PolishPal redesign: what happened to search traffic
Clicks up about 14 times, impressions up about 100 times, and a Google spam update that started seven and a half hours before the merge.
In short
- In the six weeks after the 24 September 2026 redesign, Google clicks went from 87 to 1,231 and impressions from 1,347 to 134,134.
- Google's September 2026 spam update started about seven and a half hours before the merge, so the two effects can't be separated.
- No new articles were published between 20 August and 29 October: the same roughly 100 posts moved from Google's pages three to six to around position 6–7.
- CTR fell from 6.5% to 0.9% because the site went from brand searches to thousands of non-brand queries at positions 5–10.
- The honest reading: the redesign did not hurt, and how much of the jump it caused is unknown.
On 24 September 2026 I merged a full redesign of PolishPal, the free Polish course I build and run. In the six weeks that followed, clicks from Google Search went up about 14 times and impressions about 100 times. The easy story would be "new design, more traffic". The data doesn't support that story on its own.
About seven and a half hours before the merge, Google began rolling out its September 2026 spam update. This post sets out what the redesign changed, what the numbers did, everything else that changed in the same weeks, and what I think is the honest reading. The product itself is on the PolishPal project page.
What the redesign changed#
The redesign was one pull request (#2), merged at 23:45 UTC on 24 September. The main commit touched 87 files, adding 4,747 lines and removing 3,582. Here is what changed:
- Look. The site became light-only, in a friendly language-app style. It uses Fredoka for headings and Nunito for body text (replacing Inter), with pressed 3D buttons and bordered tiles. The palette is Polish crimson with accents from Łowicz papercuts (wycinanki), and each blog topic and grammar category got its own colour.
- Homepage. A hero phrase card with working pronunciation buttons, the latest posts, a lesson path, a tools grid and an FAQ.
- Blog posts. Larger reading type in a 680 px column, a summary box, a table of contents in the sidebar, and a "start lesson 1" card at the end of every post.
- Lessons, grammar, quizzes and static pages. All restyled. Quizzes got an answer feedback bar.
- Dark mode removed. That meant the theme switcher, its hook, the script that ran before the page painted, and every dark-mode class.
- A bug fix. Images in article Markdown were rendered as a figure inside a paragraph, which caused a hydration error on posts with inline images.
- Accessibility. A follow-up commit darkened one accent colour so white button labels pass WCAG AA (5.0:1). It also swapped a white numeral that only reached 2.94:1.
Some things did not change. No page was added, removed or renamed, and the sitemap and robots files were not touched. The article text lives in the CMS database, which the redesign did not edit. Before merging, the type check and production build passed, and 13 pages were screenshotted on desktop and mobile with no overflow or console errors.
One thing I did not do: measure page speed or Core Web Vitals before and after. So I won't claim the new design is faster.
What the numbers did#
All figures are from Google Search Console for the whole polishpal.pl domain: web search, all queries, pulled on 12 November 2026. Weeks in the charts start on Monday, grouped from the daily data.
| Period | Clicks | Impressions | CTR | Avg. position |
|---|---|---|---|---|
| 6 weeks before (13 Aug – 23 Sep) | 87 | 1,347 | 6.5% | 50.3 |
| 6 weeks after (25 Sep – 5 Nov) | 1,231 | 134,134 | 0.9% | 6.4 |
| 4 weeks before | 50 | 707 | 7.1% | 41.5 |
| 4 weeks after | 605 | 66,189 | 0.9% | 6.6 |
Weekly, the shape is a step rather than a slope. Clicks hovered between 10 and 26 a week through August and September. Then they jumped to 165 in the week of 28 September and held between 146 and 165 a week for a month. They rose again in late October and early November.
The daily data makes the timing sharper. Search Console counts days in US Pacific time. On 24 September the site had 11 impressions, on 25 September 112, and on 26 September 2,253. For the next two weeks it stayed between about 2,100 and 2,500 a day. Average position, which had been in the 20s and 30s, was 9.1 on 25 September and 7.2 on 26 September.
A cleaner comparison: no new articles on either side#
New articles are the usual explanation for a traffic rise, so I checked the publish dates in the CMS. The last article before the redesign went live on 20 August. The next one went live on 29 October. That gives two windows of 34 days each with no new content at all:
- 22 August – 24 September: 66 clicks, 946 impressions.
- 25 September – 28 October: 780 clicks, 87,538 impressions.
So new content does not explain the first step. The same roughly 100 articles, which had mostly sat on Google's pages three to six, moved together to around position six or seven. A few examples, comparing the week before the merge with the week after:
- Polish swear words: position 12.2 → 5.8.
- Sto lat (the birthday song): 47.2 → 8.2.
- What is Biedronka: 26.5 → 6.1.
The number of distinct queries with at least one impression went from 223 in the six weeks before to 3,325 in the six weeks after.
The clicks are also concentrated. One article, on Polish swear words, earned 429 of the 1,231 clicks in the six weeks after, about 35%.
Why click-through rate fell from 6.5% to 0.9%#
CTR is clicks divided by impressions. Before the redesign, almost every click came from people searching for the site by name: "polishpal" and "polish pal" account for at least 76 of the 87 clicks. People who type your name click on you, so the rate was high on a tiny base.
After the jump, the site started appearing for thousands of other queries, mostly at positions five to ten, where a much smaller share of people click. Impressions grew about 100 times and clicks about 14 times, so the ratio fell.
That is not a sign that the pages got worse. It is what happens when a site goes from being found only by name to being shown for real questions. The number to watch is total clicks, and that went up.
Everything else that changed in the same weeks#
To judge the redesign fairly, here is the full list of other changes, in order:
- 24 Sep, 16:15 UTC: Google starts the September 2026 spam update, "globally and to all languages". It finished on 8 October.
- 24 Sep, 23:45 UTC: the redesign is merged.
- 6 Oct: an accuracy pass fixes 68 Polish errors in lessons, quizzes and grammar pages. Blog articles were not part of it.
- 27 Oct: scheduled publishing moves to the server.
- 29 Oct: the first new article since August (Polish Independence Day). More followed on 3, 6 and 9 November.
- 4 Nov: SEO fixes from a site scan: complete metadata on every page, heading levels, new terms and disclosure pages, security headers and an RSS feed. On the same day, 115 internal links were added across 58 posts.
- 5 Nov: keyword research with DataForSEO volumes, a new logo, the mascot, a newsletter and Amazon affiliate links.
- 6 Nov: another 140 internal links across 54 posts.
- 8 Nov: larger main images and 16:9 share crops, aimed at Google Discover. Discover is reported separately from web search. It showed no impressions at all until 6 November and then 0 to 5 a day, so it is not in the numbers above.
- 10 Nov: Google starts its November 2026 core update. I left the partial week of 9 November out of the analysis for that reason.
The second rise, in late October and early November, overlaps several of these changes. Comparing 26 October – 8 November with the two weeks before, clicks went from 307 to 720. The new Independence Day article brought 61 of those clicks, and the swear-words article roughly doubled, from 107 to 221. I can't attribute that second step to the redesign either.
One more caution runs the other way. Impressions had been falling before the redesign, from 412 in the week of 24 August to around 110 a week by mid-September. That came after Google's August 2026 spam update, which ran from 18 August. I don't know whether those two things are related.
What I think happened#
These are hypotheses, ordered by how well the data fits them.
1. Google re-evaluated the site during the spam update. The timing fits: the jump starts within hours of the rollout. The breadth fits: pages on unrelated topics moved together within about 48 hours. The size fits too. A spam update demotes spammy pages, and a clean site can rise when pages above it drop. I can't see how competing pages ranked that week, so this remains a guess, but it is the one I lean towards.
2. The redesign helped. It shipped on the same day, so it can't be ruled out. If it contributed, plausible routes are clearer page structure (a summary box near the top, a table of contents), the fixed hydration error on posts with inline images, or less work before the first paint. None of these were measured. Google would also have had to recrawl a large part of the site within a day or two to see the new design, and I have no crawl data for those days to confirm or rule that out.
3. The site was maturing anyway. PolishPal was about three months old. Young sites do climb over time, but usually gradually, and this was a twentyfold step in impressions overnight.
The most I can honestly say: the redesign did not hurt, and search traffic stepped up the day it shipped. How much of that step the redesign caused, I don't know.
What I would do differently#
- Check Google's status dashboard before a site-wide merge. This update was announced at 09:15 Pacific time, hours before I merged. Waiting a couple of weeks would have kept the two effects apart.
- Record a baseline first. Page speed, Core Web Vitals and crawl stats before and after, so the "faster pages" question has an answer instead of a guess.
- Keep a dated change log next to the traffic chart. Reconstructing this list from git history and notes took longer than reading the chart.
- Change one thing at a time when you want to learn from it. A site-wide template can't be split, but content, links and metadata changes can wait a few weeks.
Sources#
- Google Search Status Dashboard: September 2026 spam update
- Google Search Status Dashboard: ranking update history
- Search Console Help: Performance report, overview
- Search Console Help: Performance report, dimensions and data groupings
FAQ#
Did the redesign increase PolishPal's search traffic?#
Traffic rose sharply from the day it shipped, but Google's September 2026 spam update started the same day. The data can't separate the two. The redesign did not hurt, and that is as far as I can go.
Why did CTR drop if clicks went up?#
Impressions grew about 100 times and clicks about 14 times. The site started appearing for thousands of non-brand queries at positions five to ten, where fewer people click.
Were new articles behind the jump?#
No. No articles were published between 20 August and 29 October, and the jump happened in the middle of that gap.