Google Search Console guide: how to read the reports and act on them

Google Search Console shows how Google finds, indexes and presents your website, plus the searches that earned clicks. It won't fix anything on its own. Nor can it show which clicks became paying work. Start with the right property and a fair comparison period.

In practice, it answers three questions for a business:

• whether Google can find and index the pages that matter

• which searches show your pages, and which of them earn clicks

• what changed, when it changed and where

It can't count enquiries, calls or sales. That takes analytics and your own records. The tool costs nothing. The fixes it points to still take time or money.

Confirm the property before you read anything

Every chart depends on which property is open. A Domain property covers every protocol and subdomain. That can include a staging subdomain by accident. A URL prefix property covers only the address you entered, such as one version of the site. Check which one you're in before you trust a total.

The business should own its property, with the agency or developer added as a user. Then nothing is lost when a supplier leaves. Our audit sample is 100 website audits from May 2023 to September 2026. In about 1 in 4 audits, the business couldn't get owner access to its own property.

Our guide to setting up Search Console covers verification, property types and access. This guide is about reading what comes out. Two more things are worth knowing before you open a report:

• Start each visit on the Overview page. It flags warnings and broad changes. Some reports only appear when a site has the right data. So a missing report isn't proof of a broken setup.

• Platform properties now cover Instagram, TikTok, X and YouTube profiles. Google announced a global rollout in July 2026. They show how that content performs on Google Search, Discover and News, not inside each platform.

That's a chain of evidence, not a join between people. Each source answers a different question:

Search Console and GA4 will never match exactly. Search Console counts clicks from Google results. GA4 counts sessions, and only after consent and tagging have had their say. Our guide to reading GA4 reports covers the analytics side.

Read the four metrics and compare fairly

Four numbers drive the Performance report. Here's what each one means:

• Clicks: a click from a Google result through to your site.

• Impressions: a result from your site shown to someone. The counting rules vary by result type.

• CTR: clicks divided by impressions, multiplied by 100.

• Average position: the average of your site's highest position for each search. It isn't one fixed rank every user sees.

Google's page on impressions, position and clicks covers the exceptions. The tabs under the chart then split those numbers six ways:

• Queries: the searches that showed your site

• Pages: the URLs that appeared in results

• Countries: where the searcher was

• Devices: mobile, desktop or tablet

• Search appearance: special result types, such as rich results

• Dates: the same numbers, day by day

Learn the patterns below before you react to a change in any of them.

None of those patterns proves a cause on its own. Each one tells you where to look next.

Compare complete, equal periods

A 28 day comparison is practical. It isn't a rule. Whatever the length, keep the comparison fair:

• compare equal periods with the same mix of weekdays

• add year on year context where you have the history

• never compare a partial week with a full one

• leave out the newest days until they've filled in

Time zones trip people up too. Daily data uses Pacific Time. The 24 hour view uses your own browser's time zone. So a "yesterday" spike can sit on a different day than you expect.

Mark releases, tracking changes and outages with annotations. Owners and full users can add them. Everyone on the property can see them. Keep names and personal details out.

Annotations disappear after 500 days. They don't show in comparison or 24 hour views either. So keep a separate change log that lasts.

When total CTR falls but everything improved

Headline numbers can hide what happened underneath. Here's an example Australian service business. Each cell shows clicks out of impressions, with the CTR in brackets:

Both segments improved. Yet the total CTR fell by 0.16 percentage points. Most of the new impressions came from unbranded searches. Those carry a lower CTR. Clicks still rose by 248, or 88.6%.

So the total alone can't tell you a title got weaker. Never average the two segment rates either. Weight them by impressions, or the answer is wrong. Real data won't split this neatly, because privacy filtering hides some queries from both segments.

Connect queries, pages and intent

Queries describe searches, not people. Sort them by what the searcher wanted. Here's how that looks for an example Sydney plumber:

The last group isn't worth winning. A page that ranks for job seekers won't bring in jobs.

An unbranded query doesn't prove a new customer. Branded searches can also rise because of other marketing, like radio or a trade show.

Google's branded queries filter splits brand searches out for you. Eligibility depends on how much search data a site has.

An AI system classifies the queries. Google says it can make mistakes. So spot check its choices, and keep a reviewed regex filter (a saved text pattern) as a fallback. The filter is for reporting only. It doesn't change rankings.

Then reverse it. Start from an important page and check which searches reach it. If a page seems to be missing, check the canonical Google selected. That's the version Google treats as the main one. Search Console credits most performance data to that canonical URL.

Here's how that plays out. Say the plumber's hot water page loses clicks. Filter the hot water searches and open the Pages tab.

If a blog post now takes most of the impressions, compare what each page promises. The service page may need clearer signals. Or the blog post may need to link across and step back.

Two URLs getting impressions for one search isn't automatically cannibalisation. It only matters if the wrong page keeps winning, or the two keep swapping places.

Find and prioritise opportunities

This is where Search Console earns its keep for SEO. Each observation below leads to a check, not a blanket instruction to optimise:

Positions 6 to 20 aren't a literal page two. The layout changes by search, and averages blur the picture.

A low CTR doesn't always mean a weak title. Look at the actual result first. The page may sit below a map, ads or an AI answer.

Sometimes the title really is the problem. Our meta title and description checker helps you test new wording. Just remember a catchier title can also attract the wrong visitor.

Here's an example record. A plumbing services page earned 38 clicks from 2,400 impressions, a CTR of 1.58%. Another page earned 22 from 1,800, or 1.22%. A brand search earned 380 from 400, which is 95%.

None of those rates is good or bad on its own. Brand searches always run high. The useful question is which page could win more relevant work with a fixable change.

So rank issues in plain language: relevance, likely business value, confidence, effort and who can do the work. Don't invent a weighted score you can't explain. A fixable problem on a money page beats a big number on a blog post.

When a page is declining

Treat a declining page as its own investigation. Check these in order:

• URL or canonical changes

• indexing status for the page

• query demand, using year on year data

• new result formats, such as AI answers or maps

• new or stronger competitors

Broad declines can come from technical failures, not just search updates. An update date is a clue to investigate, not proof.

Use segments and the newer analysis features

Segments answer real questions. Did a service page lose mobile exposure in Australia? Compare mobile with desktop, and Australia with all countries, over equal periods. Country isn't suburb targeting. It doesn't say where customers live either.

Keep search types apart. Web, Image, Video and News results behave differently. Discover has its own report.

Since September 2026, Web results can be split into text based and multimodal searches. Multimodal covers searches that start with an image, like Google Lens or Circle to Search. It isn't the Images tab. Older exports and API integrations won't have the new split.

Insights and recommendations are prompts, not orders. The Insights report highlights trending pages and searches. Before acting, open the same pages in the Performance report for the same dates. A recommendation isn't a ranking requirement either.

Check AI powered configuration before trusting it

Search Console has an AI powered configuration for the Search results report. You type a request, such as comparing mobile clicks to the services pages across two months. It then sets the filters, comparisons and metrics for you.

Google calls it experimental and warns it can misread requests. It can't export or sort tables. Nor can it explain causes.

Read the generative AI report accurately

Google's page on AI features and your website covers traffic from AI Overviews and AI Mode. That traffic is included in the overall search traffic in Search Console. It's reported in the Performance report, within the Web search type.

Separately, Search Console's Generative AI performance report shows impressions (not clicks) from AI Overviews and AI Mode. It breaks them down by page, country, date and device. The report reached all websites worldwide as of 31 August 2026.

It still needs enough data before it appears for a site.

Don't add the AI impressions on top of your Web totals. They're already inside them. And an impression isn't praise. The report can't tell you whether an answer described the business accurately, or whether anyone enquired.

It also covers Google's own features only. Mentions in other AI tools, and mentions without a link, sit outside it. Our guide to getting cited in AI Overviews and AI Mode covers this report in depth. It also covers measuring citations more widely.

Google also added a Search generative AI control for site owners. It removes your content from AI Overviews, AI Mode and AI features in Discover. Opted out sites get no traffic or impressions from those features. That's an owner level business decision, not a fix for a low CTR.

Diagnose indexing without chasing every excluded URL

Start with the pages that should be indexed. That means useful public pages, in their canonical version. Duplicates, retired URLs and utility pages can rightly stay out. Google's page indexing report says not to expect every URL to be indexed.

So a 100% indexed figure isn't the goal. Watch for unexpected changes in the page groups that matter, such as services or products.

A common mistake is trying to get every variant indexed. Filtered pages, tracking parameters and old duplicates should stay out.

Two features make the report easier to use. The sitemap filter at the top shows only the URLs you submitted. That's usually the list that matters.

After fixing an issue, select Validate fix. Google then rechecks the affected URLs. That can take a couple of weeks or longer.

Inspect a URL properly

The URL Inspection tool first shows what Google has on record: the indexed version and its last crawl. The live test checks the page as it is now. Google says a live test only confirms its tool can access the page.

Read the parts in order. The declared canonical is the one you set. The selected canonical is the one Google chose. Then check whether crawling is allowed, whether indexing is allowed and what the fetch returned. If the live test followed a redirect, confirm the final address yourself.

A sitemap helps Google discover pages. A Success status means Google processed the file. It doesn't mean every URL got indexed. List only canonical, crawlable pages you want found.

Request indexing only after you've fixed the page. There's a daily limit on requests. And a request isn't a promise.

For a big batch of changes, update the sitemap instead. Fixing the page is in your control. Google's later decisions aren't.

Use the supporting reports selectively

The other reports answer narrower questions. Match the question to the report:

The Core Web Vitals report groups real user experience data by page type. Our guide to Core Web Vitals explained covers the thresholds and fixes. The Mobile Usability report and the Mobile Friendly Test retired on 1 December 2023. Mobile experience still matters. Google just stopped grading it there.

The HTTPS report covers indexed HTTP and HTTPS URLs for Domain and HTTPS URL prefix properties. It isn't a security audit. Enhancement reports only appear for supported structured data your site actually uses. Our guide to schema markup explained covers that markup.

The Links report shows a sample, not every link. It isn't an authority score either. So don't disavow unfamiliar domains just because they appear there. Manual actions and security issues are different matters, and both deserve prompt attention. A blank report doesn't prove the site has no problems.

Export without losing meaning

Downloads lose context fast. A few limits shape every export:

• Report tables show up to 1,000 rows.

• Private queries count in the chart but never appear as rows.

• Query filters change what's included, so filtered subsets may not add back to the total.

• Property and page totals are counted differently. Search appearance rows can overlap.

So don't sum overlapping rows or divide mismatched numbers. Reconcile the scope before you suspect a tracking fault.

Watch the blanks. Google's Performance report help says values shown as a dash or tilde become zeros when downloaded. A zero there isn't a real zero. Flag missing values before any spreadsheet maths.

Search Console keeps 16 months of performance data. So save baseline snapshots before they fall out of range, with the full filter record.

Bigger sites can go further. The Search Console API returns up to 50,000 rows a day per search type, sorted by clicks. That's more than the interface shows. It's still the top data, not every search.

The bulk data export to BigQuery, Google's cloud data warehouse, sends daily data for larger properties. It needs an owner, a billing check and a retention plan. It exports new data from the day it starts. It can't fill in older history. It still doesn't reveal anonymised queries.

Diagnose a traffic drop

Investigate before you assign a cause. Work through the same sequence every time:

Confirm the property, filters and dates first. Then check Google's list of data anomalies in Search Console. Sometimes the logging broke, not the site. A common mistake is blaming a Google update before checking any of this.

Next, find the shape of the drop. Check whether it hits:

• the whole site

• one directory or a handful of pages

• one country or one device

• one group of searches

The shape narrows the causes fast.

Check whether the site actually works before forming a theory. A failed release, a noindex tag or a broken redirect can look like a penalty. If a launch or URL move happened, our SEO migration checklist covers what to check.

Then compare demand, seasonal patterns and changes in the results themselves. If many pages fell on one day, don't assume an algorithm update. Our guide to Google algorithm updates covers how to test that idea.

Causes, in the order we check them

Each cause leaves its own clue:

Write down the competing explanations, and what evidence would separate them. Two causes often overlap, such as a seasonal dip landing in the same week as a release.

Run a monthly review and write a useful ticket

A monthly review keeps small problems small. Here's the cycle:

Each month, look at the same five things:

• Overview warnings and any emails from Google

• Performance for your money pages, against the same period last year

• Page indexing for the URLs you submitted

• manual actions and security issues

• open items in your opportunity register

Allow about 80 minutes as a planning figure, not a fixed time. Urgent problems, launches and big releases need checks between reviews.

Keep one opportunity register, not a new priority list each month. Every action gets an owner, any dependency, the evidence you expect and a review date. Separate "the change went live correctly" from "search results improved". The first is in your control. The second isn't.

Allow a result to stay inconclusive. If the data is thin, mixed or confounded, say so. Then choose a practical next observation. Search Console data shows search outcomes only. Our guide to measuring SEO results covers enquiries and sales.

Write a ticket a developer can act on

Vague tickets waste developer time. Here's an example for a page that dropped out of the index:

Attach real evidence only once you've collected it. And plan how to reverse the change if something critical breaks. Here's a reusable version:

Some problems turn out bigger than one template. Our SEO audit checklist covers the wider investigation.

Use a flexible 90 day plan

A new property, or a neglected one, needs a simple plan. Treat these as phases, not deadlines:

In month 1, fix access, the property scope and anything urgent, like a stray noindex tag. In month 2, make a small number of changes from the opportunity register. In month 3, review each one against the evidence you said would prove it.

Nobody can promise Google will reward the work inside 90 days. The plan makes sure the work itself is sound. Then the outcomes can be judged fairly.

Remember what Search Console can't show, too. A traffic gain means little if the page doesn't turn visitors into enquiries. Our guide to why your website is not converting covers that side.

What we recommend at Elev8d

FAQ

Next steps

Written by Ajay K.Co-founder, Elev8d

Ajay is the co-founder of Elev8d and writes most of what you will read here. Psychology grad turned marketer, which is a fancy way of saying he cares about why people click, not just where things rank. Every guide is written in plain English for Australian businesses, tested on real campaigns before it earns a spot on this site.