Most AI content fails because it starts with a vague topic, not a keyword with real demand. This is the keyword research method I use to decide what my agents write next.
Start with the searcher, not the topic
Ask: what would someone actually type? Real search phrases, not editorial headlines. Write down the question form, the problem form, and the how-to form of your topic.
Use the free tools first
- Google Autocomplete: type your seed phrase and read the suggestions - these are real queries.
- People also ask: mine those boxes for sub-questions.
- Answer boxes / featured snippets show what Google thinks the intent is.
Judge demand vs. effort
For each keyword, ask two questions: Is it worth ranking for? Can I realistically rank? A long-tail keyword with clear intent beats a broad head term for a new site.
Cluster into one article
Group 3-5 related questions into one in-depth article. One strong page that answers a cluster outperforms five thin pages, and matches how Google rewards topical depth.
Hand the cluster to the agent
Once you have a keyword cluster, your agent has a real brief: title, target keywords, outline sections mapped to each question. That is the difference between write about X and answer these specific searchers.
The honest metric
Track position, not vanity. If a page does not move after a few weeks, reassess the keyword or improve the page. Keyword research is a loop, not a one-time step.
That is the method: find the real queries, judge effort, cluster, then brief the agent precisely.
Top comments (0)