Every SEO blog quotes a zero-click statistic from someone else's study. We had a 16-month Search Console export sitting in the repository, so we ran the numbers on ourselves instead. The result was worse than the headlines we had been citing — and the shape of it was not what we expected.
1. 763 of 995 non-brand queries (77%) earned zero clicks, across 474,117 impressions.
2. Ranking top-10 did not rescue them: 72 queries sat inside the top 10 and still earned zero clicks across 30,449 impressions.
3. Question-shaped queries were hit hardest. At positions 1–5 they earned 0.144% CTR versus 2.749% for conventional short queries — a ~19× gap at the same positions.
4. Geography mattered more than we assumed: the US returned 876,622 impressions at 0.07% CTR while India returned 28,702 at 3.35% — about 48×.
theStacc sells SEO automation software, so we have a commercial interest in how you read this. We have tried to keep that out of the numbers: every figure below is computed directly from our Search Console export, the method is described so you can reproduce it on your own data, and the limitations section lists what this data cannot tell you — including the fact that it cannot prove AI Overviews are the cause.
The dataset
One property, thestacc.com, exported from Google Search Console on
2026-08-14 over a 16-month window. The query-level export contains 1,000 rows totalling
557,424 impressions and 975 clicks. We separated brand queries (any
containing "stacc") from non-brand, because brand traffic behaves nothing like discovery
traffic and mixing them hides the effect entirely.
| Segment | Queries | Impressions | Clicks | CTR |
|---|---|---|---|---|
| Brand | 5 | 1,890 | 643 | 34.0% |
| Non-brand | 995 | 555,534 | 332 | 0.060% |
Five brand queries produced 66% of all clicks from 0.3% of impressions. Everything interesting is in the other 555,534.
Finding 1 — position stopped predicting clicks
The classic CTR curve says position 1–3 earns somewhere around a quarter of clicks. Our non-brand data does not resemble that curve at any point.
| Avg position | Queries | Impressions | Clicks | CTR |
|---|---|---|---|---|
| 1–3 | 16 | 2,876 | 12 | 0.417% |
| 3–5 | 31 | 2,529 | 28 | 1.107% |
| 5–10 | 120 | 30,360 | 108 | 0.356% |
| 10–20 | 192 | 141,813 | 92 | 0.065% |
| 20–40 | 320 | 201,303 | 53 | 0.026% |
| 40+ | 316 | 176,653 | 39 | 0.022% |
Two things stand out. Most of our impressions (68%) sit below position 20, which is a content-quality and authority problem we own. But the top of the curve is the surprise: positions 1–3 returned 0.417% CTR. Sixteen queries, 2,876 impressions, twelve clicks.
Finding 2 — 72 top-10 queries earned nothing at all
Filtering to non-brand queries with an average position better than 10 gives 167 queries. Of those, 72 earned zero clicks across 30,449 impressions. The largest:
| Query | Avg position | Impressions | Clicks |
|---|---|---|---|
| social media scheduling tools free | 9.9 | 1,340 | 0 |
| what is a content audit | 2.4 | 1,295 | 0 |
| free social scheduling tools | 6.8 | 1,113 | 0 |
| characteristics of ai written blog posts 2026 | 6.9 | 1,070 | 0 |
| free scheduling tools for social media | 6.5 | 977 | 0 |
| free social scheduler | 6.0 | 921 | 0 |
| which ai sales rep is best for discovering and messaging… | 2.9 | 892 | 0 |
A query at average position 2.4 with 1,295 impressions and not one click is the single most useful number in this study. Whatever is happening, it is not a ranking failure.
Finding 3 — question-shaped queries were hit hardest
We split non-brand queries into two groups: conversational (starts with what / which / how / why / is / can, contains a question mark, or runs seven words or longer) and conventional (everything else — the short keyword phrases SEO has optimised for since 2010). Then we compared CTR within the same position bands, so position is held roughly constant.
| Position band | Conversational CTR | Conventional CTR | Gap |
|---|---|---|---|
| 1–5 | 0.144% (13 q, 4,168 imp) | 2.749% (34 q, 1,237 imp) | ~19× |
| 5–10 | 0.073% (32 q, 10,974 imp) | 0.516% (88 q, 19,386 imp) | ~7× |
| 10–20 | 0.312% (17 q, 3,840 imp) | 0.058% (175 q, 137,973 imp) | inverted |
At the top of page one — where it matters — a question-shaped query converted roughly 19× worse than a conventional query at the same position. Below position 10 the relationship inverts, but with only 17 conversational queries in that band we treat that row as noise rather than a finding.
The honest reading: the better a query looks like something you would ask an assistant, the less a good ranking is worth. We cannot prove why from this data (see limitations), but it is consistent with answer surfaces resolving the question before the user reaches a result.
Finding 4 — the same content, 48× different outcomes by market
| Country | Impressions | Clicks | CTR | Avg position |
|---|---|---|---|---|
| United States | 876,622 | 646 | 0.07% | 26.5 |
| United Kingdom | 112,939 | 215 | 0.19% | 31.0 |
| Germany | 67,706 | 137 | 0.20% | 22.7 |
| India | 28,702 | 962 | 3.35% | 17.3 |
| Pakistan | 8,198 | 158 | 1.93% | 12.8 |
India delivered more clicks than the United States from 3% of the impressions. Part of that is position (17.3 vs 26.5), but position alone does not explain a 48× CTR spread. If you plan international SEO on impression volume, this is the trap: the market that shows you the most impressions may be the one returning the least traffic.
Method — reproduce this on your own data
- Search Console → Performance → Search results, set the date range to 16 months, export Queries, Pages, Countries and Devices as CSV.
- Split queries into brand and non-brand by substring match on your brand name. Report them separately; do not average them together.
- Bucket non-brand queries by average position and compute clicks ÷ impressions per bucket.
- Classify each query as conversational (leading question word, question mark, or ≥7 words) or conventional, then compare CTR within each position band so position is controlled.
- Count the queries with clicks = 0 and sum their impressions. That total is your zero-click exposure.
What this data cannot tell you
One site, one niche. This is thestacc.com only, in SEO/marketing
software. It is a case study, not an industry benchmark, and our absolute CTRs are low
partly because most of our impressions sit below position 20.
The export is truncated. The Search Console UI returns at most 1,000
rows per dimension, so the query table is the top 1,000 by impressions, not the full
long tail.
No query-to-page pairing. The UI export does not join queries to
landing pages, so we cannot attribute a query's performance to a specific page.
AI Overviews are not measured. Search Console does not report whether
an AI Overview appeared. The link between conversational queries and near-zero CTR is a
correlation in our data. We are not claiming a proven cause, and neither should
anyone citing this.
Small samples at the top. The position 1–5 conversational bucket is 13
queries. It is directionally striking, not statistically robust.
Aggregate, not a time series. A 16-month total cannot show when the
change happened, so we make no before/after claim about any algorithm update.
What we changed because of it
We are acting on our own data rather than filing it. Concretely: we stopped treating impression growth as progress, because 474,117 of our impressions produce nothing; we started measuring commercial queries separately from informational ones, since the informational side is where the clicks vanish; and we stopped expanding into locales on impression counts alone — our best non-English market earned 99 clicks in 16 months, which is a signal to consolidate rather than translate more.
If you want the shorter version of the lesson: rankings are no longer the deliverable. If you are buying SEO on a ranking report, ask what it earned.
Want the same analysis on your own Search Console data?
The method above is deliberately reproducible by hand. If you would rather it ran continuously — with commercial and informational queries tracked separately — that is what theStacc does.
Data & sources
- [01]Google Search Console performance export for
thestacc.com, 16-month window, pulled 2026-08-14 — queries, pages, countries, devices. - [02]Figures computed from that export: 557,424 impressions / 975 clicks at query level; 1,329,155 impressions / 4,107 clicks at page level (the two exports cover different row limits, so totals differ).
- [03]Method and limitations as stated above. No third-party study, estimate, or modelled figure is used anywhere on this page.
