Semrush ships dozens of tools, but keyword research lives almost entirely inside two of them — the Keyword Magic Tool and Keyword Manager. Learning how to use Semrush for keyword research means chaining those two deliberately: cast wide with Magic's billions-strong database, then let Manager cluster, score and track your shortlist automatically until it becomes a content roadmap.
The workflow below runs about an hour end to end and outputs clustered, prioritized groups instead of a chaotic ten-thousand-row spreadsheet. A free account handles the fundamentals under daily quotas; paid tiers lift the caps and add historical metrics.
Quick Answer: Enter a seed term or competitor domain, filter the Keyword Magic Tool by intent, difficulty and volume, send survivors to a named list, open that list in Keyword Manager for automatic clustering and personalized difficulty scores, then export the clusters as your publishing plan.
Step 1: Choose the Right Entry Point
Two doors lead into research. Keyword Overview accepts a seed term and returns volume, global distribution, intent, variations and questions — ideal when you know the topic but not its edges. Organic Research accepts a competitor domain and lists every keyword it ranks for — stronger when an obvious incumbent already defines the niche. Experienced researchers alternate between them: rival first for breadth, seeds second for depth.
Step 2: Filter Ruthlessly in the Keyword Magic Tool
Magic returns thousands of variations, pre-grouped on the left by shared words. The right-hand filters convert that flood into a workable shortlist:
- Intent: separating informational from transactional terms makes blog calendars and product-page targets fall out on their own.
- KD%: set an upper bound matching your authority; new domains typically live below 30.
- Volume and word count: four-plus-word filters surface long-tail phrases where newer sites earn their first rankings.
- SERP features and questions: featured-snippet eligibility and People-Also-Ask mining sit behind single toggles each.
Step 3: Build Lists Instead of Spreadsheets
Tick keywords and push them to a named list via the arrow control. Lists persist across sessions and feed directly into the clustering stage — resist exporting to Excel here, because you would be abandoning the machinery that does the hardest part for you.
Step 4: Let Keyword Manager Do the Grouping
Open your list in Keyword Manager and Semrush clusters semantically related keywords automatically, nominating a main keyword plus sub-groups per cluster. Each cluster displays combined search volume, dominant intent and a top-competing-pages preview — instantly revealing whether one strong pillar page can own the entire group or rivals fragment it beyond reach. Personal Keyword Difficulty recalibrates scores against your actual domain rather than an abstract average, and metrics refresh continuously while clusters remain saved.
Step 5: Export Something Publishable
- Export clusters to CSV or XLS with every metric intact.
- Assign one cluster per planned piece — main keyword earns the pillar page.
- Schedule sub-group articles beneath each pillar and interlink them as they publish.
| Feature | Role in the Workflow | Watch Out For |
|---|---|---|
| Keyword Magic Tool | Mass variation discovery with stacked filters | Raw output overwhelms without filtering discipline |
| Left-side topic groups | Pre-clustered themes ready to explore | Verify each group truly matches one intent |
| Keyword Manager | Automatic clustering with live-updating metrics | Keywords-per-list limits vary by plan |
| Personal KD% | Difficulty relative to your own domain | Still sanity-check the live SERP before committing |
Honest Constraints to Plan Around
Free accounts ration roughly ten analytics requests per day, which punishes exploratory browsing — decide your questions before logging in. Difficulty percentages are modelled and occasionally miscalibrated against weak SERPs, and intent labels come from automation rather than human review. None of this dents the core loop's value: no other suite moves from raw database to clustered, scored roadmap this smoothly.
Key Takeaways
- Start from a competitor domain when breadth matters; from seeds when depth matters.
- Intent and word-count filters do more damage than volume filters ever will.
- Lists feed clustering; spreadsheets kill it — stay inside the tool until Step 5.
- Treat KD% as a first filter and the live SERP as the final judge.