How should you read your Search Console data every 28 days?

October 21, 2026

Read Search Console the same way every 28 days. Export the queries and pages, split brand searches from the rest, sort what is left into position bands, compare each band's click-through rate with your own average, group queries by topic and intent, and read the generative AI report next to it. Each band points to one kind of fix, so the read ends with a short list of work, and the next read tells you whether that work helped.

I started doing this for AI Agents Live because the numbers were confusing without a method. I had thousands of impressions, a handful of clicks, and a different conclusion every time I opened the dashboard. A fixed routine solved that. This article is the routine, step by step, so you can run it on your own site.

Before you start: export and save

Open the Performance report, set the date range to the last 28 days, and export the queries and pages. Do the same for the generative AI report if your property has it. Save every export in a dated folder, because Search Console keeps 16 months of data and you will want to compare against older periods later.

Know the limits of what you exported. The interface export stops at 1,000 rows per table. Rare queries are held back for privacy, so the query rows add up to less than the totals in the chart. Google explains both in its performance data deep dive. If you need more rows, the Search Console API or the bulk data export to BigQuery gives you them.

Step 1: keep what people can see

Sort queries and pages by impressions and keep those at an average position of 50 or better. Impressions show where Google already finds you relevant, and past position 50 almost nobody scrolls far enough to see the result.

Step 2: split brand from non-brand

Brand queries are searches for your name and its spellings. They come from people who already know you, so they say little about whether your content works. Report brand and non-brand clicks separately, and treat non-brand clicks as the number that matters most.

Step 3: sort queries into position bands

Each band has its own fix.

Positions 1 to 10 with a low click-through rate are a click gap. The page ranks, and the title and description fail to win the click. Rewrite the snippet.
Positions 8 to 20 are striking distance. Better on-page content and internal links can move them onto the first page.
Positions 20 to 50 are deep. The page is related to the query but answers a different question. Add a section, write a new page or rebuild the old one.
Queries with impressions and no matching page of yours need a new page of their own.

Step 4: build your own click-through curve

Calculate clicks divided by impressions for each position band, for non-brand queries only. Public averages vary a lot with AI Overviews and other result features, so compare your pages with your own curve. A page well under your own average for its band is a snippet problem.

Step 5: group by intent

Sort queries into navigational (a product or company name), commercial ("review", "vs", "alternative", "pricing", "best"), informational ("how", "what is", long questions) and transactional ("buy", "sell", "hire"). Each intent maps to a page type: a listing or product page, a comparison, a blog post, a landing page.

Step 6: group by topic and add up the impressions

A topic with many small queries is worth more than any single query in it. Give each topic a hub page and supporting posts that link to it. Skip any single query under about 20 impressions as a project of its own.

Step 7: answer long questions in the first lines

Long question queries are the ones most likely to show an AI Overview. Put the question in a heading and answer it in the first two sentences under it, so your page can be the source the answer quotes.

Step 8: read the generative AI report next to the web report

Google's generative AI performance reports count how often a link to your page appeared in AI Overviews and AI Mode, by page, country, device and day. They count impressions only, with no clicks, queries or positions.

Read them three ways. Look at the weekly trend (daily numbers are too noisy on a small site). Look at which pages are cited, since that is what AI answers trust. And look for pages that get AI impressions and web impressions but no web clicks: the answer is being given from your page, so add a reason to click, such as a comparison table or a tool.

Step 9: find the AI sub-queries

AI Mode runs many searches of its own while answering one question, and Google calls this query fan-out. Some of those searches show up in your queries export as long, formal questions nobody would type. Count them apart from human queries, so each topic counts only what people searched, and read them as a list of what AI answers want to know about you.

Step 10: compare with the last period

From the second read on, compare the last 28 days with the 28 days before. Once you have enough history, compare 3 months with the 3 before, and with the same 3 months a year earlier to remove seasonality. Flag pages whose impressions fell more than about 20% or whose position dropped by 2 or more. A position drop usually comes before a click drop.

Turn the read into a short list

End every read with five lines in a notes file: the date, non-brand clicks, blog clicks, generative AI impressions, and what went live since the last read. Then pick the three fixes with the most impressions behind them.

Give each change four to six weeks before judging it, and compare the same 28-day window before and after. Change one thing per page at a time, so you can tell what worked.

Questions people ask

How often should I check Google Search Console?

A full read every 28 days is enough for most small sites, with a quick look at the Page indexing report after big releases. The data runs two to three days behind, so start each read a few days after the period ends.

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

It depends on position, the type of query and the result features on the page, so compare with your own average for each position band. A page well below your own average for its band needs a better title and description.

Why does Search Console show impressions but no clicks?

Usually the page ranks too low to be seen, or an AI answer or another result answers the question first. Check the position band and the generative AI report to tell which one it is.