n8n - automated keyword research

The context

Finding the right keywords to target takes a tedious routine: running searches in Semrush, exporting tables, cross-checking them against the site's current rankings, discarding what is off-topic or too competitive, then grouping the rest by theme. Useful work, but repetitive, and so it ends up being done too rarely.

The idea: hand this routine over to an n8n workflow that runs on its own every week and only delivers what matters, a short list of high-potential keywords, already sorted and grouped.

The workflow

1. Scheduled trigger: the workflow starts every Monday morning.
2. Starting points: a Google Sheets spreadsheet lists the "seed" keywords (the site's topics) and a few competitor domains to keep an eye on.
3. Semrush data: for each seed, the Semrush API returns related keywords with their search volume, difficulty, CPC and trend; for the site, the positions it already holds.
4. Clean-up: duplicates removed, keywords outside the market or with no volume discarded, data brought into the same format.
5. Scoring: each keyword gets a potential score (see below).
6. AI grouping: the selected keywords are sent to a language model, which groups them by topic and search intent, and suggests for each group the type of page to create or improve.
7. The deliverable: results are written to Google Sheets, and a summary of the top five opportunities arrives by email.

Selection criteria

A "high-performing" keyword is not necessarily the one with the biggest volume: it is the one the site has a realistic chance of winning, and that brings in the right people.

• Search volume: enough demand for the effort to be worth it.
• Difficulty: below a realistic threshold for the site's authority.
• Search intent: informational, commercial or transactional, depending on what the site can actually offer.
• Current position: a keyword already on page two (positions 11 to 20) is often the quickest win.
• Trend: rising demand comes before demand that is fading.

These criteria combine into a single score, so very different keywords can be compared on one scale. AI does not decide the score: it comes in afterwards, to make sense of the list by grouping it.

The approach

• A repetitive SEO routine turned into a workflow that runs on its own, every week.
• Semrush data cross-checked against the site's current rankings, to aim for realistic wins rather than big volumes.
• A readable potential score, then AI to group keywords by topic and intent.
• A short, actionable deliverable: an up-to-date spreadsheet and an email with the best opportunities.