How many Google impressions before a comparison page gets its first click? On our 10 pages: 2,236 impressions over 84 days at average position 78, then 2 clicks on one page 16 hours apart

The question and the instrument

"Best X alternatives" pages are a common early SEO asset for a software product. We have 10 of them. The practical question for anyone who publishes one on a new domain is how long it sits in Search Console collecting impressions before the first click, and what that first click looks like when it comes. Ours came on 2026-09-17, so the wait can now be counted.

The Search Console side is the API's day-by-page table for web search from the first impression of any page on the site to the read, with fresh data included. Days are Pacific time, as the API reference states, and our script flags the last three days as provisional because Google can still revise them. "Position" is Google's average for a page on a day, weighted here by impressions when days are combined. The on-site side is our own event tape: every page view of a comparison page whose referrer is a Google search domain, with the page's own reports of scroll depth and seconds in the foreground. The artifact holds no IP addresses, places or query strings.

Impressions before the first click, by page type

The comparison pages took 17 times the impressions and 7 times the days that the news posts took. The difference that the table can show is position: 13.8 for the news posts, 77.5 for the comparison pages. Position 78 is the eighth page of results at ten to a page.

Where the comparison-page impressions sat

2,154 of 2,287 impressions (94%) came on days when the page averaged between 51 and 100. The pages did reach page one: 30 day-and-page rows average 10 or better, holding 48 impressions and 0 clicks. Both clicks sit in the 51 to 100 row. The first page-one row is dated 2026-06-28, 81 days before the first click.

The clicked page is the most-shown of the ten and the worst-positioned of the large ones: 861 impressions at 85.7, first shown on 2026-08-08. Counted alone it waited 40 days and 842 impressions. The page with the most page-one days (10) has 59 impressions and no click.

The two click days

On both days the click belongs to no query row. Google's own explanation of the performance data says why that can happen: "While the actual anonymized queries are always omitted from the tables, they are included in chart totals, unless you filter by query." The page row is a total, so the click is in it; the query that produced it was too rare or too personal to list. Over its life the page has 52 named queries carrying 794 of its 861 impressions, and none of them has ever produced a click.

The last column is arithmetic on Google's averages: the page row's position times its impressions, minus the same product over the named query rows, divided by the unnamed impressions. On 2026-09-18 there is 1 unnamed impression and the click must be on it, which puts that click at about position 76. On 2026-09-17 there are 2 unnamed impressions averaging 39, and the click is on one of them. Both days fall inside the window Google may still revise, so these two figures can move.

What the two readers did

The tape holds 2 Google-referred page views of any comparison page since the site launched, and they match Search Console's 2 clicks day for day. They share one IP address, one browser build, one window size and one time zone, and they carry two different analytics identities and two device ids. That is one person in two browser profiles or with cleared storage, or two people on one network; the tape cannot tell which. It is one source of demand, not two.

Both loads reported a scroll depth of 64% within 852 milliseconds of the page view. At that window size the page is 9,533 pixels tall and its first screen ends at 8%, so the page did not open at the top: the window opened about 5,313 pixels down, with the sixth of 7 section headings on screen. A reader cannot scroll that far in under a second, and both loads reported the same depth. A browser opens a page part-way down when the link carries an anchor or a text fragment, a link to a quoted passage. The section headings on this page carry 0 anchors, which leaves a text fragment. We cannot confirm that from the page: MDN's reference says the fragment instructions "are stripped from the URL during loading so that author scripts cannot directly interact with them." So it is a reading of the scroll data, stated as one.

The readers stayed 11 and 12 seconds. The first did not scroll; the second went on to 82%. The page's check panel starts at pixel 8,145, below where either window ended, so neither reader had a call to action on screen at any point. Nothing placed in a page's first screen is seen by a reader whose window opens two-thirds of the way down.

Reading

• One site and two clicks: these are counts. On this site, at an average position near 78, a set of 10 comparison pages needed 2,236 impressions and 84 days for a first click. News posts at position 14 needed 130 and 12. The impression count in Search Console says little about how close a first click is; the position column says more.
• The first click did not come from the pages or queries that had reached page one. It came on the deepest-ranked large page, on an impression Google names no query for. A query report filtered to named queries would never have shown it.
• Two clicks 16.1 hours apart from one network are one reader's or one office's interest in one topic. They do not show that the page class has started to rank.
• Both loads opened mid-page and ended in under 13 seconds with no call to action seen. A reader sent to a passage sees only what sits near that passage.

The next read, pre-registered

1. 1On 2026-10-03: Search Console clicks on the comparison pages for 2026-09-19 to 2026-10-01. Statement: 2 or fewer. Three or more and the class has started to earn clicks at its current position.
2. 2On the same date: the next Google-referred landing on a comparison page. Statement: its first scroll report is deeper than 15%, meaning it also opened below the first screen. A landing that opens at the top counts against the passage-link reading.
3. 3On the same date: the clicks on the two click days above, re-read after Google's revision window closes. Statement: both remain, and both remain on impressions with no query named.

A result outside its statement is published as a falsification with the same script and a dated artifact.

Artifact

`content/news/compare-first-click-cut-2026-09-19.json` holds the per-type first-impression and first-click days, the ten comparison pages, the position buckets, the click days with the named-query and unnamed-impression split, the two landings, the page geometry at the readers' window size and 13 self-checks, among them that Google-referred landings on the tape equal Search Console clicks day for day. The script is `content/news/compare_first_click_cut.mjs`; this post is rendered from the artifact by `compare_first_click_post.mjs`. The first version of the landing query matched any referrer containing "google." and counted a Gmail-app visit from August as a search landing; the day-for-day self-check failed on it and the filter now matches Google search domains only.