Most trend lists repeat the same predictions with someone else’s statistics. This post does something narrower: it reports five patterns in PixelCrayons’ own Google Search Console data, from July 2025 to September 2026, and shows how to check each one in your own account in an afternoon.
It is one site’s data, so it proves nothing about the market. It does show which popular claims held up for us, and which numbers misled us when we looked closely.
What we looked at, and the limits of one site
We ran the check on 28 September 2026, reading 15 months of our Search Console data through the Ahrefs connection: monthly clicks and impressions, weekly counts of queries by position band, and pages and queries in two equal 25-day windows, 6 to 30 August and 31 August to 24 September 2026. The latest data point was 24 September.
Two limits matter. The figures describe our previous site, before the rebuilt site went live on 29 September 2026, so our own site changes are mixed into every trend. And the totals use a page-level filter, which matches page tables but can differ slightly from the account summary.
Five patterns in our own data
Each row is a claim you will find in most trend lists, what our data showed for it, and the period. The last column is the check to run on your own property before you believe the claim applies to you.
- July 20255,349
- August 2026411
- July 20253,869,765
- August 2026453,929
| Pattern | What our data showed | How to check yours |
|---|---|---|
| 1. Organic clicks are falling | Monthly clicks fell from 5,349 in July 2025 to 411 in August 2026; impressions from 3,869,765 to 453,929 | Export fifteen months of monthly clicks and impressions; compare the same month a year apart |
| 2. Average position can improve while visibility falls | Average position moved from about 41 to about 31 to 33, while impressions fell 19% between the two 25-day windows | Count queries per position band each week, not the average |
| 3. Tool estimates and real clicks can point in opposite directions | A third-party estimate of our organic traffic rose 37% from 31 August, while our Search Console clicks fell 8% | Take the tool’s ten biggest gainers and look each one up in Search Console |
| 4. Impressions without clicks | Our Google Ads service page had 11,740 impressions and 17 clicks at an average position of 32 (31 August to 24 September) | List pages with high impressions and almost no clicks; check their position first |
| 5. Brand carries the clicks | Branded queries were about half of all our clicks in September 2026 | Split clicks into branded and unbranded before reading any trend |
What each pattern meant once we looked closer
Falling clicks. We found the drop real and steep, but we cannot separate the market from our own ageing site. September 2025 also fell sharply from August 2025, so early autumn is not a seasonal high for us. A year-on-year comparison of the same months is the least misleading view.
The average position. Our weekly count of queries ranking 51 and below fell from 8,670 in the week of 13 April 2026 to 1,531 in the week of 21 September. Deep-ranking queries dropping out lifts the average, so the line improved while we lost visibility.
Estimates against clicks. One query the tool placed fifth, “ai development companies”, had 37 impressions, an average position of 18.3 and no clicks in Search Console over the same window. Between 7 and 28 September the tool also showed about 330 new referring domains, almost all automated spam sites. Neither was a gain.
Impressions without clicks. In the same month the tool recorded our pages cited in AI Overviews for several service queries. We cannot say how many clicks those answers took from us. We can say the citations did not show up in our click data.
Brand. Clicks on “pixelcrayons” fell from 133 to 99 between the two windows, while “pixel crayons” rose from 24 to 40. Net brand clicks were roughly flat. With brand at about half of clicks, any unbranded trend is half the size it looks.
What we take from it
Three working rules follow from this check:
- Decide on Search Console clicks and impressions, never on a tool’s traffic estimate. Estimates are useful only for finding queries to check.
- Read position by band, week by week, beside impressions, so a rising average cannot hide a shrinking footprint.
- Track AI answers separately from search clicks. From 5 October 2026 we monitor a fixed set of 39 buyer questions, each tagged with the page that should be cited, and read it monthly. No reading is reported here because none has been filed yet.
How we run that monitoring is in the AI visibility audit. The search work it feeds is on our SEO services page.
How to run the same check on your account
You need Search Console access and about three hours. Run it before you change strategy because of a trend article, including this one.
- Export monthly clicks and impressions for the last fifteen months. Compare each month with the same month a year earlier.
- Filter out branded queries and repeat step one. Note how much of the change was brand.
- Export queries with their position and group them into bands: 1 to 3, 4 to 10, 11 to 20, 21 to 50, 51 and below. Repeat weekly for a quarter.
- List pages with the most impressions and their click-through rate. Pages far below page one need ranking work before any title change.
- If you use a third-party SEO tool, take its ten biggest gainers and look each one up in Search Console.
Clicks are not the last word either: a page can rank and still lose money. The keyword that is costing you money shows how to spot that in your own analytics.
Where this check misleads
A check like this reads the past. It cannot tell you whether a fall came from the market, an algorithm update or your own site, and on one site it cannot tell you anything about other businesses. Search Console data also arrives a few days late and leaves out some rare queries to protect searchers’ privacy.
If your site changed platform, structure or addresses in the period, treat every trend as mixed with that change. Ours did, which is why we report these patterns as observations, not as forecasts.
