How to Use Google Search Console: The Reports That Actually Matter

Search Console has more reports in it than at any point in its history. Almost every guide responds to that by touring all of them, which is exactly the wrong instinct — the skill is knowing which four you open, and which numbers inside them will quietly mislead you.

You can read a tab-by-tab walkthrough anywhere. Performance, URL Inspection, Page Indexing, Sitemaps, Core Web Vitals, Enhancements, Links, Removals, Manual Actions, Security. Every guide lists the same ten things in the same order, explains what each one displays, and stops there.

The problem is that knowing what a report displays is not the same as knowing what question it answers, how often it's worth opening, or where its data will lead you to a confident wrong conclusion. This guide is about those three things instead.

The three questions Search Console answers well — and two it doesn't

Start here, because it prevents most wasted time. Search Console is genuinely excellent at three jobs and structurally incapable of two others that people keep asking it to do.

That last row causes real confusion in 2026. Search Console measures performance inside Google Search, Discover and Google News. Visits from ChatGPT, Perplexity, Copilot, Claude or the standalone Gemini app are ordinary referral traffic and live in your analytics platform — which is where you'd go to work through where AI search traffic actually lands. Teams look for those visits in Search Console, find nothing, and conclude AI search sends them zero. They were just in the wrong tool.

The Performance report is most of the value — and four controls do the work

If you only ever opened one report, this is it. But the default view — sixteen months of daily data for every query, unfiltered — is close to unreadable, and that's why most people bounce off it.

Four controls turn it into something useful, and two of them are recent enough that older guides don't mention them at all.

Branded versus non-branded, without the regex

Google added a branded queries filter in November 2025, and it is the single most useful change to the report in years.

Splitting branded from non-branded traffic used to mean writing regex filters by hand and maintaining them as your brand picked up misspellings and variants. It mattered because the two behave completely differently: branded search largely tracks your marketing and PR, while non-branded is what your SEO work actually moves. Reported together, a good quarter of SEO work can be entirely hidden by a dip in brand search, or vice versa.

If you have never separated the two, do it before anything else in this article. It routinely changes the story a team has been telling itself about its own performance.

Weekly and monthly views

Added in December 2025, and a quiet fix for a real problem: daily data is noisy enough that people react to nothing. A Tuesday dip becomes a meeting. Switching to weekly or monthly aggregation removes most of that and makes genuine trend changes visible instead of buried.

Use daily granularity only when investigating a specific incident with a known date. For everything else — reporting, trend review, anything you show another human — weekly or monthly is the honest view. That principle applies well beyond this tool, as we argued in building a dashboard leadership will actually read.

Compare mode, not just filters

Filtering tells you what happened. Comparing tells you whether that's unusual. Any number you're about to act on should be viewed against the equivalent prior period first — and if you're checking whether a drop is seasonal, compare against the same period last year rather than last month.

Query groups in Insights

Search Console Insights was rebuilt in 2025 and gained query groups that October, clustering related searches rather than listing them individually. It's the fastest way to see which topics you own versus where you have scattered coverage — a useful complement to a proper keyword research process rather than a replacement for one.

Featured Recommendation AD · AFFILIATE

4.6 / 5.0

Mangools

Simple and affordable keyword research suite loved by bloggers and freelancers

Best for: Keyword Research

Visit Website

View details Compare

The data traps that produce confident wrong conclusions

This is the part almost no guide covers, and it's where the real damage happens. Each of these is documented behaviour, not a bug — but each one has caused someone to report a number that wasn't true.

Average position is not your ranking. It is the impression-weighted average of every position you appeared at. A page ranking third for a niche query and fortieth for a high-volume one will report an average that describes neither. Read position only alongside a query or page filter, never as a single site-wide figure — and be very careful about putting site-wide average position in a client report, because it will move for reasons that have nothing to do with your work.

CTR without position context means nothing. A 2% click-through rate is excellent at position 9 and alarming at position 2. Always pull the two together. When you find genuinely low CTR at a strong position, that's usually a snippet problem rather than a ranking one, and it's fixed by rewriting title tags and meta descriptions rather than by building links.

The interface truncates at a thousand rows. The tables cap out, and they do it without a warning that says "there is more." Any site with real query breadth is looking at a slice, and the slice is not random — you're seeing the top of the distribution and missing the long tail, which is often exactly where the opportunity is.

Data lags, and the recent days are incomplete. The last two or three days will look worse than reality simply because they haven't finished processing. Do not investigate a "drop" that only exists in the most recent data points.

Your property type changes what you see. A URL-prefix property covers one protocol and subdomain combination; a domain property covers everything under the domain. If your numbers look implausibly small, check which one you're in before you check anything else.

The generative AI report, and the one thing it still can't tell you

On 3 June 2026 Google launched dedicated Search generative AI performance reports, sitting alongside the standard search results view inside Performance. After two years of guessing whether you appeared in AI answers, there's finally a number.

What it shows: impressions within generative AI features on Search — AI Overviews and AI Mode — plus generative AI features in Discover, broken down by page, country, device and date.

What it doesn't show is more important. There are no clicks, no click-through rate and no query data. And all three surfaces are combined into one view, so an impression could have come from AI Overviews, AI Mode or Discover and you cannot tell which. Rollout has been phased rather than universal, beginning with a subset of sites, so it may simply not be in your property yet.

The thing to understand about this report It is a breakout, not new measurement. That data was always inside your overall performance totals. Nothing about your historical numbers changed on June 3 — only your ability to see one slice of them separately.

That last point catches people out. Nobody's traffic changed when the report launched, and if your aggregate numbers moved that week, the cause was something else. For the wider context on how these surfaces reshape organic traffic, we covered it in what AI Mode means for organic traffic and what's working in AI Overviews.

Practically: treat AI impressions as a visibility signal, not a traffic one. Rising impressions with flat clicks tells you that you're being cited but not clicked, which is a real and increasingly common state — and a separate problem from not being cited at all.

Page Indexing and URL Inspection: use them as a pair

Page Indexing tells you how many pages are in or out and groups the reasons. URL Inspection tells you the story for one specific URL, including which canonical Google picked and when it last crawled the page.

The pattern that works: use Page Indexing to spot a category of problem, then URL Inspection on two or three examples to understand it. Going straight to URL Inspection means fixing pages one at a time; going only to Page Indexing means guessing at causes.

Two exclusion types deserve attention over the others. "Discovered — currently not indexed" usually means Google found the URL and chose not to prioritise crawling it, which is often a signal about site-wide quality or internal linking rather than anything about that page. "Duplicate, Google chose a different canonical" tells you Google disagreed with your canonical tag — worth investigating rather than ignoring, especially after a migration, where the redirect and migration checklist is the right companion.

Not every exclusion is a problem. Pages you deliberately noindexed appear here and are working as intended. A rising exclusion count is only bad if you didn't cause it.

Custom annotations: the most underused feature in the tool

Google added custom annotations to Search Console charts in November 2025, letting you mark your own events directly on the timeline.

This sounds minor and isn't. The single hardest question in search reporting is "did our change cause this, or did something else happen that week?" — and the usual answer is a scramble through Slack and deploy logs to reconstruct dates after the fact.

Annotate as you go: site migrations, redesigns, large content publishes, template changes, tracking changes, confirmed Google updates. Six months later, when a chart shows an inflection nobody can explain, the answer is sitting on the chart. This is also what makes it possible to separate your own doing from an algorithm update — the analytical problem at the heart of Google's 2026 core updates.

The reports you can check quarterly instead of weekly

Not everything deserves a recurring slot.

• Core Web Vitals — field data on a rolling window, so it moves slowly by design. Watching it weekly tells you nothing you couldn't learn quarterly. Fix it as a project, using something like our Core Web Vitals guide, then check it after.
• Links — a sample rather than a complete picture, and less complete than a dedicated backlink tool. Useful for a rough sense of which pages attract links; not a reporting metric.
• Enhancements and structured data — check after implementing markup and after template changes. Otherwise quarterly.
• Sitemaps — confirm it submitted without errors. There is nothing else to do here.
• Manual Actions and Security Issues — Google emails you if either fires. Checking weekly is pure anxiety management. Look when you get the email.
• Removals — an emergency tool for getting something out of results fast, not part of any routine.

Cutting these from your routine is not laziness. It's what makes room for the analysis that actually changes decisions.

A cadence that survives a busy month

Monthly, thirty minutes: Performance on a monthly view, branded and non-branded separately, compared against the prior period and the same period last year. Note anything that moved more than your normal variance. Add any annotations you forgot.

Quarterly, two hours: Page Indexing trend, Core Web Vitals, structured data, and the exercise that pays for the whole session — filter for pages with high impressions and low clicks, which is your list of pages that rank near the click threshold and need a better snippet or a better page. That list feeds directly into a content refresh programme, and it's usually cheaper than publishing new work. If the pages themselves are the weak point rather than the snippets, that's a content production problem more than a technical one.

On demand: URL Inspection when a specific page misbehaves. A full content audit when the quarterly review keeps surfacing the same pattern.

When you've outgrown the interface

Three signs, and a fix for each.

If you keep hitting the thousand-row limit, the Search Analytics API returns more than the interface will show you. If you need history beyond sixteen months, set up the bulk data export to BigQuery — it accumulates from the day you switch it on, which is the argument for doing it before you need it rather than when you do. And if you're rebuilding the same report every month, connect Search Console to a reporting tool once and stop.

For sites large enough that crawl behaviour itself is the question, Search Console's crawl stats are a starting point but log file analysis is the real answer.

The short version

Open Performance monthly with branded and non-branded split, on a weekly or monthly view, in compare mode. Use URL Inspection and Page Indexing as a diagnostic pair when something specific is wrong. Annotate your own changes as you make them. Treat the generative AI report as a visibility signal with no clicks attached, and remember it's a breakout of data you already had. Check everything else quarterly, and stop monitoring the reports that email you.

And be sceptical of every number that doesn't have a filter applied to it. Site-wide averages in this tool are almost always describing a situation that doesn't exist.