We put a statistic on our own homepage, then took it back down

Google will tell you which of your pages it found linking to any given page of yours. It tells you one page at a time, and it keeps no history. Ask again next month and you get next month's answer with no record of this month's.

We keep it. So we can ask something Search Console cannot answer: are the pages nothing links to less likely to end up in Google?

We have now answered that question wrong twice. Here is the number we published, the two flaws we found in it, and what is left.

The number we put on our own homepage

Across every site we monitor, counting pages we had checked: 96% of the pages with something linking to them were in Google, against 58% of the pages with nothing.

A 38 point gap. It confirms what everyone in SEO already believes, it is easy to say out loud, and we put it on the front page.

First flaw: the sites disagreed with each other

Asking the same question one site at a time instead of throwing them all in together, the pattern held on some and reversed on others. That was enough to add the sample size and a caveat to the homepage, which felt like diligence at the time.

It was not the real problem.

Second flaw: half the evidence was circular

Google only reports links for pages it has actually looked at. It cannot tell you what points at a page it has never crawled, because it has not gone looking.

So every page Google had not yet discovered was recorded as having nothing linking to it, and every one of those pages was also not indexed. Both facts follow automatically from the same cause. Putting them on opposite sides of a comparison measures nothing at all.

We noticed because of our own site. We published a batch of new pages, linked every one of them from the navigation and the footer within the hour, and our own dashboard reported all of them as having nothing pointing at them. That is not what was happening. Google simply had not been back yet.

What the number becomes

Thirty-eight points becomes eight. And eight points across a couple of hundred pages is not a finding, it is noise with a direction. Site by site, corrected:

Two of the four now go the other way, and on the agency site the unlinked pages are indexed 25 points more often than the linked ones. We cannot show what we said we had shown.

What is still worth doing

Look at a page nothing links to. Not because of a percentage, but because of what the situation means: if nothing on your site points at a page, the only reason Google knows it exists is that you listed it in a sitemap. It has no signal about whether the page matters beyond your say-so. That is a weaker position than a page your own navigation vouches for, and checking costs nothing.

What we no longer claim is a rate. We do not know what share of your unlinked pages will end up indexed. Neither does anyone quoting an industry average at you, and now you know one specific way those averages get inflated.

The lesson we actually paid for

Both flaws had the same shape. We compared groups that were different for a reason we had not accounted for, and got a clean number that pointed the way we already expected. The first correction made the number look more responsible without making it more true. The second one was the one that mattered, and we only found it because the tool told us something about our own site that we knew was false.

That is a decent argument for using your own product on your own site, and a better one for publishing the method next to the number, because this article is what that looks like when the method turns out to be wrong.

Your own answer

The useful output was never our average. It is the same question asked of your site, where the comparison is between your own pages. We now exclude never-crawled pages from that count, so "nothing links to this" means we checked and there is genuinely nothing, rather than Google not having been round yet.

Connect a property to see yours, or run a single page through the free checker first.