A new site can't outrank established publishers for broad terms like "indoor gardening" or "best credit cards". What it can do is answer narrow, specific questions that bigger sites have covered badly or skipped. Finding those questions is most of what keyword research means for a small site.
You don't need a paid tool to start. Google shows you what people search for through autocomplete, People Also Ask and related searches, and Search Console shows you what your own site already appears for. The step most people skip is opening the results page for each candidate query and reading what ranks there. That tells you more about your chances than any difficulty score.
One caveat before the process: no keyword worth having has zero competition. Something always ranks. The useful goal is a query where the current results are weak enough that a focused, well-made page has a fair shot. Nothing here guarantees a ranking.
Step 1: Collect the phrases people really type
Start with one seed topic your site covers and gather wording from four free places.
Google autocomplete. Type the seed and note the suggestions, then add a letter or a question word ("how", "why", "can", "without", "for") and note those too. Google says its autocomplete predictions reflect real searches, and that they're also shaped by your language, location and past searches. Use a private window so your own history doesn't skew the list.
People Also Ask. Search the seed and expand the question boxes. Each one you open loads more. These are often good article topics or section headings as written.
Related searches. The suggestions at the bottom of the results page tend to show the neighbouring angles people take on the same topic.
Forums and Reddit. Search your topic plus "reddit" or "forum" and read the thread titles. People describe problems in their own words there ("my basil keeps dying" rather than "basil care troubleshooting"), and that wording is often what they later type into Google. Threads with a lot of replies and no clear answer are worth noting.
Put everything in a spreadsheet. Don't filter yet.
Step 2: Check Search Console for queries you already appear for
If your site has been live for a while, Search Console is the best source you have, because it's data about your pages specifically. In the Performance report, open the Queries table and sort by impressions.
Look for queries where you get impressions but sit low on page one or on page two, and queries you appear for without having written a page about them. The first group may only need a better section or a clearer heading on an existing article. The second group is a list of articles Google already half-associates with your site.
Two limits to know about. A brand-new site will have little or nothing here, so come back after a couple of months of publishing. And Google leaves some queries out of the report to protect privacy, so the list is never complete.
If you haven't published anything yet, set up the site first. My guide to starting a blog that can sustain itself covers that, and you can add Search Console the same day.
Step 3: Work out what the searcher wants
Search intent is the reason behind the query. Most queries fall into one of these:
- Informational: they want to learn or fix something ("why is my basil wilting")
- Commercial: they're comparing before buying ("best grow light for herbs")
- Transactional: they're ready to buy or sign up ("buy basil seeds online")
- Navigational: they want a specific site or brand
The fastest way to tell is to look at what ranks. If the first page is all product roundups, Google has learned that searchers want a roundup, and a how-to won't fit there however good it is. Match the format that's already winning, then do it better. For commercial queries, that usually means the approach in my guide to writing buyer's guides that help readers decide.
Step 4: Judge difficulty by reading the results page
Difficulty scores are a rough filter. Each tool calculates its own, and the formula is usually narrower than the name suggests. Ahrefs, for example, bases its Keyword Difficulty on the referring domains pointing to the top 10 pages and says the score doesn't account for on-page factors. So a score can't tell you whether the ranking pages answer the question well.
Search each shortlisted query in a private window and open the top five or six results.
| What you see | What it suggests |
|---|---|
| Small blogs or sites like yours on page one | Reachable |
| Forum or Reddit threads ranking high | Often no strong dedicated page exists |
| Results that answer a broader or different question | A gap for a page on the exact query |
| Old, thin or clearly outdated pages | Room for a better one |
| Only big brands, each with a page on this exact query | Leave it for later |
Then read the pages themselves. Check whether they answer the question directly, whether the information is current, and whether you could add something they lack: clearer steps, a comparison, real photos, a worked example. If you can't name what your page would do better, pick a different query.
Also check how much of the answer appears on the results page itself. If Google already shows a full answer at the top, a number-one ranking may bring fewer clicks than you'd expect.
