First-party research
Search Impressions Without Clicks: A SaaS Case Study
A current look at our own Search Console data: steady exposure, fewer clicks and three improving queries. Here is how those observations change the work we choose.
Cite this article
Om Patel. Search Impressions Without Clicks: A SaaS Case Study. Linkeddit. 2026-10-28. https://linkeddit.com/blog/search-impressions-without-clicks-case-study
Key takeaways
- Google impressions were almost unchanged across two equal windows: 100,720 versus 100,678. Clicks fell from 662 to 517.
- Three deliberately selected product-relevant queries improved in average position, yet together produced only one visible click in the later window.
- The useful question is which buyer query and page need work. Property-wide averages cannot choose that page for you.
- This is an observation from one SaaS property. It does not measure AI citations or establish the effect of our October changes.
1What happened when impressions stayed flat?
Linkeddit received 100,678 Google Web impressions and 517 clicks from September 29 to October 26, 2026. In the previous 28 days it received 100,720 impressions and 662 clicks. Exposure was nearly flat; clicks changed by -21.9%.
| Property measure | September 1 to 28 | September 29 to October 26 |
|---|---|---|
| Impressions | 100,720 | 100,678 |
| Clicks | 662 | 517 |
| CTR | 0.66% | 0.51% |
| Impression-weighted average position | 13.45 | 15.84 |
That does not make every page a failed page. Total impressions combine different queries, result formats, devices and countries. We needed the query/page view to find the specific discovery opportunities hiding inside the aggregate.
2Which queries improved without producing many clicks?
We selected three queries because each matches a job Linkeddit currently supports. This is a deliberate selection, not a random sample or an estimate of market demand. Each has an existing page owner.
| Exact query | Previous impressions / position | Current impressions / position | Current clicks |
|---|---|---|---|
| ai brand monitoring | 1,211 / 17.8 | 1,765 / 13.9 | 0 |
| competitive intelligence tools | 394 / 22.4 | 794 / 17.8 | 0 |
| gemini visibility tracker | 28 / 13.5 | 754 / 10.9 | 1 |
All three moved closer to the first page on this blended measure. Only the Gemini query produced a visible click. The immediate hypothesis is that these owners need to match the buying task and result format more convincingly. We must inspect the live result before attributing low CTR to a title, a missing feature or an AI answer.
The relevant owners are our AI brand monitoring comparison, competitive intelligence tools guide and Gemini visibility tracker. These observations support improving existing owners; they do not establish that we need another page for every keyword variation.
3How was the case study measured?
We queried our own sc-domain:linkeddit.com Search Console property through the GSC MCP on October 28. Both windows contain 28 complete returned dates, with Web search, all countries and all devices. Search Console dates use Pacific Time. Daily property totals, page data and query/page data were collected separately.
The API returned 3,888 query rows for the earlier window and 5,355 for the later one. Their visible clicks were 333 and 224, below the property totals. We did not fill the missing queries with guessed categories. The public download includes daily totals, the selected query observations and the limitations.
Download the case-study data and method notes (JSON)For the underlying counting rules, read Google's definitions of impressions, clicks and position and the Search Analytics API reference. The API can omit rows even when a request returns fewer than its row limit.
4What should change on a page with exposure and few clicks?
| Observation | Next check | Action if confirmed |
|---|---|---|
| A relevant query averages 11 to 20 | Does the owner answer the same buying task as the live results? | Improve its answer, evidence and decision criteria; keep the existing canonical. |
| A page average is poor but its main query is top ten | Which secondary queries lowered the average? | Protect the winning answer; correct factual or discovery gaps without a wholesale rewrite. |
| A hub appears for unrelated vendor alternatives | Would that buyer actually choose our supported product? | Clarify categories and link to fitting comparisons; do not manufacture unrelated replacement claims. |
| CTR changes while position appears stable | Do device, geography, query mix or result features differ? | Compare like with like before blaming the snippet or AI Overviews. |
| The page was changed today | Has it been published and crawled, and are complete later dates available? | Keep the baseline and wait for a fair comparison instead of rewriting it again. |
This table is our decision method, not a measured ranking recipe. Better documentation and clearer page ownership are actions we can verify; later search movement is a separate result to observe.
5Does this show AI answers taking the clicks?
No. The ordinary Web data used here contains no captured answers or source lists. Google also provides a generative AI performance report for impressions from its AI features. That report is a different view from this export and does not turn these query rows into a citation panel.
For a query-level AI Overview investigation, use the Search Console and observed-answer workflow. Keep observed citations, brand recommendations, Google clicks and activated users as separate outcomes.
6How can another SaaS team repeat the analysis?
- Choose equal complete windows and export daily totals separately from query/page pairs.
- Select buyer jobs the product actually supports. State how queries were selected and keep missing queries missing.
- Identify the existing page owner, inspect its real result format and verify every proposed factual change.
- Record the exact edit and actual publication date. Keep the baseline and repeat the comparison after enough complete dates.
- Use actual product activity to judge acquisition quality. Impressions alone do not identify buyers or revenue.
We publish this case about our own company and sell the tools discussed here. Corrections can be sent through our contact page; our editorial policy explains how we handle source claims.
Measure a page change against your own search data
Frequently asked questions
Can impressions stay flat while useful queries improve?+
Yes. In this one-property case, three selected current-product queries improved in average position while total Google clicks fell from 662 to 517. Different query families can move in different directions. The selected queries are examples, not a representative market sample.
Does this case prove AI Overviews reduced clicks?+
No. This export contains ordinary Google Web performance data, not a recorded set of AI answers. It does not establish which searches showed an AI Overview, which sources it cited, or whether an Overview caused a click change.
Why do query clicks fail to match property clicks?+
Some queries are anonymized or not returned by the API, and property, query and page aggregation have different counting rules. We calculate property totals from daily rows and keep query observations separate. Adding the views together would double-count rather than recover missing data.
How should I repeat this analysis on my own site?+
Choose equal complete windows, export property totals separately from query/page pairs, select queries tied to actual buyer jobs, inspect their ranking owners and live results, then record the exact change and publication date. Compare enough complete post-publication data before judging it.