Keyword Research for a New Business Website

For a brand-new site, keyword research is not about finding high-volume terms. It is about finding the narrow, specific questions your buyers type that established sites have not bothered to answer properly. A page targeting 80 searches a month that you can rank for is worth more than a page targeting 8,000 that you cannot.

Bottom line

• Start from what your customers say, not from a tool. The tool validates and expands; it should not generate your strategy.
• Filter by intent and competition before volume. On a new domain, difficulty is the binding constraint.
• Group keywords into pages, not one page per keyword. Twelve solid pages beat sixty thin ones.

Start with the words your customers already use

Before you open anything, write down 20 to 40 phrases from real sources:

• Sales emails and support tickets. The exact wording of questions people ask you.
• Competitor navigation and pricing pages. What they call the thing you both sell.
• Reddit, Quora and niche forums in your market. Long, messy, specific phrasing — which is exactly what you want.
• Your own service list, written as a customer would say it rather than as you would bill it.

This seed list is the part nobody can automate for you, and it is why two businesses running the same tool get different results. A US LLC owner selling remote bookkeeping to e-commerce stores has a vocabulary that no generic “small business” keyword list contains.

The only metrics worth checking at the start

Keyword tools report a dozen columns. Four matter on a new site.

Intent. Is the searcher trying to learn, to compare, or to buy? Informational queries build authority and email lists. Commercial comparison queries (“X vs Y”, “best X for Y”) convert. Transactional queries (“buy”, “pricing”, “near me”) convert best and are usually the most contested. You want a mix weighted toward comparison.

Difficulty. Every tool has its own 0-100 score, calculated differently, so never compare scores across tools. On a domain with no backlinks, anything above roughly 30 is a long game. Below 20 is where a new site can realistically place.

Volume. Useful as a tiebreaker and nothing more. Treat reported volume as directional — it is modelled, not measured, and monthly figures for long-tail terms are noisy at best.

Cost per click. The most underused signal. A term with $12 CPC and 50 searches a month has buyers behind it. A term with $0.10 CPC and 5,000 searches mostly does not. If advertisers are paying real money, the traffic is worth something.

That last row deserves a note. If a query returns an answer box, a map pack and four ads above the first organic result, the organic click is worth much less than the volume suggests. Look at the actual results page before committing a page to it.

Using a tool without drowning in it

A paid tool earns its cost in three specific jobs: expanding your seed list into hundreds of variants, showing you which competitors already rank and with what, and telling you whether you have any chance against them.

Semrush is the suite most people end up on for this, and the workflow that matters is short:

1. Keyword Magic Tool. Enter one seed phrase. Filter to questions, then sort by difficulty ascending rather than volume descending. This single change in sort order is most of the value.
2. Keyword Gap. Enter your domain and three competitors. It shows terms they rank for and you do not. On a new site every row is a gap, so filter hard by difficulty.
3. Organic Research on one competitor. Sort their pages by traffic. You are looking for their best-performing topic, not their homepage.
4. Position Tracking. Add 30 to 50 chosen terms and leave it alone for a month. Keyword research without measurement is guessing with extra steps.

Semrush offers a free trial on its paid plans and a limited free tier, which is enough for one concentrated research week — start the trial, do the research in one sitting while it is active, and export everything to a spreadsheet you own. Pricing and trial terms change, so check the current pricing page before you sign up.

Free alternatives are genuinely usable for the seed stage: Google’s own autocomplete and “people also ask” boxes, Search Console once you have any impressions at all, and Google Trends for seasonality. What free tools do not give you is reliable difficulty scoring, which is the number that decides where you start.

Group keywords into pages

This is where most research dies. A spreadsheet of 400 keywords is not a plan.

Cluster by what a single page could answer well. If three keywords would be satisfied by the same article, they are one page — the phrasing variants are not separate targets, and splitting them creates two weak pages competing with each other.

A practical grouping for a service business:

• Money pages. One per service, targeting the commercial term plus its variants. These get your best writing and your internal links.
• Comparison pages. “X vs Y”, “best X for Z”, “alternatives to X”. Highest conversion rate per visit for most B2B sites, and often less contested than the money terms.
• Question pages. The long-tail how-to and definition queries. Cheap to write, build topical authority, feed email signups.
• Location or segment pages only if there is genuine differentiated content. Twelve near-identical city pages is a pattern search engines discount and readers notice.

Then order them. Publish the three lowest-difficulty, highest-commercial-value pages first — not the hardest one you care most about. Early rankings on easy terms produce the impression data in Search Console that tells you what to write next, and that feedback loop is worth more than any tool.

Check for overlap before you write. If two planned pages would both answer “how much does X cost”, merge them. Self-competition is the most common self-inflicted SEO problem on small sites.

The hosting decision, briefly

Keyword research is wasted if the pages load slowly or the site goes down during your first traffic. Page speed is a real, if modest, ranking input, and it is a large conversion input.

For a content-driven business site on WordPress, managed hosting removes the two failure modes that actually bite: a slow server under a traffic spike, and a site you break while trying to optimize it yourself. WP Engine is the standard choice at the professional end, with caching and a staging environment included so you are not testing changes in production. It costs more than shared hosting, which is the honest trade. Our fuller comparison is in best web hosting for your US LLC website.

If you are not on WordPress, none of this applies and a static host is probably better and cheaper. The point is only that the publishing platform should not be the bottleneck once the research is done.

Common mistakes on a new site

Targeting head terms first. “Business insurance” is not available to you. “Business insurance for a single-member LLC with no employees” might be.

Trusting one tool’s difficulty score as truth. Open the results page and read the top five. If they are all thin affiliate roundups, the score is overstating the difficulty. If they are all major publications with original data, it is understating it.

Writing for volume rather than for the buyer. Traffic that cannot buy from you is a cost, not an asset.

Ignoring your own Search Console. Within weeks of publishing, it shows real queries you are being shown for. Those are better leads than any tool’s suggestions, because they are measured rather than modelled.

Publishing 60 thin pages in a month. Quantity without depth dilutes the site. Fewer, better, internally linked.

Never revisiting. Research is not a one-off project. Re-run the gap analysis every few months and update the pages that are ranking on page two, since those are the cheapest wins you will ever get.

FAQ

How many keywords do I need to start?

Enough to fill 10 to 20 pages, which usually means 40 to 80 keywords after clustering. More than that and you are planning further ahead than your publishing capacity.

Do I need a paid tool at all?

Not for the seed stage. You need one for competitive difficulty and gap analysis, which is where guessing gets expensive. A single month of a paid plan, used intensively, covers the initial research for most new sites.

How long until a new page ranks?

Realistically months, not weeks, and low-difficulty terms move first. If nothing has moved after six months on genuinely easy terms, the problem is the page or the site, not the timeline.

Should I target my competitors’ brand names?

Comparison pages that name competitors fairly and accurately work well and convert. Pages that misrepresent them do not, and create risk you do not need. Keep claims specific and sourced.

This is general information about marketing, not legal or tax advice — and if your site publishes anything touching US taxes or entity compliance, have a CPA or attorney review those pages before they go live.

If you want the research done properly in one pass, Semrush is the tool to run it in, and the trial is enough to get your first twenty pages planned. If the pages then need somewhere reliable to live, WP Engine is the version of that decision you will not have to revisit.