The Search performance report uses four core metrics: clicks, impressions, click-through rate, and average position. The labels look familiar, but their Search Console definitions are shaped by result features, canonical URLs, filters, and aggregation.
Treat them as a connected system:
- Impressions describe recorded visibility.
- Clicks describe recorded interactions that lead from Google toward the property.
- CTR describes clicks relative to impressions.
- Average position describes the average topmost search-result position credited to the property or selected page grouping.
Google documents the underlying rules in What are impressions, position, and clicks?.
What counts as an impression?
An impression means that a user saw, or could have seen, a link to the property in Search, Discover, or Google News under the rules for that result type.
For a standard paged search result, a link can receive an impression when it is in the current set of results even if the user does not scroll it into the viewport. If the user must load another results page, the link does not receive an impression until it is in the loaded set.
Result features can behave differently:
- In an independently scrolling carousel, the item generally must be scrolled into view within the carousel.
- In an expandable result feature, the link may need to be revealed by expanding the feature.
- In infinite-scrolling experiences such as some image results or Discover, the item generally must be scrolled into view.
- Scrolling away and back during the same query or session does not create another impression for the same item.
The correct interpretation is “eligible recorded visibility under Google's counting rules,” not “a person read my title.”
Property-level and page-level impressions differ
Suppose three URLs from one property appear for a query.
- In a property-aggregated view, the property receives one impression and the topmost property position.
- In a page-aggregated view, each unique page can receive its own impression and position.
This is why chart and table totals can change when you group or filter by Page or Search appearance. The Performance chart is property-aggregated. Table data grouped by Queries, Countries, Devices, or Dates is also property-aggregated; Pages and Search appearance use page aggregation. Google's aggregation documentation includes worked examples.
What counts as a click?
For most result types, a click is counted when an interaction sends the user from Google Search, Discover, or News toward an external page. An interaction that stays inside Google, such as a query-refinement link, is not normally a click for the property.
Important details include:
- Clicking a result, returning to Google, and clicking the same link again generally counts as one click.
- Clicking a different external link can count as another click.
- When data is aggregated by property, clicking two different links from the same property after returning can count as one property click.
- When data is aggregated by page, clicks to two different pages can count for each page.
- Images, carousels, AI features, AMP, and other result types can have feature-specific rules.
A Search Console click is not the same as a GA4 session or user. The systems collect data at different points and use different processing, attribution, time zones, and privacy rules.
Which URL receives the click?
Performance data is usually assigned to the canonical URL selected by Google. If a duplicate URL appears and a person opens it, the click can be credited to a different canonical URL in the Pages table. A redirect after the visitor has landed does not reassign the original Search performance event.
This is one reason Search Console clicks should not be joined blindly to analytics landing pages. Read Canonical URL data attribution before building a page-level join.
What is CTR?
CTR is calculated as:
clicks ÷ impressions × 100
If a page records 80 clicks from 2,000 impressions, its CTR is 4%.
CTR is a ratio, not an independent count. It can change because clicks changed, impressions changed, or both. Compare the underlying counts before explaining the rate.
Do not average row-level CTRs
If one query has 1 click from 10 impressions and another has 10 clicks from 1,000 impressions, their CTRs are 10% and 1%. A simple average would be 5.5%, which misrepresents the combined data.
Use the weighted calculation:
(1 + 10) ÷ (10 + 1,000) = 1.09%
Search Console calculates CTR from the aggregated click and impression totals for the selected view.
What is a good CTR?
There is no universal target. CTR varies with:
- Position and the number of results from the property.
- Branded versus non-branded intent.
- Device and country.
- Result layout and search appearance.
- Query intent and how fully Google answers it on the results page.
- Title, snippet, date, rich-result treatment, and brand recognition.
Compare like with like: the same query or query group, similar position band, device, country, result type, and season. Low CTR at position 12 is a different problem from low CTR at position 1.
What is position?
Position is available for Google Search results. It is a relative location assigned to a search-result element, where 1 is the topmost position, 2 the next, and so on. Ads do not occupy Search Console positions.
Modern results are not a simple list of ten blue links. A carousel, AI feature, knowledge panel, or other compound element can occupy a position shared by several links. Some feature-specific positions are counted inside a container.
Why the report says average position
For each impression, Search Console records the topmost position for the property or page represented by the selected grouping. It then averages those positions across recorded impressions.
Example:
- For one query, the property appears at positions 2, 5, and 8. Its property position for that impression is 2.
- For another query, it appears at positions 4 and 7. Its position is 4.
- Average property position is
(2 + 4) ÷ 2 = 3.
If you group by a particular page, Search Console uses the topmost position of that page for each impression rather than the topmost position of the whole property.
Average position is weighted by impressions. In BigQuery bulk export, Google stores zero-based position sums; the documented calculation is SUM(position sum) ÷ SUM(impressions) + 1.
Why average position is not a rank tracker
An average of 5.2 does not mean the page currently ranks fifth. It can combine:
- Many different queries.
- Different countries, devices, languages, and times.
- Personalized or contextual result variations.
- Different result features and layouts.
- Multiple pages from the same property.
- A 16-month date range in which ranking changed repeatedly.
Even a query-filtered value remains an average across the remaining dimensions and dates. A manual search may differ because of time, location, device, and recent search history. Google recommends focusing on position trends alongside clicks and impressions rather than treating the absolute value as a single rank.
How to interpret the four metrics together
Impressions rise, clicks rise, CTR is stable
The property likely gained useful visibility or demand grew. Check pages, queries, countries, devices, and position to locate the growth.
Impressions rise, clicks are flat, CTR falls
The property may be appearing for broader or lower-position queries. It could also face a weaker snippet, different intent mix, or more result features. Do not assume the title is the problem until you segment by query and position.
Impressions fall, position is stable
Demand, seasonality, query mix, eligibility, or reporting may have changed. Compare year over year, review queries and pages, and check Google's data anomaly log.
Position improves, clicks fall
The average can improve while high-volume queries lose demand or disappear from the mix. Because the set of impressions changed, the average is not a fixed cohort. Inspect query-level counts.
Clicks rise while average position worsens
New lower-position impressions can pull the average down numerically while adding clicks. This can be healthy growth.
CTR changes after adding a Page filter
Page filtering switches relevant table and chart behavior toward page aggregation. Multiple URLs from one property can be counted separately. Check the aggregation rules before comparing the filtered and unfiltered values.
A reliable analysis sequence
- Select the correct property and Search type.
- Use a complete date range and an equivalent comparison period.
- Check whether the newest data is preliminary.
- Start with clicks and impressions.
- Segment by Page to locate the site area.
- Filter to a page, then inspect its reported Queries.
- Compare device, country, and search appearance.
- Use CTR and position as supporting evidence.
- Check canonical attribution and aggregation if totals seem inconsistent.
- Record filters and limitations with the conclusion.