How to Use the Google Search Console Performance Report for Keyword Research

What the Performance report actually measures

The Performance report shows how your site appears in Google Search: clicks, impressions, CTR and average position, broken down by query, page, country, device, search appearance and date ^[1]. It is the only first-party source of Google query data, so it doubles as a keyword research tool. Open Search Console, choose your property, then go to Performance and Search results; Discover and Google News have separate reports. By default you see the past three months. Most mistakes come from reading one metric without the others, so know what each one counts before you sort a single column.

Clicks and impressions

A click counts when someone clicks a result that leads outside Google. Clicks on links that stay inside Google are not counted, and clicking the same result twice in a session counts once ^[2]. That makes clicks the most reliable number in the report. Impressions are looser. "In general, an impression is counted whenever an item appears in the current page of results, whether or not the item is scrolled into view", with carousels and expandable sections as the exception. So impressions measure visibility, not demand. They are not search volume. A page on page three for a popular query may record only a handful.

CTR and average position

CTR is the click count divided by the impression count, so a page with 10,000 impressions and 300 clicks has a 3% CTR. It depends heavily on position and on what else appears on the results page, so compare it only between pages that rank in similar positions. Average position is the average position of the topmost result from your site, recorded only when that result gets an impression. Read it per query. A page that ranks first for a query seen 100 times and twentieth for a query seen 900 times shows an average close to 18, which hides that it already owns the first query.

Where AI Overviews and AI Mode fit

AI Overviews and AI Mode are counted in the Web data. An AI Overview occupies a single position, and all its links share it. Clicks from both features count as ordinary clicks, and a follow-up question in AI Mode counts as a new query, so long conversational questions now show up in your query table. Google itself recommends watching trends in impressions and clicks more than position alone.

Finding your way around the report

Below the chart sit tabs for Queries, Pages, Countries, Devices, Search appearance and Dates. Clicking a row filters the whole report to that row, which is the fastest way to drill from a query to the pages that rank for it. Move through the tabs in a fixed order, from Queries to Pages to Countries to Devices, and by the fourth tab you usually know which page, market and device explain a change. The search type filter switches between web (text-based or multimodal), image, video and news, and data is recorded separately for each type.

Ranges, granularity and comparisons

You can choose a range of up to 16 months, with Hourly, Daily, Weekly or Monthly granularity. Weekly and Monthly views, added on December 10, 2025, smooth out swings from weekends and holidays. The 24-hour view shows hourly data in your local time, while every other view uses Pacific Time. Use it on launch days. For keyword research it only adds noise. To compare periods, open the date filter, choose Compare and pick two ranges, and a Difference column appears in the table. Google suggests Weekly or Monthly granularity here, and you can run only one comparison at a time.

Filters and regex: turning rows into segments

Click + Add filter and choose Queries or Pages ^[3]. Containing and not-containing filters ignore case but are otherwise exact. The real power sits under Custom (regex). Search Console uses RE2 syntax with partial matching by default, matching ignores case unless you prepend (?-i), and several values go inside parentheses, separated by a vertical bar. Regex turns the report from a list into a set of segments. A store can see pricing, comparison and how-to queries as three groups in minutes. A good starting set:

• Question queries: ^(how|what|why|when|which|can|does)
• Pricing intent: (price|pricing|cost|cheap)
• Comparisons: (vs|versus|alternative)
• Long queries of six or more words: ( .*){5,}
• Your brand excluded: choose Doesn't match regex and enter your brand name

Branded and non-branded queries

Test each expression on a small date range before you trust it. For the brand split, Google now does the work: open the filter bar, choose Query, then Branded or Non-branded. Branded queries include the brand name, domain, products and common misspellings. The data starts on March 11, 2025, and the filter is not available for sub-properties or low-traffic sites. Watching the non-branded share is one of the most honest measures of SEO progress, because branded searches grow with advertising and word of mouth, while non-branded clicks grow only when new people find the site.

Why the totals never add up

Queries, Countries, Devices and Dates tables are aggregated by property, while Pages and Search appearance are aggregated by page ^[1]. CTR and average position are generally higher when aggregating by property, so the same query can show a better CTR in the Queries tab than in the Pages tab, and both numbers are correct. Filtering by query or URL can also change totals, because of data truncation and anonymized queries. Match and do-not-match totals may not add up to the unfiltered total, and a page filter can even raise totals ^[3]. Note the unfiltered total first.

Anonymized queries and the 1,000-row ceiling

Some queries are left out of the table to protect user privacy. These anonymized queries still count in chart totals unless you filter by query ^[4]. For a niche B2B site, a large share of clicks can come from queries the table never shows; one developer reported that 29 of the 31 clicks to his best-performing article came from such queries. That is extreme. On top of that, the table shows at most 1,000 rows, and rare long-tail rows may drop out of the table while staying in the chart total. Bulk export gives the most complete query list.

Exporting, and when to move to the API

Export sends chart and table data to Google Sheets, Excel or CSV, up to 1,000 rows, and values shown as ~ or - become zeros. Export with filters already applied, since five filtered exports of 1,000 rows each beat one unfiltered one. Beyond that, the API returns up to 50,000 rows per day per search type per property and allows 1,200 queries a minute per site. For a site with 30,000 queries a month, the report table shows about 3% of them. Our Search Console API guide covers the connection.

Keyword research from queries you already earn

Search Console cannot find new keywords from nothing. It finds the ones you already earn impressions for, which is often more useful, because Google already connects those topics to your site. Start in the Queries tab and sort by Clicks, CTR or Impressions. Clicks show what pays today, and impressions show where Google already shows you. Queries are exact matches, not case-sensitive, so group near-identical phrasings with regex before drawing conclusions.

Queries sitting on page two

Filter for queries with high impressions and an average position around 8 to 20. These are the closest to page one. For each, use the Pages tab to find the page that ranks, and improve that page rather than creating a new one. This is usually the fastest win in the report. A query with 5,000 monthly impressions that climbs from position 11 to position 4 can multiply its clicks several times over, which can mean hundreds of extra clicks a month from a page that already exists.

Queries with impressions but few clicks

Select CTR as the metric, open Pages and sort ascending. High impressions with low CTR usually mean "other search results are more compelling than yours". The fix is the title and description. Check whether the title uses the words of the queries that bring the page most impressions; if people search pricing and your title says plans, change it. Google decides titles and snippets algorithmically, but your title is the strongest hint. Recheck after four weeks. A page with 40,000 impressions and a 1.5% CTR earns 600 clicks, while a 3% CTR earns 1,200 at the same position.

Questions, long-tail topics and query groups

A question regex lists the how, what and why queries, each a candidate for an FAQ section, a heading or a new guide. For new content, look for queries where your best page is only loosely related: gluten-free bread impressions on a general bread page signal a topic for its own page, provided no other page targets it. Since October 27, 2025, Insights also groups similar queries for properties with a large query volume, so you can plan clusters around topics rather than spellings.

Winners, losers and pages that compete

Compare the last three months with the previous period, open Pages, and sort by the click Difference column, exactly as Google's own drop guide does. Inspect the biggest losers first, and check whether the query mix changed or the position fell. Then sort the other way. Pages that grew show what Google rewards now, so find what they share, whether freshness, format or depth, and repeat it. Growth is easier to copy than decline is to reverse.

Cannibalization between your own pages

Click a query, then open the Pages tab. If two or more of your pages share impressions for it, they may be competing. Merge them, or make each target a clearly different intent, and check again after the next crawl. Position usually stabilizes once one page clearly owns the query. Three pages splitting one query is a warning sign. If none holds a stable position, merging them into one stronger page often lifts the combined clicks within a few weeks.

Devices, countries and search features

Compare can put two devices side by side. A large mobile gap in CTR or position for the same query points to layout, speed or snippet problems on phones. Unexpected countries in the Countries tab can mean you rank for terms in another language or market, so if you sell only in the US, filter to it before keyword research. In the Search appearance tab, clicks shown under a feature are not guaranteed to be clicks on that feature, because clicks are assigned to a URL.

Diagnosing a traffic drop

Google's guide sets the range to the last 16 months, compares the last three months with the previous period or year over year, checks each search type separately, then sorts pages by click difference. The guide separates a small drop, such as position 2 to 4, from a large one, such as position 4 to 29. The first is normal movement. The second usually points to a technical, policy or quality problem. Weekend dips and holiday swings are normal too. A December dip for a B2B site is not a penalty.

Data anomalies and data delays

Ranking or reporting changes may be announced on the Search Central blog and the data anomalies page, so check both before acting. Google admits a logging error misreported impressions from May 13, 2025 until April 27, 2026, while clicks were unaffected. That alone explains odd impression trends. Data normally arrives in 2-3 days, recent preliminary days are drawn as a dotted line, and the report shows complete days by default, so end research ranges at least three days ago.

Recording changes and reporting them

Right-click the chart on a date to add an annotation of up to 120 characters for title rewrites, launches and migrations; annotations older than 500 days are deleted automatically. To track a content refresh, annotate the date, filter to the page, and compare four weeks before with four weeks after on Weekly granularity. Watch impressions first, because they move before clicks. If nothing changes after six weeks, the refresh missed what searchers wanted. The only CTR benchmark worth trusting is your own history at the same position.

When you report to a client or boss, lead with clicks, impressions and top queries rather than average position alone, and explain anonymized queries once so nobody expects totals to add up. The Share button shares only the current view, filters included, and share links are read-only; for ongoing access, add the person as a restricted user. Plan for history, too. The report keeps 16 months and bulk export has no backfill, so start it now if you will ever need longer comparisons.

A weekly and monthly routine

A weekly check of about 15 minutes catches problems early and slowly builds a list of pages worth improving. Pages with many impressions in the generative AI report and falling clicks here may be answered inside the AI feature itself, so that report belongs in the check too. Run it on the same day each week, so the comparisons line up:

• Compare clicks week over week.
• Sort pages by the Difference column.
• Scan the top ten queries for surprises.
• Glance at the generative AI report.
• Annotate any changes you made.

Once a month, go deeper. Export queries with impressions, filter positions 8 to 20, list low-CTR pages, check the branded share, and review device and country splits. Then turn the findings into three small tasks: one page to improve, one title to rewrite and one new topic to write. Keep them small. A single improved page that gains 200 clicks a month is worth more than a long list of findings nobody acts on, and twelve such pages a year compound into a measurable share of organic traffic.

What the report cannot tell you

It cannot show search volume, competitor rankings or queries you do not appear for, and it hides rare queries. For volume and competitors you need a paid tool; for your own site, this report is the source. The common mistakes come from forgetting those limits: treating impressions as search volume, averaging position across all queries, expecting filtered totals to match, and reading yesterday's preliminary data as final. If some pages earn no impressions at all, the problem sits upstream, and our Page indexing errors guide is the place to start.

The bottom line

The Performance report is the best free keyword research tool for your own site, because it shows the queries Google already connects to you. Sort, filter and compare, then improve the pages closest to page one before writing anything new. Annotate every change so the next comparison means something. Add the API when 1,000 rows stop being enough. If your property is new, our setup guide makes sure the report has data from day one.