260 domains that have ever linked here, 15,421 — 69.3% — are Trust Flow 0. Scrapers, aggregators, expired-domain churn, auto-generated directories, comment spam, sites that copied a post wholesale.
I did not buy any of them. I did not build any of them. They showed up on their own, because that is what happens when you publish on the internet for two decades.
If you have run a backlink audit and found thousands of junk domains in your profile, this is your calibration point. Junk links are the default state of having existed online. They are not evidence that something went wrong, that a competitor attacked you, or that a previous agency did something regrettable.
The same table calibrates something else. In twenty years this domain accumulated 446 referring domains at Trust Flow 41 or above — roughly 22 a year. When somebody offers you fifty TF 50+ links in a quarter, they are offering more than double what this site earned annually, at higher quality, in a tenth of the time. That is not aggressive. That is not a thing that occurs.
One sitewide link is 16.8% of my entire profile
The single largest referring domain in the dataset sent 219,159 links. Of those, 218,030 carry the identical anchor text, pointing at the identical URL — a post I wrote called “In a Post Google Penguin World, It Is Still Okay To Link Out.”
One post. One sitewide link. One person’s editorial decision, years ago. Nearly a sixth of a 1.3 million link profile.
Widen it out and the top 100 domains — out of 22,260 — account for 84.3% of every link ever recorded here. This is why you count referring domains and not links, and it is why I am always sceptical when someone quotes a backlink number without saying how many domains it came from.
What I found wrong with my own site
I said some of this was uncomfortable. Here is that part.
My own domain hartzer.net was once the third-largest referring domain in this entire profile, with 14,194 links. It now has zero live links pointing here. That is the single largest recoverable loss in the dataset and it is entirely my own fault.
Two cross-links from this site to hartzer.com have gone dead in the last three months — one on “buying brand mentions” and one on “seo expert witness.” I own both ends of both links. I should probably fix those links.
Three of my commercial service pages have collectively lost 20,748 links, which usually means a URL changed and nobody put the redirect in.
And 20,768 links that are flagged live resolve to connection failures, DNS failures or timeouts when Majestic.com tries to fetch the target. My working hypothesis is that my host or WAF is intermittently blocking their crawler, which would mean every link tool in the industry is seeing a degraded picture of this site. I need to check the server logs before I can say that for certain.
There is also a pleasant surprise. An image I uploaded in 2013 — a chart about keyword “(not provided)” — has 3,200 live links pointing at it. I had no idea.
The full case study, and what I am building with it
Everything here is a summary. The complete case study — all the figures, the technical audit, the interactive survival curve, and the full list of what I need to fix — is going up on AdvancedLinkTraining.com, a free link building course I am building.
The course is nine modules and fifty lessons, self-paced, with no signup and no email gate. This dataset is the worked example that runs through the module on reading the link graph, because I would rather teach from a real profile with real problems than from a hypothetical.
If you have a Majestic.com subscription and a domain with some history, I would genuinely like to see whether the Trust Flow survival relationship holds on your data. One profile is an observation. Two would be a finding.
Methodology note: figures are computed from the Majestic Historic and Fresh exports for billhartzer.com, downloaded 2 August 2026. The Historic export returned 1,301,839 rows against the 1,382,639 links the Summary screen reported, so all figures are computed on what the export contained. Trust Flow is measured at export time rather than at the moment each link was acquired, which will bias the low bands’ survival downward by an unknown amount — that is the most important limitation of the segmentation above.