Google Search Console Performance Report: Reading Queries, Clicks, CTR, and Position

The Search Console Performance report shows how a site performs in Google Search with four metrics (clicks, impressions, click-through rate and average position) broken down by query, page, country, device, search appearance and date. Each metric has a precise definition, and the misreadings below come from ignoring it: average position is the position of your topmost result, not of every result; impressions count appearances, not views; and the table does not contain every query. Since 2026 the report also sits beside three newer views (generative AI performance, multimodal search and platform properties) that answer different questions and should not be merged with it.

The four metrics, as Google defines them

Google’s help page for the report defines the metrics this way (Performance report help):

• Clicks: “The number of times a user clicked your site from Google Search results”.
• Impressions: “How many times your site appeared in Search results”.
• CTR: clicks divided by impressions.
• Average position: “The average position of the topmost result from your site”.

Average position has two readings. In the chart it is the average position of the topmost result from the whole site. In the table it is “the average position in Search results for the specific URL or grouping dimension shown in the table row”. A page can therefore show an average position of 4 in the table while the site’s chart value for the same query is 2, because a different page ranked higher.

Google’s traffic-drop guide adds two cautions worth keeping beside the definitions: “you shouldn’t focus too much on your absolute position”, and “Impressions and clicks are ultimately the measure of success for your site” (debugging Search traffic drops).

Dimensions and how they aggregate

The table groups data by one dimension at a time: Queries, Pages, Countries, Devices, Search appearance and Dates. The grouping changes the arithmetic. Google’s help page states that data grouped by queries, countries, devices or dates is aggregated by property, while data grouped by pages or search appearance is aggregated by page.

In practice:

• When two of your pages appear for the same query, the Queries view counts one impression (the property appeared once), while the Pages view counts an impression for each page.
• Chart totals and table totals can differ. Google lists this as a known discrepancy caused by the aggregation difference, not an error.
• Pages are reported under their canonical URL, so a duplicate URL that Google folds into a canonical will not appear separately.

What the table leaves out

Three limits shape every export:

1. Privacy filtering. Google’s data-discrepancy help states that, to protect user privacy, the report “omits some queries that are searched a very small number of times”. On the query tab these anonymized results are left out of the table but included in the chart totals, unless a query filter is applied (Performance data discrepancies). That is why the sum of query rows is lower than the total clicks shown above the table, and why “matches” and “does not match” filters may not add up to the unfiltered total.
2. Row limits. The interface shows up to 1,000 rows per table. Larger datasets require the Search Console API or the bulk data export.
3. Time range. The report keeps 16 months of data.

A practical consequence: never compute “share of clicks from brand queries” by dividing the brand rows by the chart total without noting that the unlisted queries are in the denominator but in neither group.

Search types, including multimodal

The search type filter splits the data into web, image, video and news searches. Since September 24, 2026, web search is itself split into two options (Google’s September 2026 announcement):

• Web: text-based covers text queries typed into the standard search bar.
• Web: multimodal covers searches made with images, which Google describes as including Google Lens, Circle to Search on Android, uploaded images and Chrome image search.

The multimodal filter is available both in the Search results performance report and in the Generative AI performance report. Image-led searches are therefore not a separate world you cannot see; they are a filter on reports you already use. Google notes the data appears only if a site receives traffic from these searches.

Where AI features fit

Two facts are true at the same time and need to be held together.

First, traffic from AI features is already inside the standard report. Google’s AI features guide states that sites appearing in AI features such as AI Overviews and AI Mode are included in overall search traffic and reported within the “Web” search type (AI features and your website). Those clicks and impressions are not labelled separately there.

Second, a dedicated Generative AI performance report was announced on June 3, 2026 and rolled out to all websites worldwide as of August 31, 2026 (Google’s June 2026 announcement). Its help page documents impressions broken down by pages, countries, dates and devices (Generative AI performance report help). Google defines an impression in this report as the number of times links to your site were shown in a generative AI feature. The page does not describe clicks, CTR, queries or average position for this report, so none of those should be reported as coming from it. Discover has its own generative AI report.

Four reports, four questions

Platform properties are a separate property type introduced on July 7, 2026 and made globally available on July 29, 2026 (Google’s July 2026 announcement). They report on posts hosted on those platforms, not on pages hosted on your domain, so their numbers belong in a separate column of any report.

Reading the report: a working sequence

1. Set the comparison first. Use a date comparison (last 3 months against the previous period or the same period a year earlier) before reading any single number. Google’s traffic-drop guide recommends exactly these comparisons.
2. Check whether the change is site-wide. Look at the chart before the table. A site-wide drop points to technical, policy or demand causes; a drop concentrated on a few pages points to those pages.
3. Open Pages, then Queries for each page. Filter by one page and switch to the Queries tab to see what it is shown for.
4. Separate impressions from clicks. Stable impressions with falling clicks suggest a change in how results are presented or in title and snippet appeal; falling impressions suggest lost visibility or lower demand.
5. Filter by search type. Compare text-based and multimodal web traffic, and image search separately, if the site relies on visual content.
6. Read position as a trend, not a rank. A move from 8.4 to 6.1 across a group of queries is a signal; a single query’s position is an average across many searches, devices and locations.

Worked example: why position “improved” while clicks fell

Take a hypothetical site whose average position for a query group improves from 9.2 to 7.5 in a month while clicks drop by a fifth.

• In the Pages view, the main guide lost impressions while a newer, more specific page began appearing for the same queries in higher positions.
• The site’s topmost result moved up, so the chart’s average position improved.
• The newer page is shown for fewer searches and draws a lower CTR, so total clicks fell.

The metrics are not contradicting each other. Average position follows the best-ranked page; clicks follow the full set of appearances. The next step is to decide which page should own those queries (keyword cannibalization is the usual frame for that decision), not to celebrate the position change.

Misreadings to avoid

• “Position 1 means everyone saw us first.” Position is an average of the topmost result across searches; personalization, location and features above organic results vary.
• “Impressions mean people looked at the result.” Google defines an impression as an appearance of your site in results; it does not mean the result was read.
• “The query table is complete.” Privacy filtering removes rare queries, and the interface shows 1,000 rows.
• “AI Overview clicks are separated somewhere.” In the standard report they are part of web traffic; the generative AI report shows impressions only. Visits from AI assistants such as ChatGPT are a different measure, found in GA4’s AI Assistant channel.
• “Platform property numbers are site numbers.” They describe posts on other platforms.

Using the report with other data

The Performance report measures Search behavior up to the click. What happens after the click (sessions, engagement, conversions) is measured in analytics tools, which count different things and will not match. Reconciling the two is a separate task; the short version is that a Search Console click and an analytics session are different events recorded by different systems. Bing reports its own search data separately in Bing Webmaster Tools.

The report is most useful when every number is read with its definition beside it. Clicks, impressions, CTR and position each answer a narrow question, and the newer AI, multimodal and platform views add narrower questions still. Keep them in separate columns and the report tells a consistent story.