# Three months of Search Console on an essay-led site Over equal 99-day windows before and after this site's rebuild, Google impressions rose from 118 to 1,263 and clicks from 3 to 11. Eight reference pages earned 45% of post impressions and 23 short logs earned 25%. Only 17% of impressions came with a visible query, and 16 of 29 new posts got their first impression within a day of going live. Published: 2026-11-04 Canonical: https://umar.codes/three-months-of-search-console ## Revisions - 2026-11-04 — Created from the Search Console API, queried that day by date, page, query, device and country for 2025-07-01 to 2026-11-02, and from the URL Inspection results of 2026-11-02. - 2026-11-04 — Published on its Nov 4 slot, the day its three-month gate matured; auto-drafted by the publishing run because no draft existed for the row. --- In the 99 days after I rebuilt this site, Google showed it 1,263 times and sent 11 clicks. In the 99 days before, it showed it 118 times and sent 3. Impressions grew more than ten times; clicks did not keep up. Nearly half of the impressions on posts went to eight reference pages, not to the 23 logs. The rebuild went live on 27 July 2026. Search Console data lags about two days, so the window here ends on 2 November: 99 days, which is a little more than a quarter. I queried the Search Console API on 4 November by date, page, query, device and country, for the whole range Google keeps, which starts in July 2025. One detail matters before the numbers. Site totals in this post are property totals. Per-page numbers are page totals, and they add up to more, because Google counts a search once for the site even when two of my pages appear in it. Over the full range that is 1,660 impressions for the site and 1,803 across pages. ## Impressions grew ten times after the rebuild, and clicks did not | 99 days | Impressions | Clicks | CTR | Average position | |---|---|---|---|---| | 19 Apr – 26 Jul 2026 (old site) | 118 | 3 | 2.5% | 4.1 | | 27 Jul – 2 Nov 2026 (rebuild) | 1,263 | 11 | 0.9% | 14.4 | The whole year before the rebuild had 381 impressions and 16 clicks. The rebuilt site passed that year's impressions in its first 37 days. Average position got worse, from 4.1 to 14.4, and that is not a sign that the pages got worse. The old site was mostly a homepage: 307 of its 421 page impressions since July 2025 were on it. Its one visible query with volume was my name, with 39 impressions at position 1.8, all before the rebuild. The rebuilt site turns up for topic searches, where it sits on page two. A worse average here means more kinds of search, not worse results for the same ones. The 11 clicks are spread over 10 pages, and no page has more than 2. All 11 came from desktop, which also had 1,134 of the 1,263 impressions. ## Two and a half times the posts, 38% more impressions The site went from 6 posts on launch day to 40 three months later. Impressions did not follow at the same rate. | 28 days | Posts live at the end | Impressions | |---|---|---| | 3 – 30 Aug | 16 | 339 | | 31 Aug – 27 Sep | 25 | 349 | | 28 Sep – 25 Oct | 37 | 451 | | 6 Oct – 2 Nov (latest) | 40 | 467 | From the first window to the latest one, posts went up 150% and impressions went up 38%. More pages do add impressions, but each new page has added much less than the pages before it. The next table shows why. ## Reference pages earn impressions, and most logs do not The site has three kinds of post. A reference is a living page that I revise for years, an essay is an argument, and a log is a short note about one finding. Most posts in this quarter were logs. | Kind | Posts | Median days live | Impressions | Per post per day | Median, posts live 28+ days | |---|---|---|---|---|---| | Reference | 8 | 68 | 581 | 1.20 | 115 (5 posts) | | Essay | 9 | 78 | 382 | 0.57 | 25 (9 posts) | | Log | 23 | 36 | 317 | 0.32 | 7 (14 posts) | Logs are younger, so I controlled for age two ways. Per post per day, a reference earns almost four times what a log does. Among posts live for at least 28 days, the median reference has 115 impressions and the median log has 7. The single largest page, [testing Claude skills](/testing-claude-skills), has 235 impressions; [sdx](/sdx) has 142. Both are references from launch day. A log answers one narrow question. That makes it easy to write and easy to cite, but few people search for exactly that question. A reference answers the broad question that the search was about. ## Only 17% of impressions came with a visible query Google hides queries that few people search for. On a small site, that is most of them. In the 99 days, 215 of the 1,263 impressions came with a query I can see, across 60 queries. None of the 11 clicks came with a visible query. The one query with real volume is "astro to next.js conversion prompt": 73 impressions at position 12.8, no clicks. I wrote [the Next.js to Astro conversion log](/next-to-astro-conversion) because of that query, and that page now has 94 impressions. The query asks for the other direction, from Astro to Next.js, which may be why page two of the results has not earned a click from it. The query report on a site this size is a sample, and a small one. I treat any query under 10 impressions as noise when I decide what to write. ## Requesting indexing got most new posts seen within a day 29 posts published since the rebuild have at least one impression. For 16 of them the first impression came on the day they went live or the day after, and the median is one day. I requested indexing through URL Inspection for almost every one of them, most on the day I published it. The exceptions show what the request does and does not do: - **Three launch-week posts waited 13 to 17 days.** Their impressions started on or after 9 August, the day I requested indexing for them. - **[The AI crawler essay](/what-ai-crawlers-read) waited 23 days**, although I requested it on publish day. Then it took 86 impressions in two days, at position 21 to 25: on 10 and 11 October the site had 106 impressions, 23% of the latest 28-day total. - **[The widget post](/widget-in-someone-elses-page) took two requests.** It was still "Discovered – currently not indexed" 38 days after the first. Google crawled and indexed it 19 days after the second. - **[The URL state comparison](/url-state-in-react) is "Crawled – currently not indexed".** It was requested and crawled on publish day and not crawled again in the next 26 days. That is Google's choice, not a technical block. - **[One post was never requested](/optimistic-concurrency).** Its first impression came one day after publication, and it has 52 impressions. One data point does not prove that the request no longer matters. Nine of the 40 posts have no impressions. Four of them were indexed when I checked on 2 November, one is the URL state page above, and four went live between 21 and 30 October, which is too recent to read. An indexed page with no impressions is a demand problem, not an indexing one. ## Against a target of 10,000 impressions a quarter The publishing plan set a target of 10,000 impressions a quarter. The site had 1,263 in 99 days, about 13% of it. In September I wrote down the three things that could close the gap: more indexed pages, topics that people search for, and links from other sites. Three months of data says this about each: - **More pages:** 150% more posts gave 38% more impressions. - **Topics with search demand:** I published six posts from the September demand pass. Two of them have more than 10 impressions: 94 and 22. - **Links from other sites:** still the open item. I have not measured it, because I have not changed it. The plan does not change because of this post. The year-end essay compares the three publishing cadences of this quarter with the data that will exist by then. ## What Search Console cannot show on a site this size - **Most queries.** 83% of impressions arrive without one. - **CTR per page.** With 11 clicks in 99 days and no page above 2, a CTR for any one page rests on one or two clicks. - **AI features.** The API's search appearance breakdown returned no rows for this site, so I cannot separate AI feature impressions from normal results. - **Anything after 2 November.** The data lags two days, and posts from the last two weeks of October are still too new to judge. What it does show is enough for the decisions I make with it: which pages Google knows about, which kind of page earns impressions, and how far the site is from its target.