Most teams pick blog topics by intuition and pay a keyword tool to confirm it. Meanwhile the property they already own is sitting on the only demand data that's actually about their site: every query Google chose to show them for, whether anyone clicked, and where they landed. That's what Search Console's Performance report is — a free topic engine, tuned to your own audience. The catch is that the best ideas are invisible at the defaults: the report opens on three months of data, sorted by clicks, which shows you what already works. Topic mining means flipping all three of those settings. Here's the sequence we run.
Open the right report and turn on all four metrics
In Search Console, go to Performance → Search results. The report tracks four metrics — clicks, impressions, CTR, and average position — but only some tiles are lit by default. Click each tile at the top so all four are on. Everything that follows depends on seeing impressions and position next to clicks: clicks tell you what you've already won, and the other three tell you what you could.
Stretch the date range to 16 months
The default view covers the last three months. Click the date filter and take it to the maximum — Search Console keeps 16 months of data and permanently deletes anything older. The long window matters for two reasons: it smooths out flukes, and it's the only way to see seasonal demand that a three-month view slices in half. If a query spiked last September, you want to be writing for it now, not discovering it in October. (Two caveats on the data itself: the most recent days are still filling in, and Google withholds rare, long-tail queries for privacy — so query rows will never quite sum to your page totals. Neither affects topic mining at this altitude.)
Sort by impressions, not clicks
In the Queries tab, click the Impressions column header. This is the single most useful re-sort in the product. Impressions are demand — searches where Google considered you relevant enough to show. A query with high impressions, few clicks, and an average position on page two is Google telling you: people keep asking this, I keep almost recommending you, and you haven't written the thing that would close it. Those rows are topic candidates, ranked by search volume you've already been measured against — no third-party tool estimate involved.
Pull the question queries with one regex
Click + New → Query above the chart and choose Custom (regex) from the dropdown. Search Console uses RE2 syntax, the filter field takes up to 4,096 characters, and matching is case-sensitive unless you prepend (?i). The workhorse pattern:
^(?i)(who|what|when|where|why|how|can|does|should)\b
That isolates every question your audience is typing. Questions map one-to-one onto content: a question with real impression volume and no ranking page is a post title waiting to happen, and clusters of smaller related questions become the H2s inside one post. A second pattern worth keeping: ([^" "]*\s){7,} surfaces queries seven-plus words long — long-tail phrasings that almost never appear in keyword tools but show you exactly how people describe the problem.
Let your pages tell you what deserves its own post
Switch to the Pages tab, click your highest-impression post, then flip back to Queries. You're now looking at every query that one URL ranks for — and somewhere in that list is a query the page answers in two sentences when searchers want two thousand words. That drift is a topic assignment: spin the query out into its own dedicated post and link the two. This is the closest thing to a self-refilling queue in SEO — every post you publish starts collecting evidence for the next one.
Compare date ranges to catch rising demand
In the date filter, use the Compare tab — last three months against the previous three. Sort the resulting table by impression difference. Queries growing quarter-over-quarter are demand curves you can get in front of while the SERP is still soft; queries collapsing tell you which planned topics to quietly drop. This is also the check that keeps a topic queue honest: intuition ages, comparison data doesn't.
Know the report's edges
The interface and its spreadsheet export cap out at 1,000 rows per table. For a small or mid-size site, that's plenty. If your property outgrows it, the Search Console API returns the deeper rows the UI hides — but that's an engineering ticket, not a prerequisite. The 1,000 rows you can see, re-sorted and regex-filtered as above, will keep a weekly publishing cadence supplied with evidence-backed topics for a long time.
From query list to publishing queue
The failure mode isn't finding ideas — after an hour in this report you'll have forty. It's that the list dies in a tab. The fix is operational: pull the winners into a prioritized queue with the query, its impressions, and its current position attached, so every topic carries the evidence that justified it, then work the queue on a fixed cadence. That's the difference between doing SEO research and running a content engine — and it's the part that doesn't require hiring. We've written up the math on 3x-ing content output without adding headcount and what an in-house content team actually costs per year if you want to see why the queue, not the writing, is usually the bottleneck.