1,510 impressions, 6 clicks: what a 28-day Search Console teardown actually showed

A small English site, roughly 70 posts, written for foreigners dealing with practical things in China. It is about three weeks old. Two weeks ago I pulled the query-level data for the first time and got 157 impressions and 4 clicks over 28 days.

Today the same window reads 1,510 impressions and 6 clicks. Impressions up almost tenfold in twelve days.

The easy read is "growth". What I actually found after sorting all 71 pages with any impression is a single, concrete leak: 1,131 of those impressions land on pages already ranking in positions 1 to 10, and that whole group converted into 5 clicks. A 0.44% click-through rate on pages that are already on page one.

The site is no longer in the "not indexed" phase. It gets impressions, it has rankings. What it does not have is anyone clicking.

Three numbers on one screen, and they disagree

Before any conclusion, the accounting. Google Search Console reports this period three ways:

• Header: 6 clicks, 1,510 impressions, 0.4% CTR, average position 11.1.
• Pages tab: 71 pages with impressions, 8 clicks, 1,559 impressions.
• Queries tab: 47 queries, 104 impressions total, zero clicks.

The queries tab accounts for 6.9% of the impressions shown in the header.

This is not a bug. At low volume, Search Console applies privacy thresholding to queries: it only lists the ones it is willing to list, and it rounds the numbers. The consequence matters more than the mechanism: if you treat the query table as a complete picture, you will draw the opposite conclusion from the one the data supports.

Everything below uses the pages tab, because that layer is complete. The three totals are printed here so the discrepancy is visible rather than silently resolved.

Impressions are already on page one

Sorting the 71 pages by average position into four buckets:

The four buckets add up to 71 pages, 1,559 impressions and 8 clicks, matching the pages tab.

The first row is the interesting one. More than half the pages already sit in the top 10. They take about 72% of all impressions. They convert at 5 clicks per 1,131 impressions, which is under half a percent.

A new site's usual problem is not being indexed. This one is past that. The bottleneck is the two lines of text Google shows: the title and the description are not giving anyone a reason to click.

So the next move is not writing fifty more posts.

The six biggest pages, and two counter-examples

Top six by impressions (clicks and average position in brackets):

• LPR and mortgage rates: 367 impressions, 0 clicks, position 6.1
• FX quota and paperwork: 128, 0, position 7.1
• Paying in China as a foreigner: 94, 0, position 15.6
• Air quality index explained: 84, 0, position 5.8
• Sleep data by group: 71, 2, position 31.8
• Setting up Alipay and WeChat Pay: 70, 0, position 8.0

Five of these share a shape: they answer a factual question, but the title is written as a topic label while the reader typed a question. The LPR page is the extreme case. 367 impressions, zero clicks, position 6.1. It was seen 367 times and never once looked worth opening.

Two counter-examples are worth copying.

One page, explaining why speed is the default in China, got 4 impressions, 2 clicks, a 50% CTR at position 5.0. Almost no impressions, but when it was seen, people clicked. Its title is a standalone sentence; you do not have to guess what the page argues.

The other is the sleep data page: position 31.8, still 2 clicks, 2.8% CTR. It is picking up non-English long-tail queries, which says the audience is not confined to English-language markets. I have not decided whether to follow that thread yet.

Query-level optimization at this volume is blind work

My first instinct was to go query by query. The numbers killed that plan.

The 47 listed queries total 104 impressions and zero clicks. All 6 header clicks land on words the query table does not show. The terms that actually convert are the ones Search Console is not telling me. Optimizing against the visible list means optimizing traffic that was never going to click.

Two things in the query table are still usable.

The first is a clear "do not bother". Five name queries total 22 impressions at positions 1.0 to 5.7, the best positions on the site, with zero clicks. They come from note pages that quote specific authors; the searcher wants the person, not the page. No amount of title work will change that.

The second is "real demand, far away". Three variants of the word for a Chinese tax invoice total 23 impressions at positions 55 to 91. Four phrasings of the eSIM question sit at 75 to 82. The demand exists, but it is dozens of positions from a click, and working on it now changes nothing.

The usable substitute is the pages tab, which is complete. The rule I settled on: decide per page, grouped by position band, never per query.

What I changed, and the test I will judge it by

Five pages: FX quota, LPR, paying in China, Alipay setup, the leaving-and-returning checklist. All five had heavy impressions and zero clicks.

Two edits only. Move the answer to the front of the title, and put the real question phrasing into the description. Body copy untouched. The reasoning is simple: at position six, the title line is the only thing anyone reads.

Baseline is the table above, window 10 to 25 October, pulled on 28 October.

The pass criteria are fixed in advance. In two weeks I look at exactly two things:

1. Whether CTR on those five pages moves off 0%.
2. Whether average position falls out of the top 10.

Not total impressions. Impressions moving is an indexing story; clicks moving is a title story. Mixing them means never knowing which lever did anything.

One risk, recorded up front: three of the five are utility pages. A punchier title that overpromises sends people into a page that does not answer their question, which is worse than no click. The descriptions have to say what the page actually contains, not just bait.

Open questions

Why the query table covers only 6.9% of impressions: no official answer found. I can only confirm the gap at low volume, and I do not know when it closes.

Why the invoice cluster gets 23 impressions but sits at 55 to 91: my guess is the page explains the process but never defines the concept, and English searchers are looking for the concept. That is a guess, untested.

Why the sleep page converts at 2.8% from position 31.8: the non-English long-tail lead. Whether to build a line around it is undecided.

Reproduction

Data: Google Search Console, Performance report, Pages tab, 28-day window (the page itself displayed 10 to 25 October), pulled 28 October 2026, with the dashboard reporting "last updated 3 hours ago".

The bucket table was produced by exporting the 71 page rows and summing by average position. Bucket totals match the tab total.

Drafted with an AI assistant from my own measurements; every number and link here was checked against the source.