Keyword Research with Free Tools, step by step

What is keyword research, and what is it not?

Keyword research answers one question before you write a page: *what would the person I want to reach actually type into Google?* The answer is a short list of phrases, each matched to a page, so every page you publish has a job: to be the best answer for one specific search.

What it is not: stuffing a page with a keyword repeated twenty times. Google's ranking tutorial in this series already covered why that fails, ranking follows relevance and usefulness, not repetition. Research decides *which* phrase a page targets; the writing itself must then genuinely answer it.

It is also not a one-time task. You research before you write, then you check later, through Search Console, which the SEO course covers, whether the page is actually being found for the phrases you targeted, and you adjust.

How do I choose a seed topic?

A seed is a short topic in your own words, 1–3 words long, that describes what the page is about. It comes from the audience, not from a tool:

1. Name the audience. Who is this page for, and what are they trying to do? "Parents ordering a birthday cake" is an audience; "everyone who likes cake" is not.
2. Write their problem as they would say it. Not your business language ("celebration confectionery solutions") but their words ("birthday cake", "cake delivery", "cake price").
3. Pick one seed per page. If you cannot state the page's topic in three words, the page is trying to do two jobs, split it.

A good seed is broad enough to expand (it should generate dozens of related phrases) and specific enough to stay on topic (it should not generate phrases about unrelated products).

How do I expand a seed with Google's own suggestions?

Before opening any specialist tool, use Google Search itself. It is free, needs no account, and its suggestions come from real searches:

Step 1: Autocomplete. Type the seed slowly into Google's search box and read the dropdown suggestions before pressing Enter. These are completions real searchers have typed. Note the ones that match your audience's problem, and ignore the ones that do not.

Step 2: "People also ask". Run the search and read the "People also ask" box on the results page. Each question there is a search in its own right, and often a better page topic than the seed itself, because it is already phrased as a question someone wants answered.

Step 3: Related searches. Scroll to the bottom of the results page. The "related searches" list shows variations Google considers close to your query. These are strong candidates for supporting phrases on the same page.

Expected result: from one seed you now have a raw list of roughly 15–30 phrases, all in searchers' own words. Write them down exactly as shown, including the question forms. Do not tidy them into business language.

How do I compare phrases with Google Trends?

Google Trends (trends.google.com) is free and needs no account. It does not show how many people searched. It shows *relative* interest on a 0–100 scale, where 100 is the peak for the phrases and period you are comparing. That is enough for a beginner's decision, because you are comparing phrases against each other, not counting searches.

Step 1: Compare up to five phrases. Enter your seed and your best candidates together. Trends draws one line per phrase, so you can see at a glance which phrases are searched more, and which are barely searched at all.

Step 2: Check the shape, not just the height. A phrase that spikes once a year (birthday cakes before a holiday season, for example) is seasonal, useful, but plan its page ahead of the spike. A phrase with a flat, steady line is an everyday topic. A phrase falling steadily year after year is a topic in decline; prefer a rising or steady alternative.

Step 3: Check the region. Trends can break interest down by region. If your audience is local, a phrase that is popular nationally but flat in your city is the wrong target. The related-queries section will often suggest the local wording instead.

Expected result: your raw list is now shorter. Drop the phrases with almost no line at all, keep the steady and rising ones, and note any seasonal phrases for timing, not for immediate writing.

What does Keyword Planner add: and what does it not?

Google Keyword Planner is inside Google Ads, and opening it needs a Google Ads account. Opening an account does not mean running an ad. Planner is a research tool inside the account, and this tutorial uses it for research only, with no ad spend.

Step 1: "Discover new keywords". Enter your seed. Planner returns keyword ideas with a search-volume *range* (for example, a broad band like "100–1,000 searches a month") rather than an exact number, and a competition label (Low, Medium or High) that describes competition *among advertisers*, not how hard the phrase is to rank for organically. Treat both as rough signals, not measurements.

Step 2: Filter for relevance. Planner's ideas include phrases for advertisers selling different things. Delete anything your page could not honestly answer. A long list of loosely related phrases is worse than a short list of exact ones.

Step 3: Read it with Trends, not instead of it. Planner's ranges tell you roughly how big a phrase is; Trends tells you whether it is rising, steady or seasonal. A phrase with a wide range and a falling Trends line is less attractive than a smaller phrase with a steady line.

What Planner does not do: it does not tell you whether *you* can rank for a phrase. That judgement is the next two sections' work, and no free tool makes it for you.

How do I group keywords by search intent?

Search intent is what the searcher wants to *do*. Every phrase in your list belongs to one of four groups, and the group decides what kind of page should target it:

Why this matters: a guide page will not satisfy someone who wants to order, and an order page will not satisfy someone who wants to learn. When a page's content mismatches the intent behind its keyword, visitors leave immediately, and that is a ranking problem, not just a writing problem.

Group your list now: write each phrase in a row, add its intent, and check that each group has at least one page idea. A group with phrases but no page is a missing page; a page with phrases from two different intents is two pages pretending to be one.

How do I pick the winner for a page?

For each page, score its candidate phrases on four questions. No tool gives you a score, you judge, honestly, from what you can see:

1. Relevance. Could this page genuinely be the best answer for this phrase? If not, drop it, however big it looks.
2. Specificity. Prefer the longer, more specific phrase. It says exactly what the page is about, and it faces fewer competing pages.
3. Intent match. Does the page you plan to write match the phrase's intent group? A mismatch disqualifies the phrase for that page.
4. Winnability. Search the phrase and look at who ranks. If the first page is all large, established sites, a new page will struggle; if smaller blogs, local businesses or forums rank, the phrase is winnable. This is a judgement from the results page, not a number from a tool.

The winner is the phrase that passes all four. The runners-up become supporting phrases: use them naturally in headings and text on the same page, but do not build a second page around them unless their intent differs.

Worked example: keyword research for a practice bakery

The practice business is the one from the Digital Marketing pillar: an imaginary home bakery taking cake orders through Instagram and WhatsApp. It is a practice example, not a client and not a result. Follow the method, not the specific phrases. Your own phrases will differ.

Step 1: Audience and seed. Audience: parents in the bakery's own city, ordering birthday cakes a few days ahead. Seed: "birthday cake".

Step 2: Expand. Autocomplete, "People also ask" and related searches grow the seed into a raw list. Example phrases in the style you would collect (your own list will differ): "birthday cake delivery", "birthday cake price", "how to order a birthday cake", "birthday cake designs for kids", "birthday cake near me".

Step 3: Compare. In Trends, the delivery and price phrases show steady lines, the "how to order" phrase is steadier than expected for a question, and the design phrase spikes around celebration seasons, a seasonal note, not a rejection.

Step 4: Planner check. Planner confirms broad ranges for the delivery and price phrases and a High advertiser-competition label on the broadest phrases, a reminder that "High" measures ad competition, not organic difficulty, so it does not disqualify them on its own.

Step 5: Group by intent. The list splits cleanly: "how to order a birthday cake" is informational (a guide page); "birthday cake designs for kids" is commercial (a gallery/comparison page); "birthday cake delivery" and "birthday cake price" are transactional (an order page); the bakery's own name, when it has one, would be navigational.

Step 6: Pick. For the bakery's order page, the main keyword is the transactional phrase "birthday cake delivery" (specific, intent-matched, and, from the results page, contested by local bakeries rather than only national chains). "Birthday cake price" becomes a supporting phrase on the same page, in a pricing section. The informational phrase gets its own guide page later. It does not get stuffed onto the order page.

Expected result: three page ideas from one seed, each with one main keyword and an intent that matches the page. That table is the research output. Keep it, because the on-page SEO tutorial in this series starts from exactly this table.

Common mistakes

• Targeting the broadest word. "Cake" or "marketing" looks attractive and is contested by every established site in the field. A specific phrase you can actually rank for beats a broad phrase you cannot.
• Trusting Planner's competition label as ranking difficulty. It measures advertiser competition, not organic difficulty. Judge winnability from who actually ranks on the results page.
• Quoting free-tool volumes as exact numbers. Planner shows ranges, Trends shows relative interest. Reporting a range as "1,000 searches" invents precision the tool never gave you.
• Putting two intents on one page. A guide and an order page are different pages, even when they share words. When one page serves two intents, it serves neither well.
• Researching once and never revisiting. Search behaviour changes and new competitors appear. Re-check a page's phrases in Search Console after publishing, and adjust the page. The research table is a living document, not a certificate.

Frequently asked questions