One of my sites went from about 850 Search Console impressions a day to 16,000 in a week. Before rewriting anything to "capture the opportunity", I spent ten minutes on the numbers. I am glad I did.
The numbers
Over 28 days the site showed 133,505 impressions. One page carried 88,640 of them: an article about what NordVPN really costs over 24 months. One query carried most of that page:
| Query | Impressions | Clicks | Avg. position |
|---|---|---|---|
| nordvpn cost | 80,721 | 0 | 9.0 |
| nord vpn preis | 6,729 | 0 | 2.6 |
Zero clicks at position 9 on eighty thousand impressions is unusual. Zero clicks at position 2.6 on nearly seven thousand is not something a human audience produces.
Three checks that took the "opportunity" apart
1. Compare with the site's own pages at the same rank. On the same property, in the same period, a French article at position 7.3 got 18 clicks from 985 impressions, and a Spanish one at 6.7 got 9 from 986. So the site does get clicks at these positions. This one query does not.
2. Look at who is supposedly searching. The property's top countries by impressions were Austria (18,895, 0 clicks), Denmark, the United States, Norway and the Netherlands. For an English page about a price in euros, Austria and Norway leading with no clicks at all is not a reader profile. 83% of impressions were on desktop, while clicks were split almost evenly between desktop and mobile.
3. Look at the shape over time. Daily impressions sat around 800, jumped to 4,953 on 24 September, stayed between 9,000 and 16,700 for six days, then dropped to 2,579 on 1 October. Demand for a product price does not behave like that. A scheduled job does.
I cannot prove what generated those impressions. Rank trackers and scrapers querying Google at scale are the obvious candidates, and Search Console counts an impression whenever the result is rendered in a results page, whoever asked.
What I changed
Nothing on the page. Its title already answers the query, and rewriting a title to chase readers who are not there risks losing the ones who are.
What I changed is the rule I work by: an impressions number is not a demand number until the clicks at that position look like the rest of the site. The check is cheap:
- Take the page's query with the most impressions.
- Find two other pages on the same property at a similar average position.
- Compare click-through rates. An order of magnitude apart, with zero clicks, means the impressions are not people.
- Split by country and device. If the split looks nothing like the pages that do get clicks, stop.
The dashboard total had told me this site was sitting on its biggest opening of the year. Excluding that single query, its impressions for the period were about 52,800, with the same 180 clicks. That is the number I plan with now.
Top comments (1)
check two is the one i'd lead with. austria sending 18,895 impressions with zero clicks is the tell, and you don't need the position comparison to see it. a query with real demand spreads. it spreads across countries and devices. scraper traffic clusters, because scrapers live in datacenters in specific places and they don't keep office hours. so i'd flip your order: split by country and device first, and if the pattern looks like a scraper's sleep schedule, the ctr comparison is just confirmation.