Do more Google impressions mean more clicks? Over 28 days on a small AI-search site, no: 2,884 impressions, 13 clicks, the 7 biggest days earned 1 click, the 7 smallest earned 2

Method, stated before the numbers

On 2026-08-29 at 13:54Z we called the Search Console searchAnalytics.query API for sc-domain:askedabout.com with search type web only, dataState all, over 2026-07-01 to 2026-08-28, four times: dimension date, date × page, date × query, and no dimension for the property total. Discover and Google News rows are not in this cut. Days newer than 2026-08-27 are flagged fresh because Google may still revise them, so every 28-day figure uses complete days only, 2026-07-30 to 2026-08-26. A page counts as shown on a day if it has at least one impression on that day's page row; page-one means that page's position on that day was 10 or better; position is Google's impression-weighted average. Clicks and impressions per day are the property-level row, which is what Google reports in the Performance chart. The rank correlation is Spearman's, over the 28 daily pairs. The script `content/news/gsc_daily_rows_cut.mjs` writes the artifact `gsc-daily-rows-cut-2026-08-29.json` with 12 self-checks (per-day clicks and impressions sum to the property total, every window day present once, click days plus zero days equal 28, streak spans contain only zero-click days, and so on); all 12 pass. Small numerator, stated up front: 13 clicks over 28 days is few enough that a rank correlation is a description of this window, not an estimate of anything beyond it.

Biggest days against smallest days

The seven biggest days carried 1,090 impressions and 1 click; the seven smallest carried 377 and 2. The nine days with a click averaged 97.4 impressions (median 103); the nineteen days without averaged 105.6 (median 96). Page-one impressions on click days averaged 30.8 against 34.9 on zero days; pages shown averaged 29.9 against 28.7. Spearman rank correlation, daily impressions against daily clicks: -0.06. Page-one impressions against clicks: 0.08. Where the big days' impressions came from: on 2026-08-21, the record, 123 of 199 page-level impressions were /compare pages and 39 were /guide pages, at positions in the 60s and 70s; the /news posts that hold 12 of the 13 clicks contributed 41. The biggest seven averaged position 52.6 and 34.3 pages shown; the smallest seven averaged 42.2 and 22.0. More impressions on this site has meant more deep-ranked pages on screen, not more people at the top of the page.

The 28 complete days, one row each

Zero-click streaks since 2026-07-01

The 29 complete days before the window, 2026-07-01 to 2026-07-29, held 1,010 impressions, 1 click and 28 zero-click days; the longest streak, 19 days from 2026-07-01, ran while the site was being shown 477 times. Inside the 28-day window the streaks are shorter, four days at most, but they sit on the impression highs: the three-day streak from 2026-08-21 spans 462 impressions and the record day; the three-day streak from 2026-08-25 spans 456 and reaches into the fresh days. The two fresh days, outside every 28-day figure and still revisable, read 2026-08-27: 162 impressions, 0 clicks, position 58.1; 2026-08-28: 122 impressions, 1 click, position 47.4.

Where the 13 clicks came from

Twelve of the thirteen clicks landed on /news posts, one on the homepage; /compare and /guide, which supplied 1,337 page-level impressions between them, supplied none. Six of the twelve /news clicks went to one post, the one on whether the Search Console API exposes generative-AI data, on four separate days at positions between 4.3 and 11.8. Nine of the thirteen clicks came from page-days at position 10 or better, which carried 940 page-level impressions. The date × query table discloses a query for 0 of the 13 clicks: every clicking query sits in the anonymised remainder Google withholds, so this cut cannot say what anyone searched, only which page they chose and where it ranked that day.

What this does and does not show

• It shows that on this site, in this window, the days with the most impressions were not the days with clicks. It does not show a negative effect: -0.06 on 13 clicks over 28 days is the sign of a coin toss, and we state it as a description of these days, nothing more.
• Property-level impressions (2,884) and page-level impressions differ because one query can show two of our URLs at once: over the full 07-01 to 08-28 window the page table sums to 103.9% of the property total. The class table and the page-one figures are page-level; the daily rows are property-level; the artifact's self-checks state which is which.
• The impression growth is coming from pages ranked in the 60s and 70s (/compare 800 impressions at 0 clicks, /guide 537 at 0), not from the page-one posts. A bigger Google day for this site currently means Google tried more of our pages deeper in the results, which is consistent with the clicks not moving.
• Zero disclosed queries on 13 clicks is a property of Google's anonymisation at this volume, not of our pages. We cannot test title changes against the queries that clicked, because none is visible; the weekly title work runs against the queries Google discloses on impressions instead.
• Web search only. Discover and Google News, which Google reports separately, are outside this cut; a Discover impression day would not appear in these rows.
• Fresh days are excluded from every 28-day number and named separately. Google's own documentation says recent days may be revised; the 2026-08-28 click may still be added to, or removed.

Pre-committed read, 2026-09-26

On 2026-09-26 we run the same script over the next 28 complete days, 2026-08-27 to 2026-09-23, and publish the result whichever way it falls. (A) If the Spearman correlation between daily impressions and daily clicks reads +0.30 or below, the finding stands: a bigger impression day did not mean more clicks. Above +0.30 falsifies it. (B) If the seven biggest-impression days earn 40% or less of the period's clicks (seven of 28 days would take 25% by chance), volume is still not converting; above 40% and the deep impressions have started turning into visits. (C) If 14 or more of the 28 days are zero-click days, the daily pattern held; 13 or fewer means it changed. The artifact for that read will carry the same 12 self-checks.