How Do I Know What Keywords to Target on My Website?

Start from what you can win

Keyword targeting goes wrong in one predictable way: starting from the biggest term in the niche and working down. The biggest terms are where you are weakest, so the plan starts with months of nothing.

Reverse it. Start from the searches you can actually win, most of which are already in your own data, and climb from there. The order below is deliberate: the sources are ranked by how quickly the targets they produce turn into clicks. Three sources produce the list:

Source 1: your own almost-rankings

Open Search Console, filter queries to position 8 to 30, sort by impressions. These are terms Google already shows you for, just not high enough to earn clicks. They pay fastest because the trust half of the work is done.

This is also the one source every site has from day one of having traffic, and the one most people skip while paying for tools. Which keywords your website ranks for walks the reading in full.

Source 2: questions customers ask

Sales calls, support tickets, onboarding emails, and the forums where your customers gather. The words they use are the searches they type, and those searches carry buyer intent that no volume number captures. Type your topics into Google and read the autocomplete suggestions: that is live demand, spelled in customer language, free.

Source 3: competitor keywords

A keyword tool's competitor report shows what already works in your market: proven demand, mapped to pages. Treat it as a menu, not a mandate, because you inherit the competition along with the term, and why competitors rank higher shows how to size that inheritance honestly. Our competitor analysis checklist turns that list into decisions rather than envy.

The four-question filter

The sources produce candidates, usually far more than you can serve. The filter turns candidates into targets, and it is deliberately strict: a keyword passes all four questions or it drops. Strictness here is what keeps the list short enough to actually work.

The second question is where honesty earns its keep. Search the term and read page one as a stranger: if every result is a major brand, your page is not real competition yet, and the term goes on the someday list, not the target list. The SERP preview tool helps you see how your listing would look in that company.

Write the list down, with URLs

A target list that lives in your head is a mood. The written version has four columns and forces the useful decisions:

• The keyword, in the customer's words.
• The planned URL: the one page that should rank for it, existing or to be written. One target, one page, which is also the cure for keyword cannibalization.
• The intent: learn, compare, or buy, so the page matches the format page one rewards.
• Priority, set by demand times winnability, not by which term sounds most impressive.

CrawlRaven keeps this list as a first-class object: targets carry their planned URLs, intent groups, and priority, and imported Ahrefs or Semrush lists land in the same workspace, so research you already paid for keeps working.

A worked example: picking one target

Say you run accounting software for freelancers, and Search Console shows you at position 14 for "invoice template for freelancers" with 900 impressions a month. Run the filter:

• Demand? Yes: 900 impressions is real, and autocomplete confirms people type it.
• Winnable? Page one is a mix of template galleries and blog posts, not just giants. Position 14 says Google already rates you. Yes.
• Right intent? Searchers want a template, not an essay. You can offer a free downloadable one. Yes.
• The page? You have a blog post nearby, but the winner needs to be a template page. Planned URL: a new one.

Verdict: target it, build the template page, link the old post to it. That is the whole method, run once. Twenty of these decisions is a quarter's plan, and each one takes about five minutes once the filter is habit.

Track the plan against reality

The list is a plan, and plans drift. Once a quarter, put it next to what Search Console actually returns and ask three questions: which targets show progress, which show nothing after a fair run, and what are you ranking for that was never planned.

That last group deserves special attention, because it often contains your best next targets: demand you proved by accident. With the MCP server connected, the whole comparison is one prompt:

Compare my target keywords for [your site] with the queries Search Console actually returns. Which targets show no sign of ranking yet, which are progressing, and what am I ranking for that was never in the plan?

Find queries for [your site] between position 8 and 30 with real impressions that are not in my target keyword list, and rank them by how many clicks reaching page one would be worth.

The three targeting mistakes

Most targeting failures are not research failures. They are one of three habits, and each is fixable the day you notice it:

• Targeting your own vocabulary. You sell "workforce enablement"; customers search "staff training software". The customer's words win every time.
• Chasing volume you cannot convert. A thousand visits from a definitional query can be worth less than fifty from a comparison query. Intent decides value, not volume.
• Never retiring targets. A term that has shown nothing after two honest quarters is telling you something. Move it to someday and spend the effort where the data says you are close.