Plan SEO topics on your Mac without juggling AI tabs

Bring a real customer problem into the overlay

Start with the offer you actually sell, who needs it, the stage of their decision and the language and market you serve. Add permitted, de-identified customer questions if you have them. A broad request for the best SEO keywords often produces a list that could fit any company in the category.

On macOS, the default Colay shortcut is ⌘⇧K; it can be changed in Settings. It opens the overlay. Paste the brief yourself and select available agents rather than assuming the app reads your browser or keyword tool. Keep your research sheet open underneath so each useful idea has somewhere to go.

Use different lenses to expose missing intent

Ask one pass to focus on first-time buyers, another on people comparing alternatives, and another on users trying to solve a specific problem. Compare the meaning of the questions, not just their wording. An agent suggesting appointment reminder software and another suggesting how to reduce missed bookings may be describing different journeys, even though both concern the same product.

For a fictional booking service, suppose you want to reach independent tutors. The table is an editorial planning example with no invented search-volume figures. A topic earns a place because the service can answer it credibly and the next step suits that reader.

Copy this prompt, then keep the disagreements

Ask separately keeps the initial proposals visible. A synthesis can help organize them, but retain disputed groupings in the research sheet. For example, a model may merge pricing questions with implementation questions because their vocabulary overlaps. Your content decision should follow what a reader needs to know, not the majority label.

Product and real capabilities: [brief]. Audience: [customer type]. Target language and market: [details]. Existing page list: [titles and URLs]. Generate search-question hypotheses through three lenses: learning a task, comparing options and checking product fit. For every idea give the underlying need, suggested page format, evidence the page would need, and a suitable next action. Do not invent search volumes, rankings or competitor facts. Group synonymous ideas. Flag ideas already covered by an existing page and disagreements about whether two questions need separate pages.

Add demand evidence outside the generated answer

Google's Keyword Planner documentation describes keyword discovery, average monthly search filters, and location and language settings. Record the settings alongside any figures you use. Do not compare estimates from different markets as though they measured one audience. Existing search-query data and actual customer questions can add context to the shortlist.

Inspect the current search results for promising queries and note which need they serve. Keep three separate columns: model suggestion, observed evidence and your editorial decision. A long generated list does not show that anyone searches for those exact phrases. If you cannot validate a candidate yet, mark it as untested rather than adding a plausible number.

End the session with a small publication queue

Compare model use against the credits and limits of your Colay plan. You do not need every agent on every keyword. Use additional perspectives where they reveal a missing intent or an unsupported assumption; spend the remaining time obtaining evidence and improving the page.

1. Merge candidates that answer the same reader question and assign one intended URL to each surviving topic.
2. Select a few topics where you can contribute a specific example, original explanation or useful decision aid.
3. Write a brief containing the reader's starting question, answer, supporting material, related page and next action.
4. After publication, review actual impressions, visits and registrations attributed to the page. Use those observations to revise the queue.

Questions, answered