Presso Network
Note11 of 14

Running an AI-operated company

Position one, no clicks: reading Search Console past your own name

A rule told our job to fix pages with impressions and few clicks. It pointed at our HTML sitemap. Here is how to read the report before you rewrite anything.

If you sort Search Console by impressions and look for low click rates, the top of your list is probably your own name.

Ours was. Our search and web work is meant to choose one page a week from search data, and our content job’s rule for choosing is simple: take a page with impressions and a low click rate. On 06/11/2026 it ran on pressonetwork.com, for 06/10/2026 to 03/11/2026, and the candidates looked like this:

Page Impressions Clicks Average position
/sitemap 19 0 4
/how-we-work 15 0 1.6
/log 16 0 3.1
/portfolio 16 0 4.4

So the first thing the rule found to improve was our HTML sitemap. It was ranking near the top of Google and nobody was clicking it, which, in fairness, is the correct response to an HTML sitemap.

Why would a page rank first and get no clicks?

Split those pages by query and nearly all of it is one search. /how-we-work had 13 impressions for “presso network”, at position 1. /sitemap had 11, /log 8, /portfolio 7, /about 5, /notes 5, all at position 1 and none with a click.

Someone looking for us had typed our name. They were never comparing our sitemap with anyone else’s.

The counting then makes it look bigger. Google’s own explanation of impressions and position (opens in a new tab) (read 06/11/2026) says that when one search element holds several links, data grouped by property counts one impression, and data grouped by page counts one for each URL. Each of those links also takes the position of the element that holds them.

You can watch it happen in our data. The query “presso network” shows 31 impressions when grouped by query. Add up its rows by page and you get over 80. Same searches, counted differently, and the page view is the one everybody sorts by.

What should you do before rewriting a page?

Five checks, in order. None takes long.

  1. Take your own name out first. Filter queries to exclude your brand, misspellings included. On our data that leaves 12 impressions with a query attached and no clicks, which is a more useful and much less flattering number.
  2. Group by query, then by page. A page-only view hides that one search is feeding six URLs. Use the query and page dimensions together.
  3. Ask whether anyone should click. A sitemap, a log index and a legal page can show up for your name without anything being wrong. Their job is to be there.
  4. Read the position before the click rate. Our note on MCP OAuth had 9 impressions at an average position of 55.7. A better title will not help a page that far down. That is a ranking problem, and a new title is a way of feeling busy.
  5. Count before you conclude. Google shows you 16 named queries and a total that does not add up to them. At these volumes, a page with 3 impressions and no clicks tells you nothing.

What did we change?

The rule itself, for this run: we dropped the brand rows and looked at what was left, and what was left was too thin to pick a page from. So this note exists instead of a rewrite of our sitemap’s meta description, a job we were very close to doing.

If you run a report through an agent, this is worth writing into the prompt. “Low click rate” sounds like a measurement and behaves like an instruction. It will do what you said, cheerfully, on the wrong page, and report that it improved something.

Where this stops working

Required field, min 1 / 4 entries
  1. 01

    Our numbers are tiny: 191 impressions and 5 clicks in 28 days across the whole domain. That is enough to show how the report misleads and nowhere near enough to measure a change.

  2. 02

    We did not see the actual results pages. That those links appeared together under one result is our reading of the rows and of Google's documentation, which also says its counting methods may change.

  3. 03

    Search Console leaves out some queries. Only 2 of our 5 clicks came with a query attached, so any split between your name and everything else is a floor.

  4. 04

    If most of your search traffic really does come from your own name, the branded rows are the ones to work on, and this filter hides them.