Seventy Five Impressions in Eleven Weeks

01The number I kept not looking at

For eleven weeks I counted the machines that read this blog. Crawlers, live fetchers, agents, their visits by the day. I did not look hard at the one number that decides whether any of that reading can turn into a citation: how often a search engine actually showed a page from here to a person. From the first entry on 25 July to 9 October, Search Console recorded 75 impressions for the whole domain. Not per day. In total. Seventy five impressions in eleven weeks is not a traffic problem. It is a visibility problem, and visibility comes first.

Where the 75 went

There were two clicks, both on the same day, both to the homepage, both from one country, at an average position of 1.7, which is what a search for a person's name looks like. The blog index collected 11 impressions. One entry, the deindexing recovery story, collected 10. The topic hub about the agent web was shown 5 times at an average position of 68.8, page seven, where nobody looks. Of the 63 addresses in the sitemap, 26 were shown at least once. The other 37 were never shown to anyone.

02Two theories, both wrong

It was not the country code

My working theory was that the domain itself held the blog back. A country code domain is tied by Google to that country, and Search Console offers no setting to change it. If that were the wall, the few impressions would cluster in one place. They do not. The United States produced 27, the United Kingdom 10, Romania 9, the Netherlands 4, and the country on the domain 3. When a page is shown almost nowhere, the country it is not shown in stops being the explanation. The code may cost something at the margin. It is not why 37 pages were never seen.

It was not the crawlers

The other theory, that machines cannot reach the blog, was already dead. In the fourteen days to 8 October, Bing's crawler fetched entries about 344 times, Amazon's about 94 and GPTBot about 20. Bing then showed the domain 3 times in roughly six months, and 2 of those 3 were searches for the domain itself. Being read and being shown are separate gates, the split I described in indexed, not quoted. Every machine reads. Almost none presents.

One honest limit. Google attaches a visible query to only 7 of the 75 impressions, things like agent web and get cited. It hides the rest as too rare to report. So I cannot say which questions the other 68 answered, only that there were very few of them, and that the visible ones sat far down the results.

03Measure the first gate first

Count pages shown, not pages published

The useful number is not total impressions. It is how many of your addresses were shown at least once in a period, divided by how many you published. Here that is 26 of 63, or 41%. Coverage, the share of your pages shown at least once, is the number to watch before clicks and long before citations. In Search Console open Performance, choose Pages, set the range, export, count the rows and compare them with your sitemap. It takes ten minutes and it tells you whether your writing has a chance of being retrieved at all. A page that is never shown cannot be quoted, however well it is written.

The same check on the agency site I run with my brother, over the same eleven weeks, returned 3,432 impressions, about 46 times as many, from a domain that is older, has links pointing at it and publishes how to guides people actually search for. Same author, same server, same habits. The difference is the domain and what points at it, not the prose.

What it means for this archive

I wrote in sixty eight entries in that being findable comes before being quoted. This is the measurement behind that sentence. Writing more entries into a domain shown 75 times in eleven weeks mostly adds pages to the 37, not to the 26. That is why the how to entries already go to the agency site, and why this one stays: it is a record, and records belong in the first person.

The machines will keep reading the archive whatever search does with it, and that has its own value, because training crawlers do not need a ranking to store a page. But I no longer expect search to find it unaided. An archive the machines read but search never shows is a library with no door on the street. The door is links and mentions from places people already visit, which is the slow work behind pages you do not own.

Tomorrow: an email offering to swap links, and why I said no.