Keyword Research

Discover and cluster keywords with intent and difficulty signals pulled from the live web into a saved file.

When to use

A content or SEO (search engine optimization) plan needs a fresh list of target keywords for a topic, product, or competitor space.
You need to expand a seed keyword into related terms, questions, and long-tail variants grouped by intent.
Someone asks for a content calendar or brief and there is no existing keyword research file to build from.

browser-search: run SERP (search engine results page) queries against seed keywords and topic phrases to surface related searches, autocomplete-style suggestions, and competing pages.
browser-navigate: open the search result pages, competitor articles, and keyword-tool pages returned by search.
browser-extract: pull the visible keyword lists, "people also ask" questions, related searches, headings, and on-page copy from the pages you navigate to.
write-file: save the clustered keyword list, with intent and difficulty notes, to a file for downstream content work.

Confirm the seed topic, target audience, and any exclusions with the task details before starting.
Run an initial query for the seed keyword using browser-search to see the current SERP landscape.
Use browser-navigate to open the top-ranking pages and any "related searches" or question panels shown in the results.
Use browser-extract on each opened page to capture headings, question phrasing, and repeated terms that signal related keywords.
Repeat browser-search for 3-5 keyword variants (synonyms, "how to", "best", "vs", buyer-intent phrases) to widen coverage.
Group the collected terms into clusters by intent (informational, commercial, navigational, transactional) and note apparent difficulty based on how many high-authority pages dominate each SERP.
Draft the final table (keyword, cluster, intent, estimated difficulty, source URL) as markdown or CSV text.
Save the draft with write-file to a clearly named keyword-research file.
Summarize the cluster count and top opportunities when completing the task.

Search results return 401/403 or a blocked/consent page: treat as a dead end for that query, note it, and continue with the next seed term via browser-search.
browser-extract returns an empty or near-empty result: re-check you navigated to the right URL and that the page finished loading before extracting; retry once.
Rate limited or repeatedly blocked by a search engine: back off, reduce query frequency, and record the limitation via update-memory so future runs know to pace themselves.
A results page layout has changed and expected sections (e.g., related searches) are missing: fall back to plain browser-search queries on narrower phrasing instead of relying on page structure.

A keyword-research file exists (via write-file) containing clustered keywords with intent and difficulty notes.
Each cluster has at least one supporting source URL captured during research.
The task is completed with a summary of top clusters and recommended priority terms.