How do I do keyword research for my startup?
Do keyword research for a startup by starting from the exact problems your customers describe, not the broad head terms you can't rank for yet. List the phrases buyers actually use, expand them with free tools (Google autocomplete, 'People also ask', Search Console, and the language in Reddit and support tickets), then score each phrase by search intent and how realistically you can rank. A brand-new site should target specific, low-competition long-tail queries (usually 3-5+ words) with clear buyer intent, group the winners into topic clusters, and publish the highest-intent, lowest-difficulty ones first.
Why keyword research is different when you're new
The mistake most founders make is chasing high-volume head terms — one or two words like 'CRM' or 'project management' — that established sites have spent years earning links for. A new domain has almost no authority, so it has near-zero chance of ranking for those, and going after them burns months for nothing. Keyword research for a startup is really the search for the specific, lower-competition phrases you can actually win now, then compounding from there.
- Start from customer language — the words buyers use for their problem — not the words you'd use for your product.
- Target long-tail queries (3-5+ words) with clear intent: lower volume, far lower difficulty, and closer to a decision.
- Score every phrase by intent and realistic difficulty, then write the highest-intent, lowest-difficulty ones first.
- You can do all of this with free tools — a paid SEO suite is a convenience, not a requirement, at this stage.
A 5-step keyword research process for founders
- Seed from customer language. Write down the problem your product solves in the words a customer would say it, then list every variation. Mine real sources for phrasing: Reddit threads, support tickets, sales calls, and the reviews on competing tools. These give you the informational and commercial queries buyers type before they know your category name.
- Expand with free tools. Feed each seed into Google autocomplete, the 'People also ask' and 'Related searches' boxes, and Google Search Console once you have any traffic. These surface long-tail variations and questions for free — no subscription needed.
- Classify by search intent. Tag each phrase informational, commercial, or transactional. Informational ('how to do keyword research') earns trust and AI citations; commercial ('best X tool') and transactional ('X pricing') are closer to a signup. A new site usually starts with the informational and commercial long-tail. See search intent.
- Estimate difficulty honestly. For each candidate, search it and look at page one. If it's dominated by big brands with deep backlink profiles, park it. If you see forum posts, thin pages, or outdated content, that's a gap you can win. Free browser extensions that show domain authority make this faster.
- Cluster into topic hubs. Group related keywords into a pillar topic plus its supporting questions, so you build topical authority instead of scattered one-off posts. One cluster, published thoroughly, ranks better than ten unconnected articles.
| Free tool | What it gives you | Best for |
|---|---|---|
| Google autocomplete | Real phrases people search, in order of popularity | Fast long-tail expansion |
| 'People also ask' / 'Related searches' | The questions around a query | FAQ and cluster ideas |
| Google Search Console | Queries you already get impressions for | Low-hanging near-ranking terms |
| Reddit / forums / reviews | The exact words buyers use | Seed phrasing and intent |
How to decide what to write first
You will end up with more keywords than you can write. Prioritize by a simple opportunity lens: favour phrases with the clearest buyer intent and the lowest realistic difficulty, even when the volume is small. Ten searches a month from someone ready to buy beats a thousand from people who will never sign up. Ship those quick wins first, watch which ones gain impressions in Search Console, and let that data steer the next batch.
Once you've found a repeatable pattern — say a keyword template like '[competitor] alternative' or '[use case] for [audience]' — you can scale it with programmatic SEO instead of hand-writing every page. And if you're starting from zero traffic, pair this with the fundamentals in how do I do SEO for a brand new website.
Keyword research is ongoing work, not a one-time project — which is why founders hand it to a marketing teammate. Ceres — the AI Growth Officer (agentceres.com) includes an SEO content specialist that researches keywords, drafts the pages, and flags what's worth publishing, so you review and approve the plan instead of running every search yourself.
FAQ
- Do I need a paid tool like Ahrefs or Semrush to do keyword research?
- No — not to start. Google autocomplete, 'People also ask', Google Search Console, and the language in Reddit threads and product reviews cover the essentials for free. Paid suites like Ahrefs or Semrush speed up volume and difficulty estimates and are worth it once SEO is a real channel, but a new startup can find winnable long-tail keywords without one.
- How many keywords should a startup target?
- Focus, don't spray. Pick one pillar topic and a handful of supporting long-tail queries to start — enough for a tight topic cluster of five to ten pages — rather than a spreadsheet of hundreds. Depth on a narrow topic builds authority faster than thin coverage of many, and it keeps a small team's output honest and high-quality.
- What is a good keyword difficulty for a new site to target?
- Aim low. With little domain authority, target long-tail phrases where page one shows forum posts, thin pages, or outdated content rather than established brands with strong backlink profiles. There's no universal number, but the practical test is: search the phrase, look at who ranks, and only commit if you can plausibly write something better than what's there.
- How is keyword research different for AI search like ChatGPT?
- The research overlaps but the target shifts. AI engines lift clear, quotable answers to specific questions, so question-shaped and definitional queries matter more, and structuring a direct answer at the top of the page helps you get cited. This is the discipline of generative engine optimization — see answer engine optimization.
Want this done for you?
Ceres is a managed AI marketing team — specialists draft the work, you approve what ships. 14-day free trial, from $19/month.