serpStart with a purchasing situation
A useful seed describes something a customer needs to accomplish. A category name can help, but it often returns a mixture of students, job seekers, consumers, and buyers. Before opening a keyword tool, write down the event that makes someone look for your product, the task they need to finish, and the constraint that makes their current solution inadequate. This is an editorial method, not a Google ranking rule.
For a hypothetical scheduling product, those notes might read: a practice adds a second location, reception staff coordinate appointments manually, and patients need reminders. That creates several separate research directions. Starting only with “calendar” would hide the purchasing situation under a broad category. Starting with the actual job makes it easier to reject irrelevant suggestions later.
Build three separate seed lists
Create one list for the category, another for the problem, and a third for meaningful constraints. Keep the lists separate during the first research pass so you can identify where useful ideas originated. Product-category phrases may uncover comparison pages; problem phrases may uncover tutorials; constraints may uncover high-fit use cases.
For the scheduling example, category seeds could concern appointment software, problems could concern missed appointments, and constraints could concern multiple locations. These are hypothetical seeds, not verified keywords. Use customer interviews and support language to refine them. Do not add every adjective you can imagine. A modifier earns its place when it changes the required answer, the buyer, or the product capability being evaluated.
Inspect the tool output before expanding it
Keyword Planner accepts words and website inputs. Treat its suggestions as candidates rather than a ready-made publishing calendar. Review a sample from each seed family, record irrelevant themes, and look at actual search results for promising phrases. If a seed mostly produces consumer calendar templates, adding more variations of that seed will probably multiply the same mismatch.
Maintain a short rejection log with reasons such as wrong audience, unsupported feature, hiring intent, or purely navigational search. This log becomes especially useful when AI helps expand the list: provide the exclusions before generating additional ideas. Save the original seed and source beside every accepted phrase so another person can reproduce the research instead of inheriting an unexplained spreadsheet.
Turn the shortlist into a small experiment
Choose a few candidates that represent different buyer problems, then decide what a useful page would contain. A multi-location scheduling guide might need a workflow diagram and an explanation of staff permissions. A missed-appointment guide might need a reminder policy and practical examples. If you cannot describe the useful evidence, the topic is not ready for a brief.
After publishing, compare the queries that actually produce impressions with the seed assumptions. Revise the seed library when the audience differs from your expectation. A seed list is a working hypothesis about customer language, not a permanent taxonomy. The output of this exercise should be a smaller, explainable set of topics with a clear reader and a reason your business can help.
Keep a seed only when you can explain the customer situation behind it and the useful page it could produce.
Sources & further reading
Official documentation informs this guide. Examples are illustrative, and product interfaces or limits can change.
From search signals to a useful first draft.
SERP.so is in development. Join the waitlist for a workspace that connects research, content review, and publishing.
Join early access ↗