Search Console Impressions Spiked, CTR Fell: What I Check

A spike in impressions with a falling CTR is not a ranking win and not a penalty. It’s one page being shown to a lot of people who mostly won’t click, and your sitewide averages hide which page.

Here is the real case. I pulled my own Search Console data for the last 28 days (10 October to 7 November 2026): 1,160,544 impressions, 7,865 clicks, 0.68% CTR, average position 11.4. Then I looked at one day.

What the spike looked like

Read in isolation, that is contradictory: better position, more clicks, and a worse CTR. Nothing about my content changed that day. Average position most likely improved because a very large batch of impressions arrived at position 8, which pulled the average up. That is not a ranking gain on the pages that earn my clicks.

The three-step read

1. Split by page. Over the 28 days, my SEO tools list alone had 599,884 impressions, 48 clicks and a 0.01% CTR at position 10.2. One page, about half of all site impressions, almost none of the clicks. That is the whole story of the sitewide CTR.
2. Split by query. On 28 October the query “seo tools” showed 4,366 impressions at position 8 and 4 clicks. A head term with a huge audience, where a results page full of tool brands beats a list page for most people.
3. Decide if you can win it. If the impressions come from a query your page doesn’t actually serve, rewriting the title won’t help. I wrote about that split in why your best-ranking pages get zero clicks. If the query does fit, follow the playbook in how to fix pages stuck on page two.

I haven’t confirmed why Google surfaced that page so widely on that day, and I’m not going to guess in public. The useful move is the same either way: stop reading the sitewide line and read the pages.

What not to do

• Don’t rewrite titles sitewide because the CTR line dipped. The dip came from one page.
• Don’t report the average-position improvement to a client as a win. Position got better while nothing earned it.
• Don’t compare a spike day to a quiet day and call the difference a trend. Check at least 14 days either side.

The prompt I paste into Claude or ChatGPT

Export the Pages tab and the Queries tab from the Performance report as two CSVs for the 28 days around the spike, then paste both:

Here are two Google Search Console exports for the same 28 days:
one by page, one by query. Impressions jumped on one day and CTR fell.

1. Tell me which single page contributed the most impressions and what
   share of the total that is.
2. For that page, list the queries driving the impressions, with position
   and CTR, and flag which are head terms I'm unlikely to win.
3. Tell me whether the average-position change is explained by that page
   alone. Show the arithmetic.
Do not suggest title rewrites until you've answered 1 to 3. If a number
isn't in the data, say so instead of estimating.

Check the arithmetic yourself on one row before trusting the rest. These models will round confidently and sometimes get the share wrong.

The numbers above came from Search Console and GA4 for this site. GA4 also rated the same period’s traffic quality as mixed, with 12 seconds average engagement, so I treat sessions there with care and lean on Search Console clicks instead.