Seed keywords are the short, broad terms you feed into a research tool to generate everything else — one to three words describing what you sell, what problem you solve, or what category you compete in. “Running shoes.” “Payroll software.” “Roof repair.” They are not the keywords you target. They are the doors you open to find the ones you will.
Your seeds set the boundary of everything a research tool can return. Feed in three seeds and you will get one slice of the market. Feed in fifteen well-chosen seeds and you get the map. Most weak keyword lists trace back to a weak seed list, not to a weak tool.
How many seeds you actually need
Ten to twenty for a typical business, covering four distinct angles. Fewer than five and your research will be lopsided toward whatever corner of the market that handful happens to sit in. More than about thirty and you are mostly generating duplicates, since good research tools expand each seed into synonyms and related terms anyway.
The four angles to cover:
- Product and category terms — what you sell, in the words your industry uses: “crm software”, “hiking boots”
- Problem terms — what hurts before someone knows the solution exists: “losing track of leads”, “wet feet hiking”
- Audience and use-case terms — who and where: “crm for contractors”, “boots for wide feet”
- Competitor and alternative terms — brand names in your space, which unlock the comparison and alternatives clusters
Miss the problem angle and you will never reach the top-of-funnel audience that has not yet worked out what to buy. Miss the audience angle and you skip the long tail where a smaller site can actually win.
Where to find seeds that are not guesses
Guessing produces the terms you already think in, which are usually industry jargon your customers do not use. Six better sources:
- Your own Search Console. The Queries report shows what you already get impressions for. Terms sitting in positions 15–40 are pre-validated demand — Google already thinks you are relevant. Free, real, and specific to you.
- Customer language. Sales call notes, support tickets, live chat transcripts, the exact words in your reviews. This is where you learn your customers say “invoice app” when your marketing site says “receivables automation platform.”
- Competitor navigation. Open three competitors’ menus, category pages, and blog hubs. Their information architecture is a free market taxonomy, usually built after their own research.
- Google’s own suggestions. Autocomplete, People Also Ask, People Also Search For, and the related searches at the bottom of the page. These come from real query logs, which makes them more reliable than any modeled source.
- Marketplace and community language. Amazon category names, Reddit thread titles in your niche, question titles on Quora and specialist forums. Unpolished phrasing that matches how people actually search.
- Your paid search terms report. If you run ads, the search terms report shows queries that actually converted. Nothing beats that as a seed source.
Choosing seeds: the three tests
Not every candidate deserves a slot. Run each through three quick checks.
Is it broad enough to expand? “Waterproof hiking boots size 9 wide” is a target keyword, not a seed. It will expand into almost nothing. Seeds should sit one or two levels above the terms you plan to rank for.
Is it specific enough to stay relevant? “Software” as a seed for a payroll product will return thousands of irrelevant ideas and bury the useful ones. If a seed’s expansion is more than about a third off-topic, it is too broad.
Is it commercially connected? Every seed should trace to something you sell or a problem you solve. Seeds chosen because they look high-volume produce traffic that never converts — the classic mistake of a blog that gets 40,000 monthly visitors and zero customers.
Expanding seeds into a working list
Feed your seeds into a multi-seed explorer rather than running them one at a time. Running fifteen seeds together deduplicates overlap and surfaces the terms that connect several angles — usually the most valuable ones on the list.
You want each returned idea carrying volume, difficulty, CPC, global volume, and SERP features so you can filter without leaving the screen. A run of 150 ideas per search is plenty; the value is in the filtering, not the raw count. Set the country index to the market you actually sell to before you run anything — a Singapore business researching against the US index gets a list built for the wrong audience, and the numbers will look convincing while being irrelevant.
Then filter to your thresholds. On a young site that usually means difficulty under 25, volume between 50 and 1,500, and clear intent you can serve with one page. Save the survivors into a project keyword pool so they feed your writing and rank tracking rather than dying in a spreadsheet nobody opens again. SEO Rocket’s keyword explorer does exactly this at a flat $50 a month, with filtering, sorting, and CSV export.
A worked example: seeds for a small B2B tool
Take a company selling scheduling software to dental practices. A weak seed list would be “scheduling software”, “appointment software”, “booking system” — three product terms, one angle, and an expansion that returns results dominated by enterprise calendar platforms the company will never outrank.
A strong list covers all four angles and stays specific enough to keep the expansion relevant:
- Product: “dental scheduling software”, “patient booking system”, “appointment reminder software”
- Problem: “patient no shows”, “double booked appointments”, “front desk phone volume”
- Audience and use case: “software for dental practices”, “scheduling for multi location clinics”, “dental office front desk tools”
- Competitor and alternative: the three named platforms this company loses deals to, each of which unlocks a comparison and alternatives cluster
Twelve seeds, four angles, all commercially connected. The problem seeds alone will surface dozens of long-tail informational queries — “how to reduce dental no shows”, “no show policy template” — that a small site can realistically rank for and that reach buyers before they start comparing vendors. Those queries do not appear in any expansion of “scheduling software”, which is exactly why the angle matters.
What seeds cannot tell you
Two honest limits. First, every volume and difficulty figure attached to an expanded idea is an estimate — typically a modeled twelve-month average from periodic crawls. Useful for ranking terms against each other, unreliable as an absolute count, and blind to seasonality unless you check the trend line.
Second, expansion is only as good as the tool’s index. Genuinely new topics, brand-new product categories, and hyper-local phrasing often show zero volume simply because the sample missed them. If a “zero volume” term keeps appearing in your Search Console impressions, believe Search Console.
Refreshing your seed list
Seeds are not a one-time exercise. Revisit them each quarter and after any product change. New features create new seeds. Competitors entering the market create new brand seeds. Shifts in how people talk about a category — and the last few years have produced plenty — create new problem seeds that did not exist eighteen months ago.
A practical rhythm: pull your Search Console query report quarterly, look for terms you get impressions for that are not in your pool, and promote the recurring themes to seeds. That single loop keeps the list grounded in what people actually type rather than what you assumed they would.
