Google Search Console data analysis: queries most people never run

TL;DR: Most people open Search Console, glance at the clicks graph, and close the tab. The report holds two years of query-level, page-level, device-level, and country-level data once you export it to BigQuery, and it is the only free source that tells you exactly which query drove which click on which page. This guide covers the queries that surface decay before it shows up in your traffic dashboard, the filter combinations that catch cannibalization, and the CTR benchmarks (Backlinko) you need to tell a real problem from normal variance. Search Console now retains 16 months of history, not the 3 months most people assume (Search Engine Journal).

---

Table of contents

1. What Search Console actually measures
2. The default view hides almost everything
3. CTR benchmarks: what counts as underperforming
4. Finding query-level decay before it hits your traffic chart
5. Detecting keyword cannibalization with query and page filters
6. Striking-distance queries: the fastest wins in the report
7. Low-CTR, high-position pages: the SERP feature audit
8. Filters, comparisons, and the queries most people never run
9. Beyond 16 months: exporting to BigQuery
10. Page indexing and crawl-side signals worth checking weekly
11. A repeatable weekly and monthly GSC workflow
12. Common mistakes that produce wrong conclusions
13. Frequently asked questions
14. Key takeaways

---

What Search Console actually measures

Search Console’s Performance report has four metrics, and Google defines them precisely: clicks are “the number of times a user clicked your site from Google Search results,” impressions are “how often your site appeared in Search results,” CTR is clicks divided by impressions, and average position is “the average position of the topmost result from your site” for a given query or page (Google Search Central). That last one trips people up constantly. Average position is not “where my page ranks.” It is the average rank of whichever URL on your domain showed up first, for every impression, blended together.

The default view shows the past three months. Google expanded actual retention to 16 months, so if you have only ever looked at the default window, you have been analyzing roughly a fifth of the data sitting in your own account (Search Engine Journal). Every date-range comparison in this guide assumes you extend that window manually first.

One more mechanical detail that changes how you should read every chart: Google states plainly that “the newest data can be preliminary,” shown with dotted lines, and can still change for a few hours after it first appears (Google Search Central). Do not diagnose a traffic drop from the last one to two days of data. Wait for the line to solidify.

The default view hides almost everything

Open Search Console today and the landing view gives you one aggregate line for clicks and one for impressions, filtered to nothing. That view answers exactly one question well: is my total traffic going up or down. It answers almost nothing else, because every interesting finding in Search Console comes from a query, a page, or a segment you filtered for on purpose.

The report supports filtering across queries, pages, countries, devices, and search appearance types, and it supports comparison mode across two date ranges or two filter values at once, for example desktop versus mobile CTR for the same query set (Google Search Central). That comparison view, not the default line chart, is where you find the specific pages losing rank while your aggregate traffic still looks flat, because a handful of winners can mask a longer tail of quiet decay.

CTR benchmarks: what counts as underperforming

You cannot judge a query’s CTR in isolation. A 4% CTR is excellent at position 8 and alarming at position 1. Backlinko’s analysis of roughly 4 million search results, spanning 1.3 million pages and 12.2 million queries pulled from Search Console accounts via Semrush, gives you the baseline curve: position 1 averages 27.6% CTR, position 2 about 15.5%, position 3 about 11.4%, and position 10 about 2.8% (Backlinko). The top three positions combine for 54.4% of all clicks in that dataset, and results on page two capture just 0.63%.

The practical takeaway is a shape, not a single number. Moving from position 2 to position 1 produced a 74.5% relative CTR lift in that dataset, while positions 8, 9, and 10 sit close enough together that chasing rank inside that band barely moves clicks (Backlinko). Build your own position-banded CTR baseline from your own site’s data, because niche, brand strength, and SERP features all shift the curve, but use these numbers as the sanity check before you decide a query is broken.

Low-impression queries need a separate read. The same Backlinko dataset found that 90.3% of queries received 10 or fewer impressions, which means most rows in your query table are too small a sample to draw a CTR conclusion from at all (Backlinko). Filter those out, or at minimum sort by impressions descending, before you start hunting for CTR problems.

Finding query-level decay before it hits your traffic chart

Aggregate traffic charts lag the real story by weeks, because a site with hundreds of pages can lose a third of them to ranking decay while three or four large pages keep the total line flat. The fix is a page-level, six-month comparison, not a query-level one. Search the Performance report by page, turn on comparison mode for two six-month windows, and sort by the clicks difference column ascending, so the pages that lost the most traffic surface first (Ahrefs). Every page at the top of that sorted list is a candidate for a content refresh, and the pattern across them (a topic cluster, a content type, a publish-date cohort) usually tells you the actual cause faster than guessing page by page.

This is also the fastest way to catch decay early, before it shows up as a dip in the homepage traffic chart. A page dropping from position 4 to position 9 on its main query loses most of its clicks under the CTR curve above, long before it drops off page one entirely. Run this page-level decay check monthly, not quarterly. A page that has been quietly bleeding rank for three months is much cheaper to fix than one that has fallen off the results entirely.

flowchart TD
    A[Open Performance report] --> B[Set date range to 16 months]
    B --> C[Filter by Pages]
    C --> D[Enable comparison: last 6mo vs prior 6mo]
    D --> E[Sort by clicks difference, ascending]
    E --> F{Top losers share a pattern?}
    F -->|Same topic cluster| G[Check for cannibalization]
    F -->|Same publish cohort| H[Schedule a content refresh]
    F -->|Isolated page| I[Check SERP for new competitors or features]
    G --> J[Consolidate or differentiate pages]
    H --> K[Update, republish, request reindex]
    I --> L[Rewrite for the gap the new SERP leader fills]

Detecting keyword cannibalization with query and page filters

Cannibalization is when two or more of your own pages compete for the same query, splitting clicks, links, and relevance signals across URLs instead of concentrating them on one. Ahrefs’ Patrick Stox pushes back on how often the term gets used as a catch-all explanation, calling the panic around it “preposterous” and noting that Google is not confused by multiple pages covering a related topic (Ahrefs). That is a useful check on over-diagnosis, but it does not mean cannibalization never happens. It means you need evidence, not a hunch, before you merge anything.

The evidence lives in Search Console. Filter the Performance report to a single query, then switch the breakdown to pages. If you see two or more of your own URLs each collecting meaningful impressions for that query, especially if their rankings bounce between each other week to week, you have a real case. A second confirmation method that does not require GSC at all: search site:yourdomain.com "your exact topic" and append &filter=0 to the Google results URL, which removes Google’s own deduplication and shows every one of your ranking pages for that query side by side (Ahrefs).

Once you confirm it, the fix is consolidation, not deletion. Merge the weaker page’s unique content into the stronger one, then 301-redirect the loser to the survivor so its backlinks and internal links transfer instead of evaporating (Ahrefs). Ahrefs’ own broken-link-building guides were merged this way in 2018 and organic traffic to the consolidated page rose meaningfully afterward. Avoid deleting the weaker page outright or noindexing it as a first move; both throw away link equity that a redirect would have kept.

Striking-distance queries: the fastest wins in the report

Striking-distance queries are the ones ranking positions 5 through 15, close enough to page one that a modest content update, not a rebuild, can move them. To find them, open the Performance report, enable the average position metric, sort the query table by position, and scroll to the 5-to-15 band, then prioritize the queries with the highest impressions in that range, since those represent the most existing visibility you are currently wasting (Search Engine Journal).

This is the single highest-ROI recurring task in Search Console, and most teams never do it because the default view never surfaces it. A query already generating impressions at position 8 needs a smaller lift to reach page one than a page you have not written yet needs to rank at all. Build a standing list, refresh the page that owns the query, and recheck the position in two to four weeks once Google recrawls.

Low-CTR, high-position pages: the SERP feature audit

The inverse problem: a page ranks well, in the top three, but its CTR sits well under the benchmark for that position. Filter for top-3 average position, then sort by CTR ascending. Every page at the top of that list is either miswritten (a weak title tag or meta description that does not earn the click) or losing clicks to a SERP feature above it, like a featured snippet, an ad block, a video carousel, or an AI Overview (Ahrefs).

Check the actual SERP for the query before you touch the title tag. If a competitor holds the featured snippet, rewriting your title will not fix a click you were never going to get; you need to restructure the answer near the top of the page to win the snippet instead. Ahrefs documents a case where winning a featured snippet produced an 8.6% average CTR and became 38.5% of that page’s total traffic, which is the size of the prize when this diagnosis is right (Ahrefs).

Filters, comparisons, and the queries most people never run

Most Search Console users run one filter at a time: this query, or this page, or this country. The report is more useful stacked. Filter to a single page and a single device, then compare against a prior period, to see whether a mobile-specific issue (a slow-loading interstitial, an unindexed mobile-only element) is dragging down what looks like an overall ranking problem. Filter to branded queries only, isolated from non-branded, before you report a CTR or click trend to anyone, because branded query volume moves for reasons that have nothing to do with your content work and will otherwise contaminate every other number in the report.

Two comparisons worth running on a recurring basis and almost nobody runs:

• Non-branded clicks, six-month trailing, by page type. Segment blog, product, and category pages separately. A site-wide “traffic is flat” verdict often hides one page type growing and another declining, and averaging them together tells you to do nothing when you should be reallocating effort.
• Query count with at least one click versus query count with zero clicks, same date range. A widening gap between these two, more impression-only queries relative to click-getting ones, is an early warning that AI-generated answers or SERP features are absorbing demand for topics you already rank for, before your aggregate CTR line shows it.

Beyond 16 months: exporting to BigQuery

Sixteen months covers most seasonal and year-over-year comparisons, but it will not hold a full historical record through a multi-year strategy shift, an acquisition, or a domain migration. Search Console supports a native BigQuery export that stores data beyond the 16-month retention window and lets you query it with SQL instead of the report UI’s row limits (Search Engine Journal). Set this up before you need it. The export only backfills from the date you enable it forward; it does not retroactively pull years of history you never captured.

Two other limits worth knowing before you build a workflow around the UI alone: the URL Inspection tool caps manual indexing requests at 50 per day per property, and the separate Indexing API is capped at 200 publish requests per day per project and restricted to pages carrying JobPosting or BroadcastEvent structured data (Google Search Central). If your workflow depends on bulk reindexing after every content refresh, plan around those caps instead of discovering them mid-migration.

Page indexing and crawl-side signals worth checking weekly

The Performance report tells you what already ranks. The Page Indexing report tells you what Google is willing to rank at all, and it deserves a weekly glance, not a monthly one. A sudden drop in indexed page count is a signal something changed on the crawl or rendering side, not just the ranking side, and it should be investigated before it shows up as a traffic drop three weeks later (Semrush).

Pair that with a basic crawl-budget check on large sites. Google’s own guidance is direct: duplicate and near-duplicate content wastes crawl budget that could go to unique pages, and if Google spends too much time on URLs it should not, it may not get to the rest of your site or may not increase your crawl budget over time (Google Search Central). A rising count of crawled-but-not-indexed pages in the indexing report is often the earliest visible symptom of that problem, well before it costs you rankings on pages you actually care about.

A repeatable weekly and monthly GSC workflow

Cadence	Task	What it catches	Tool needed
✓ Weekly	Page Indexing report scan for sudden index-count drops	Crawl or rendering breakage	GSC only
✓ Weekly	Preliminary data check on last 48 hours	Real-time incident detection	GSC only
✓ Monthly	Six-month page-level comparison, sorted by clicks lost	Query-level decay before it hits aggregate traffic	GSC only
✓ Monthly	Striking-distance query review (positions 5 to 15)	Fastest available ranking wins	GSC only
✓ Monthly	Top-3 pages sorted by CTR ascending	SERP feature loss and weak titles	GSC + manual SERP check
✗ Never (skip)	Judging CTR against a flat 2 percent to 3 percent target regardless of position	False positives and false negatives on almost every query	N/A
✗ Never (skip)	Reporting aggregate clicks without segmenting branded from non-branded	Contaminated trend line that hides real problems	N/A
✓ Quarterly	BigQuery export health check and historical query pull	Multi-year trend analysis beyond the 16-month window	GSC + BigQuery

Search Console cannot tell you what keyword to target next, only how the ones you already rank for are performing. That gap is where dedicated keyword research tooling earns its place in the stack: Vrid.ai pairs keyword research with AI article generation and multi-channel publishing, so the striking-distance queries and content gaps you surface in this workflow can go straight into a brief without a manual handoff between tools.

Common mistakes that produce wrong conclusions

The most common Search Console mistake is treating average position as a page’s rank. It is a blended average across every impression, so a page holding position 3 for a branded query and position 14 for a non-branded one will show a mid-range average that describes neither situation accurately. Always break position out by query, never read it at the page level alone.

The second mistake is diagnosing a drop from the last two or three days of data. Google states outright that the newest data is preliminary and can still shift (Google Search Central); a drop that looks alarming on Tuesday can partially resolve by Thursday once the numbers finalize. Wait for the dotted-line window to close before you act.

The third is applying a single CTR threshold, “anything under 3% is broken,” across every position. Position 10’s benchmark CTR sits near 2.8% in Backlinko’s dataset, meaning a “3% minimum” rule flags an entirely normal position-10 result as broken while missing a genuinely underperforming position-2 result sitting at 8% against a 15.5% benchmark (Backlinko). Compare CTR to the benchmark for that specific position, not to a flat number.

Frequently asked questions

How far back does Google Search Console data go?

Search Console retains 16 months of Performance report history, expanded from the original 3-month window (Search Engine Journal). The default view only displays the past 3 months, so you have to manually extend the date range to see the full history. For anything beyond 16 months, set up the BigQuery export before you need it, since it only captures data going forward from activation.

What does “average position” actually mean in Search Console?

Average position is the mean rank of the topmost URL from your site across every recorded impression for a query or grouping, not a single page’s fixed rank (Google Search Central). A query where you rank at position 2 for branded searches and position 12 for non-branded ones can show an average around position 6, a number that describes neither reality. Always segment by query and, ideally, by branded versus non-branded traffic before trusting this metric.

How do I detect keyword cannibalization in Search Console?

Filter the Performance report to one query, then break results down by page. If two or more of your own URLs collect meaningful impressions for that query and their rankings swap places week to week, that is direct evidence of cannibalization. Confirm with a site:yourdomain.com "topic" search plus &filter=0 appended to the results URL, which disables Google’s deduplication and shows every ranking page from your domain at once (Ahrefs).

Is keyword cannibalization as big a problem as people say?

Not always. Ahrefs’ Patrick Stox argues the panic around cannibalization is overstated, since Google is not confused by related pages on the same topic (Ahrefs). Treat it as a hypothesis to confirm with the page-breakdown check above, not a default explanation for every ranking problem. Real cases show up as multiple owned URLs competing for one query with unstable rankings between them.

What is a good click-through rate in Search Console?

There is no single good CTR; it depends entirely on position. Position 1 averages 27.6% CTR, position 3 averages about 11.4%, and position 10 averages about 2.8%, based on Backlinko’s analysis of roughly 4 million search results pulled from Search Console data via Semrush (Backlinko). Compare your query’s CTR to the benchmark for its actual position, not to a flat target like 3%.

Why do my clicks look flat even though some pages are clearly losing rank?

Aggregate traffic charts blend winners and losers into one line. A handful of large pages holding steady can mask a longer tail of pages quietly losing rank. Run a page-level, six-month comparison sorted by clicks-lost ascending to surface the specific pages in decline before the aggregate chart shows any movement (Ahrefs).

What are striking-distance keywords and how do I find them?

Striking-distance keywords rank in positions 5 through 15, close enough to page one that a content update can move them without a full rebuild. Find them by sorting the Performance report’s query table by average position and reviewing the 5-to-15 band, prioritizing by impression volume (Search Engine Journal). These are typically the fastest wins available in the report because the page already has partial relevance and some backlink or authority signal.

Why is my top-ranking page getting so few clicks?

Two likely causes: a weak title tag or meta description that fails to earn the click despite the ranking, or a SERP feature above your result, such as a featured snippet, ad block, or AI Overview, absorbing the click before the searcher reaches you. Filter for top-3 positions and sort by CTR ascending to find these pages, then check the live SERP before rewriting anything (Ahrefs).

Can winning a featured snippet meaningfully increase my traffic?

Yes, in documented cases. Ahrefs cites an example where winning a featured snippet produced an 8.6% average CTR and accounted for 38.5% of that page’s total traffic (Ahrefs). Whether that opportunity exists for your query depends on whether the snippet is currently held by a competitor and whether your page’s structure can be reformatted to answer the query directly near the top.

How often should I check Search Console?

Check the Page Indexing report weekly, since a sudden drop in indexed pages signals a crawl or rendering problem worth catching early (Semrush). Run the deeper page-level decay comparison and the striking-distance query review monthly. Quarterly, check your BigQuery export if you have one set up, and review whether your CTR benchmarks still match your niche.

Should I trust the data from the last day or two?

Not for diagnosis. Google states the newest data can be preliminary and shown with dotted lines while it is still being finalized, and it can change for a few hours after it first appears (Google Search Central). A drop that looks severe in the most recent 1 to 2 days often partially resolves once the data solidifies. Wait for the dotted line to close before acting.

What is the difference between a content refresh and pruning a page based on Search Console data?

A refresh updates the same URL, keeping its backlinks and indexing history, and applies when a page still gets impressions but ranks lower than its content quality justifies. Pruning changes the URL’s fate entirely, through deletion, consolidation, or a noindex tag, and applies when the page is thin, duplicative, or has decayed past the point a content update would fix. See content refresh vs new content for the decision criteria on which action fits which page.

How do I know if a traffic drop is a Google algorithm update or something on my end?

Cross-reference the date of your drop against confirmed core update windows before assuming an algorithm cause. If the timing does not line up with an announced update, check crawl-side signals first, since indexing and rendering issues cause traffic drops that look identical to ranking drops in an aggregate chart. A page-level Search Console comparison, not the homepage traffic graph, is what separates these two causes.

Does Search Console show data for AI Overviews or ChatGPT referral traffic?

No. Search Console only reports Google web search performance data. AI Overview appearances are counted within regular web search impressions where applicable, but Search Console does not break out AI Overview-specific metrics, and it has no visibility into ChatGPT, Perplexity, or other AI assistants at all. Tracking those requires GA4 referrer segmentation, a separate workflow from anything covered in this guide.

Can I filter Search Console queries by more than one condition at once?

Yes. The Performance report supports filtering by query, page, country, device, and search appearance simultaneously, and it supports comparison mode across two date ranges or two filter values in the same view, such as desktop against mobile CTR for one query set (Google Search Central). Stacking two filters, not one, is usually what turns a vague hunch into a confirmed diagnosis.

What is the Indexing API and does it help with SEO reporting?

The Indexing API lets you request crawling for specific URLs programmatically, but Google restricts it to pages carrying JobPosting or BroadcastEvent structured data, with a default quota of 200 publish requests per day per project (Google Search Central). For general content, the manual “request indexing” button in URL Inspection is the correct tool, capped at roughly 50 requests per day per property. Neither tool improves reporting; both only affect how fast Google recrawls specific URLs.

Why do 90% of my queries show almost no impressions?

That is normal, not a sign of a problem. Backlinko’s dataset found that 90.3% of all queries received 10 or fewer impressions, meaning the vast majority of rows in any query table are long-tail, low-volume searches (Backlinko). Sort by impressions descending and focus your analysis on the queries with meaningful volume; drawing conclusions from single-digit-impression rows produces noise, not signal.

How do I compare mobile and desktop performance for the same query?

Filter the Performance report to your target query, then use comparison mode to split the same date range by device instead of by two time periods (Google Search Central). A large CTR or position gap between devices for the same query usually points to a mobile-specific rendering or usability issue rather than a ranking problem, since Google evaluates mobile and desktop results somewhat independently.

Should keyword research come from Search Console or a dedicated tool?

Search Console only shows queries you already have some visibility for; it cannot surface keywords you rank nowhere for. Use it to find striking-distance and decaying queries within your existing footprint, and use dedicated keyword research for new-topic discovery. Vrid.ai builds keyword research directly into its content workflow, so gaps found through GSC analysis and net-new opportunities both feed the same briefing process instead of living in separate tools.

What is the fastest way to catch a query cannibalization problem before it costs meaningful traffic?

Run the query-to-page breakdown check monthly on your top 20 non-branded queries by impressions, not just when you suspect a problem. Cannibalization often develops gradually as a site publishes adjacent content over months, and by the time rankings visibly bounce in the position column, both pages have usually already lost ground they would have kept had one URL owned the query from the start.

Key takeaways

Search Console’s default view answers one question, whether total traffic is trending up or down, and hides everything that actually explains why. The 16-month retention window, comparison mode, and query-to-page filtering are where the real diagnostic work happens, and almost none of it is visible until you go looking for it.

Treat CTR as position-relative, not a flat target: benchmark against the actual curve, roughly 27.6% at position 1 down to 2.8% at position 10 (Backlinko), before deciding a query underperforms. Confirm cannibalization with a page-breakdown filter before merging anything, since the panic around it is often bigger than the actual problem (Ahrefs). Run the page-level, six-month decay comparison monthly so you catch losing pages while a refresh is still cheap, not after they have fallen off page one.

If your content pipeline already needs keyword research to act on what Search Console surfaces, Vrid.ai connects that research to AI article generation with word-count control and publishing to WordPress, Ghost, Shopify, and webhooks, so a striking-distance query found this week can become a briefed, published update without a manual handoff.