How to read the Search Console performance report

What does the performance report show?

The performance report in Google Search Console shows how often your pages appeared in search results, how often they were clicked, for which queries and from which countries. By default you see clicks and impressions for the last three months. You can go back up to 16 months.

Discover, Google News and generative AI features have their own reports. This guide covers the report for search results. Indexing and crawling are handled elsewhere, see the guide on crawl errors in Search Console. For link data, see exporting backlinks.

Clicks, impressions, CTR and position

Four metrics carry the report. Google defines each of them precisely.

When does an impression count?

• Carousels and infinite scroll. The item must be scrolled into view.
• FAQ and “People also ask”. The impression counts when the entry is expanded.
• Image search. In the combined tab, it is enough for the image to be visible. In the image tab, it must be scrolled into view.
• AI Overviews. They occupy one position, and all links inside share it. Clicks on external links count.
• AI Mode. Position is determined as for standard results. Each follow-up question counts as a new query.

How do you read average position?

For each query, Google takes the topmost position of a link and averages these values. If a URL appears in position 2 for one query and position 3 for another, the result is 2.5. Each element in the search results occupies one position. Links inside a carousel or knowledge panel share the position of their container. Positions are counted from top to bottom, first in the main column, then in the secondary column.

The value remains an average and describes no fixed rank. If new queries join at weak positions, the average gets worse even though no individual position has dropped. So check individual queries before you talk about ranking losses. Your own search can differ too, because location and search history play a part.

Why don't chart and table match?

There are three reasons.

1. Counting method. The chart always counts by property, the table by property or by page depending on the tab.
2. Anonymized queries. Google does not show queries individually for privacy reasons if only a few dozen users issued them over two to three months. They are missing from the table and API, but included in the chart totals.
3. Row limit. Search Console stores and shows only the most important rows.

By property, CTR and average position usually look better when several pages of a site appear. Google also credits most data to the canonical URL, even if users visited a duplicate URL.

Which filters are there?

The tabs above the table group the data by queries, pages, countries, devices, search appearance and dates. In the filter bar you can combine several dimensions, such as country and device. Search type and date are always set. For search type, you can choose Web (text-based), Web (multimodal, for example via Google Lens), Image, Video and News. Multimodal searches have no queries dimension.

You can filter queries and URLs by text or with regular expressions in RE2 syntax. By default, an expression also matches parts of the text and ignores case. Use ^ and $ to anchor the start and end, and (?-i) to make it case-sensitive.

• Comparison. You compare two values of a dimension, such as two date ranges or two countries. The table then shows the difference. Only one comparison is possible at a time. A ~ stands for values that are too rare.
• Branded and non-branded. A dedicated filter separates queries with and without your brand name. It is missing for sites with few impressions, the data starts in March 2025, and the classification can be wrong.
• Search appearance. Clicks are tied to the URL. If you filter on a result type, not every click is guaranteed to come from it.
• Filters on queries. As soon as you filter on queries, anonymized queries drop out of the totals.

How fresh is the data?

By default, the report shows only complete days. Preliminary data appears as a dotted line and can still change over the next few hours. Through the API, data is usually available after two to three days.

The 24-hour view provides hourly values with a delay of a few hours, in your local time. All other views use Pacific Time. For readers in central Europe, a report day therefore usually starts at 9 am. Weeks run from Sunday to Saturday, months are calendar months. Weekly and monthly views smooth out fluctuations from weekends and public holidays.

What are the report's limits?

For small and medium-sized sites, the data is usually enough. According to Google, very large sites are the ones mainly affected.

If you query without the query and URL dimensions, for example by country or device, the row limit does not apply. For the most complete list of queries, Google points to the bulk export. The Search Console BigQuery export guide shows how to set it up.

Where do AI Overviews and AI Mode show up?

Both are part of the Web search type data in the standard performance report. The generative AI performance report gives you a separate view. Google announced it in June 2026 and had rolled it out to all sites by August 31, 2026.

This report shows impressions only, broken down by pages, countries, devices and dates. Search Labs experiments are excluded. If it does not appear in your property, there are usually too few impressions in AI features. Chart and table count the same way as in the standard report.

Four analyses for everyday use

1. Narrow down a traffic drop. Compare two date ranges and go through the pages, countries, devices and search type views. This shows whether a few pages or the whole site are affected. Compare the same weekdays and check the data anomalies first. The next steps are in the traffic drop analysis use case and the Google core updates guide.
2. Find pages with low CTR. Sort the pages tab by impressions and show CTR. Many impressions and few clicks are a reason to check title and description.
3. Separate branded and non-branded. The filter shows whether growth comes from your brand or from queries by users who do not know you yet.
4. Compare sections. Filter on a directory with a regular expression and compare it with the previous period, for example after a release.

Common mistakes

• Reading table totals as the overall figure. For the overall figure, use the chart without a query filter.
• Treating the last few days as a drop. Preliminary data can still change.
• Reading position as a fixed rank. It is an average and depends on the mix of queries.
• Trusting zeros in exports. A 0 can hide a value the report does not show.
• Overlooking the time zone. A day in the report follows Pacific Time. A release early in the European morning still falls on the previous day.

FAQ