Search Console is the only tool that tells you what Google already thinks your site is about. Not what a third-party crawler guesses, not what a rank tracker samples from one location, but the actual queries your pages appeared for and where they sat. It's free, it comes straight from the source, and most owners open it once a month, look at whether the line went up, and close it again.
This guide is the other use of it: reading it deliberately, in a fixed order, to produce a ranked list of things to fix. Nine steps, roughly ninety minutes, and four filter recipes you'll reuse every quarter.
Every screen here's a recreation, not a screenshot. Search Console sits behind a Google login and blocks automated capture, so the interfaces below are rebuilt in the site's own styling. The numbers inside them aren't invented: they're this site's real Search Console figures from the 2026-08-06 snapshot, including the parts that are unflattering.
Set the date window so it can tell you something
Open Performance, then Search results. The default view is three months. Before you read a single number, do two things:
- Set the range to Last 3 months, then choose Compare and pick the previous period.
- Turn on all four metric toggles: clicks, impressions, CTR and average position. Only the first two are on by default.
The comparison is the point. A single window tells you the state of things; a comparison tells you the direction, and direction is what you act on. The default 28 day window is too short for a small site, because a handful of impressions either way will look like a trend when it's noise.
Two things to know before you trust any date on this screen. Search Console reports in California local time, so its days don't line up with your analytics days. And data arrives on a delay, with Google saying collected data should normally be available in two to three days, which means the last couple of days on every chart are incomplete rather than genuinely down.
How to know it worked Your chart shows two lines per metric, solid for the current period and dotted for the previous one, and all four metric toggles are lit. If you only see one line, the comparison isn't on.
Common trap Panicking at the cliff on the right-hand edge of every chart. That's the two to three day reporting delay, not a collapse. Read the trend up to about three days ago and ignore the tail.
Read the four numbers without fooling yourself
Clicks are people who arrived. Impressions are times you appeared. CTR is clicks divided by impressions. Average position is where your topmost result sat, averaged across everything.
The useful move isn't reading them individually, it's reading them as a pair of diagnoses:
- Low impressions means a visibility problem. Google isn't showing you. Fixing your titles won't help, because nobody is seeing them.
- Decent impressions, low clicks means a click problem. You're being shown and passed over, which is a titles, descriptions and intent-match problem.
Those two need opposite work, and mistaking one for the other is the most expensive misreading on this screen.
- Total clicks
- 7 1
- Total impressions
- 1,193 2
- Average CTR
- 0.6%
- Average position
- 78.4 3
Now the trap that catches almost everyone, and the reason the number above is 1,193 rather than the 668 you'd get by adding up the query rows.
Google omits some queries from the query table to protect privacy. Its documentation calls these anonymized queries: rare queries, and queries containing personal or sensitive information. They are still counted in the chart totals, but they never appear as rows. So if you export the query table and sum the impressions column, you'll always get a smaller number than the real total.
On this site the gap isn't small. The true date totals are 1,193 impressions and 7 clicks. Summing the query rows gives 668 impressions and 2 clicks, which hides about 44% of impressions and five of the seven clicks. If I reported the query sum as my traffic, I'd be understating it by nearly half.
There's a second edge to this. The moment you apply any query filter, the anonymized queries drop out of the chart totals as well. So a filtered chart total and an unfiltered one aren't measuring the same universe.
How to know it worked You can say in one sentence whether you have a visibility problem or a click problem, and you have written down the unfiltered totals from the chart rather than a sum of the table.
Common trap Presenting a per-query sum as your site's traffic. It's always low, sometimes by half, and once you have quoted it to someone you're stuck with it.
Find the queries you almost rank for
This is the highest-value filter in the tool and the first of the four recipes. You're hunting for queries sitting in positions 8 to 20: real demand, page one or two, close enough that a title rewrite can move them.
- In the Queries tab, add a filter on Position, greater than 7.
- Add a second filter on Position, smaller than 20.
- Sort by impressions, descending.
- Copy every row into your spreadsheet under a heading of striking distance.
What you have is a list of queries where Google already considers you a plausible answer and simply prefers someone else slightly more. Those are cheap wins compared with anything sitting at position 60.
| Query | Clicks | Impr. | CTR | Position |
|---|---|---|---|---|
| report fraudulent seo company1 | 0 | 14 | 0% | 19.9 |
How to know it worked You have a list, even if it's one row long, where every query is between position 8 and 20 and you can name the page that ranks for it. If the list is empty, note that as the finding and move on: it means you have no cheap wins available and the work is further upstream.
Common trap Treating brand terms as striking distance. Your own business name ranking at position 12 is a different problem, and rewriting a title won't fix it.
Find demand you have no page for
Recipe two. Same screen, opposite end of the range: queries with real impressions sitting past position 20. Google has validated that people search this and that it's somehow relevant to you, and you're nowhere.
- Filter Position, greater than 20.
- Sort by impressions, descending.
- For each row, ask one question: do I have a page that deserves to rank for this?
- Label every row either page exists, needs work or no page exists.
That labelling is the whole exercise. The first group is a rewriting job. The second is a content plan, and it's a content plan built from demand Google has already confirmed rather than from guesses.
| Query | Impr. | Position | Verdict |
|---|---|---|---|
| google maps seo company1 | 143 | 80.4 | Page exists |
| google maps seo vaughan | 116 | 56.0 | Page exists |
| google maps seo toronto | 48 | 80.5 | Page exists |
| google maps seo services | 38 | 88.2 | Page exists |
| my google business profile isn't working | 28 | 54.3 | Page exists |
| apple maps seo services2 | 11 | 68.5 | No page |
| vaughan seo3 | 7 | 83.4 | No page |
How to know it worked Every row in your buried list carries one of the two labels. A row you can't label means you don't yet know your own site well enough, which is itself worth knowing.
Common trap Writing a new page for a query where you already have a page sitting at position 80. You now have two weak pages instead of one, which is exactly the problem in the next step.
Find the pages competing with each other
Recipe three, and the one nobody runs. Two of your own pages chasing the same query means Google has to pick, and it frequently picks the weaker one.
- Stay in the Queries tab and click a single query to filter to it.
- Switch to the Pages tab, still filtered.
- If more than one of your URLs is listed, that query is being split.
- Repeat for your top ten queries by impressions.
It's tedious and it's worth it. Here's what it found on this site.
| Page | Clicks | Impr. | Position |
|---|---|---|---|
| /google-maps-seo-vaughan1 | 2 | 43 | 1.9 |
| /local-seo2 | 0 | 41 | 94.0 |
| /google-maps-seo-toronto | 0 | 32 | 79.9 |
The fix is almost never to delete a page. It's to decide which page owns the query, then remove the competing signals from the other one: change its title and headings so it stops targeting that phrase, and link from it to the page that should win.
How to know it worked You can name at least one query where two of your URLs appear, and for each one you have written down which page should own it. Most sites have several. Finding none on a site with more than a handful of pages usually means you didn't check enough queries.
Common trap Assuming the highest-ranking page is the one that should win. Sometimes the general page outranks the specific one purely on age and internal links, and the specific page is the one that would actually convert.
Find rankings nobody is clicking
Recipe four, and the cheapest work on the entire list. You're looking for queries where you rank respectably and still get no clicks, which means the ranking is fine and the listing isn't.
- Filter Position, smaller than 11.
- Sort by impressions, descending.
- Read the CTR column. Anything in the top ten with a CTR near zero is a candidate.
- Open the page that ranks and read its title and meta description as a stranger would.
On this site the standout is the brand term itself. rmcm shows 18 impressions at position 6.8 with a 0% click-through rate. People are searching the name, seeing the site, and not clicking, which is a brand SERP that hasn't been locked down rather than a ranking problem.
Don't chase a specific CTR number here. Expected CTR varies enormously by query type and by how much of the page Google fills with other features before your result. Use the comparison of similar queries on your own site as the benchmark, not a table you found online.
How to know it worked You have a shortlist of pages to rewrite titles and descriptions for, and for each one you can say what a searcher wanted that the current listing doesn't promise.
Common trap Rewriting for keywords instead of for the click. The page already ranks. The job here's to make somebody choose it, which is a copywriting problem.
This is the analysis the Search Roadmap runs across sixteen months of your data, with a written ninety day plan on the end.
SEE THE SEARCH ROADMAPCheck indexing before you act on any of it
Everything above assumes your pages are in the index. Before you spend a week rewriting titles, confirm that.
- Go to Indexing, then Pages.
- Read the indexed count, then open the table of reasons underneath.
- Look specifically for money pages under Discovered, currently not indexed or Crawled, currently not indexed.
Those two states are different. Google describes Discovered as a page it found but hasn't crawled yet, often because crawling it then would have overloaded the site. Crawled, currently not indexed means it looked and chose not to index, and Google explicitly says there is no need to resubmit the URL.
That last point is worth sitting with, because the common reflex is to hit Request Indexing repeatedly. Google is telling you that does nothing. A page that was crawled and rejected usually needs to be more distinct or more useful, or needs internal links pointing at it so it looks worth indexing. It doesn't need to be resubmitted five times.
How to know it worked You know exactly which of your commercially important pages Google is refusing to index. This finding outranks everything else in this guide. A page Google won't index can't rank no matter how good its title becomes.
Common trap Reading the total indexed count and stopping. The total is meaningless on its own. What matters is whether the specific pages that make you money are in it.
Check whether you appear in AI results
Search Console gained reporting on AI surfaces in 2026, which means the question clients keep asking is finally answerable with data instead of opinion.
Two newer things worth opening while you're here:
- The 24-hour view, which shows the last day with an hourly breakdown instead of waiting the usual two to three days. Useful for checking whether something you shipped this morning is being seen at all, and not much else.
- The AI performance reporting, covering impressions in AI Overviews and AI Mode. This tells you whether you're being shown inside AI answers, which is a completely different question from whether AI assistants send you traffic.
Keep the two separate in your head. Appearing in an AI Overview is an impression. Somebody clicking through from ChatGPT is a referral, and that lives in your analytics, not here.
These reports are new and still moving. Confirm the exact report names and layout when you open them, because this part of the interface has changed twice in 2026 and will likely change again.
How to know it worked You can state whether you appear in AI surfaces at all, and you have that number written down separately from your ordinary search impressions.
Common trap Reading an AI impression as a visit. It's the same relationship as a normal impression to a normal click, except the click is much less likely to happen because the answer is right there.
Rank the list
You now have five lists: striking distance, buried demand, cannibalization, low CTR, and indexing problems. Ranking them is the actual deliverable, and the order is driven by dependency rather than by effort.
Within each band, sort by impressions. Volume is the only tiebreaker you have that isn't a guess.
For what it's worth, running these nine steps on this site produced eighteen documented items, and the top of the list wasn't a content idea. It was four money pages sitting in Discovered, currently not indexed, plus a general page quietly outranking three specific pages it should have been feeding. Neither of those would have shown up in a rank tracker.
How to know it worked A numbered list where item one is small enough to start this week and every item names the query or page it came from. If any row doesn't cite its evidence, it's an opinion that snuck in.
Common trap Starting at the bottom because new pages are more fun to write than fixing an indexing problem. The bottom of this list is where enthusiasm goes to die.
STRIKING DISTANCE Position > 7 AND Position < 20 sort by impressions
BURIED DEMAND Position > 20 sort by impressions
CANNIBALIZATION filter to one query, switch to the Pages tab
LOW CTR Position < 11 read the CTR column
Always read totals from the unfiltered chart, never from a sum of the table.Finding | Recipe | Query or page | Impressions | Position | Band | Fix order | DoneThe whole review on one page
The monthly version of this
The nine steps above are the full review. You don't need the full review every month, and running it monthly is the fastest way to stop running it at all.
Do the whole thing quarterly. In between, the monthly check is the one in the article this pairs with: five outcome numbers, ten minutes, written down and compared to last month. That routine watches whether search is producing work. This guide is what you run when the answer is no, or when the quarter is up.
The part nobody writes down is what to do when a monthly number moves. Most people either ignore everything or panic at noise, and the difference between the two is a rule set decided in advance, when you're calm and not looking at a graph.
ACT THIS WEEK
a money page left the index
form or phone stopped producing leads entirely
clicks down >40% month on month, no known cause
business details wrong on the profile
site down, slow to the point of unusable, or
showing a security warning
These are breakages. They do not recover on
their own and every week costs money.
NOTE AND WATCH (needs 3 months, not 1)
position drifting a few places
impressions up, clicks flat
one query gained or lost
a competitor appearing above you
traffic down but leads steady
These are noise until they repeat. One month
is not a trend, and reacting to it is how
people rewrite pages that were fine.
IGNORE
daily rank fluctuation
a single lost keyword
total traffic, on its own
THE RULE
A number that moved once is weather.
The same number moving three months running
is climate. Act on climate.
Before acting, write down what you think
changed and why. If you cannot name a cause,
you are guessing, and guessing costs a page
rewrite you did not need.Where this caps out
The nine steps are entirely doable alone, and if you run them once a quarter you'll be ahead of most businesses in your market, who aren't opening this tool at all.
What ninety minutes can't give you is weighting across time. This is one window. Reading sixteen months, watching which queries decayed and which are quietly climbing, separating a seasonal dip from a real decline, and knowing which of eighteen findings will actually move revenue in your specific market: that's a different exercise, and it's mostly pattern recognition from having done it repeatedly.
It's also worth being honest about scale. On this site, nine steps produced eighteen findings from 1,193 impressions. On a site with real traffic, the same nine steps produce hundreds, and the hard part stops being finding problems and becomes choosing between them.
That's what the Search Roadmap is: this analysis across your full history, with the findings ranked and a ninety day plan written on the end. If you'd rather run it yourself, everything you need is above.
Questions people ask
Why does Search Console show fewer clicks than my analytics?
They count different things. Search Console counts clicks from Google Search results only, and it removes duplicates and filters bot traffic. Your analytics counts sessions from every source, including direct, referral and social. Search Console also reports in California local time, so a day boundary in one tool isn't the same day boundary in the other. Expect them to disagree, and never try to reconcile them exactly.
What's a good average position?
The question doesn't really work, because average position mixes every query and every page into one figure. A site averaging position 12 across two hundred queries and a site averaging position 12 across four queries are in completely different situations. Read position per query instead, and treat the site-wide average as a trend line rather than a score.
How much Search Console data does Google keep?
Sixteen months of Performance data, which is why exporting matters. If you want to compare this spring to the spring before, you have to export before the older window rolls off. Data also arrives with a delay: Google says collected data should normally be available in two to three days, so the last couple of days on any chart are incomplete rather than bad.
Why do the query totals not add up to the chart totals?
Because of anonymized queries. Google omits rare queries and queries containing personal or sensitive information from the query table to protect privacy, but still counts them in the chart totals. So summing the query rows always gives you a smaller number than the real total. The moment you apply a query filter, the anonymized ones drop out of the chart totals too.
My site is new and has almost no data. Is this still worth doing?
Yes, but change what you expect from it. On a new site the useful signals are indexing coverage and which queries Google has decided you're relevant to at all, even at position eighty. Striking distance and low CTR analysis need volume to mean anything. Set a baseline now, and the same nine steps get sharper every quarter as data accumulates.
