How to Find Seed Keywords: A Practical Guide

A seed keyword is the broad one-or-two-word term a whole area of your content hangs off. Not the thing you expect to rank for tomorrow, and often not a thing you write a page about at all. It is the root the rest of the plan grows from.

Most keyword research skips this step and starts in a tool, sorted by volume. That produces a list of terms that are either too broad to win or too disconnected from each other to compound. The work below is what to do first.

Three places your seed keywords already exist

You do not start by generating candidates. You start by collecting the language that is already in the building.

What you sell, in two vocabularies

Write out everything you offer. For each one, note four things: what it does, what problem it removes, the term you use internally, and the term a customer uses. The gap between the last two is where most of the value is.

A security company talks about threat detection; the person with the problem searches for website security or malware protection. Both belong on the list, but only one of them gets typed into a search box.

What customers actually say

Support tickets, sales call notes, chat transcripts and reviews are the only unfiltered source of customer vocabulary you have. Read a quarter's worth and write down the recurring nouns.

People describe their problem in words that rarely match your product page, and those words are seed keywords with the intent already attached. Don't go by what customers say they want in a survey. Go by the words they used when they were annoyed enough to open a ticket.

What your own data already knows

Two reports repay the time immediately. Search Console, filtered to queries with impressions but few clicks, shows terms you already surface for and lose. Something is close and not close enough. Your internal site search log shows what visitors could not find once they were already on the site. That is stated intent from an audience that already knows you exist, and almost nobody reads it.

Where to get the data

Every tool below answers a different question, and the free ones cover the first pass completely.

Google Search Console. Which queries you already surface for, and at what position. The impressions-without-clicks filter is the single most useful view for seed discovery, because it shows demand you are already adjacent to.

Google Analytics. What people did once they arrived. Useful for judging whether a seed brings the kind of visitor who stays, which a volume column cannot tell you.

Google Ads Keyword Planner. The official volume source, and the one the estimates in every other tool are ultimately calibrated against. Export and work on the data elsewhere; the interface fights you.

Bing Webmaster Tools. Genuinely good now, and underused. Its keyword tool carries volume, competition, question phrasing and newly created terms, which is a category the Google tools do not expose directly.

Looker Studio. For joining the above into one view you will actually open again next quarter. Google gives you all the pieces and joins none of them.

Reddit. Not a keyword tool, which is why it works as one. Find the subreddit your customers actually use, sort by top posts, and take the nouns out of the titles. It is the same unfiltered vocabulary as a support ticket, from people with no reason to be diplomatic.

Mapping seeds to the buying stage

The same topic needs different seeds at different points, and the mistake is treating one seed as though it serves the whole journey.

Stage	Intent	Seed	What it expands into	Format
Awareness	Informational	digital marketing	what is digital marketing, digital marketing guide	Guides
Consideration	Commercial	marketing tools	best marketing tools, marketing software comparison	Comparisons
Decision	Transactional	marketing agency	hire marketing agency, marketing agency pricing	Service and pricing pages
Retention	Support	email marketing	email marketing best practices, improve open rates	Tutorials

A plan built only from the transactional row has nothing to say to anyone who is not already buying. A plan built only from the awareness row brings traffic that never converts.

Not too broad, not too specific

The hard part is the width of the seed. Too broad and you are competing with national publishers on a term that describes your whole industry. Too specific and you have a keyword, not a seed, with nothing to expand into.

Working on keywords for a management coaching client, the obvious seeds were "coaching" and "career". Both are unwinnable and neither says anything. The seeds that worked came from the pain points instead: leadership development, executive presence, navigating office politics. Each one is narrow enough to enter on and broad enough to hold a dozen pages underneath it.

That is the test. If you cannot list five real pages under a seed, it is too specific. If you cannot imagine beating anyone on page one, it is too broad.

Who you are actually competing against

Your business competitors and your search competitors are different lists. For any given term you compete with whoever is on page one, and that is often industry publications, universities, large media sites or a forum thread — organisations that do not sell what you sell and cannot be beaten with a better product.

Look at page one for each candidate seed before committing to it. If the first ten results are all national publishers, the seed is real but you need a narrower one to enter on.

Semrush and Ahrefs both do gap analysis: feed them the domains that keep appearing for your topics, and they list terms those domains rank for and you do not. Treat the output as a candidate list, not a plan. The volume and difficulty numbers in every one of these tools are estimates from different sample sets, and they disagree with each other on the same term. Use them to rank candidates against each other, not as measurements.

Expanding a seed into a topic

Once a seed holds up, it expands along predictable axes. Working through them systematically produces better coverage than free association.

By audience. By industry (digital marketing for lawyers), by role (marketing tools for small business), by experience (beginner digital marketing guide).

By place. City, region, country — but only where the service or the answer genuinely differs by location.

By intent. How to do X, best X services, buy X. Three different pages, not three headings on one.

By time. Trends this year, seasonal campaigns, quick wins. These date, which is the point: they are worth writing when the date is the reason someone searched.

Once you have two or three lists of modifiers, crossing them by hand is tedious and you will miss combinations. The Word Combinator multiplies the lists out in one pass, which turns three seeds and a list of industries into a full candidate set.

Questions

Question phrasing keeps growing as a share of search, and questions are also what an AI answer engine is built to match. Run each seed through three shapes:

• Definition: what is X, how does X work, why does X matter.
• Comparison: which X is best, X versus Y, what types of X are there.
• Process: how to choose X, when to use X, where to find X.

Each of these is a real page if you have a real answer. None of them is worth writing if the answer is three sentences padded to eight hundred words.

Turning the list into a structure

The point of expansion is not more pages. It is a set of pages that reinforce each other.

Build one broad page for the seed itself, covering the whole topic at a level that stands alone. Then build individual pages for each expanded term, each answering one question properly. Link every cluster page up to the broad page, and link the broad page down to all of them.

The structure does two things. It gives a search engine an unambiguous signal about which page is your answer for the broad term. And it gives a reader who arrives on a narrow question a route to the rest of the topic, which is the part most people building clusters forget to check.

What to do this week

Pull one quarter of support tickets and one quarter of Search Console queries with impressions and no clicks. Write down every recurring noun in both. Cross out the ones you have already covered properly.

What is left is your seed list, and it came from your own customers rather than a volume column.