To learn how to do keyword research, start with what your business sells, collect the words people actually type, check what Google already ranks for them, and only then pick the terms your site can realistically win. Tools give you numbers. The search results page tells you what to build.
Key takeaways
- Keyword research is a decision process, not a list. The output is "which page targets which query", not a spreadsheet of 2,000 terms.
- Always look at the live search results before you commit. Volume and difficulty scores are estimates; the SERP is evidence.
- Judge difficulty against your own site's authority and content, not against a tool's 0–100 score alone.
- Group keywords that share the same results into one page. One topic, one URL.
- Free tools (Google Keyword Planner, Search Console, Google Trends, autocomplete) are enough to start. AI chatbots help with ideas, not with search volumes.
Why keyword research still matters

Keyword research tells you what your audience calls things, how many of them search for it, and what kind of page Google thinks answers it. Skip it and you end up writing pages nobody searches for, or ten pages chasing the same query.
Google's own SEO Starter Guide puts it plainly: think about the words a user might search for to find your content. The same guide also warns that keyword stuffing breaks Google's spam policies. So the job is to understand language and demand, not to repeat a phrase.
AI Overviews and chat-style search have changed how answers are shown, but they haven't removed the need to know what people ask. If anything, the questions people type are longer and more specific now, which makes the research more useful, not less.
How to do keyword research in 9 steps

This is the process I use for client sites and for this blog. It works for a new blog, a service site or a small online shop.
Step 1: Start with the business, not the tool
Before opening any tool, write down:
- What you sell or teach (products, services, topics).
- Who buys it, and what problem they're trying to solve.
- What a "conversion" is for you: a sale, an enquiry, a sign-up.
From that, write 5–15 seed keywords. These are short, plain phrases a customer would use: "soy candles", "epoxy flooring", "keyword research". Don't overthink them. They're starting points.
My rule of thumb: if you can't say which page on your site a seed would lead to, or which product it would sell, park it for later.
Step 2: Expand the seeds into a raw list
Now widen each seed. Use several sources, because each one sees a different slice of demand:
- Google autocomplete. Type the seed into Google and note the suggestions. Add "how", "best", "for", "vs" and letters of the alphabet after it.
- Google Keyword Planner. Enter your seeds under "Discover new keywords". Google describes it as a free keyword search tool for advertisers, but you need a set-up Google Ads account to use it.
- A keyword tool (Ubersuggest, SE Ranking, Semrush, Ahrefs or similar). Pull "related", "questions" and "also ranks for" lists.
- Google Search Console, if your site already gets impressions. The Performance report shows the actual queries your pages appear for. This is the most underrated source of keywords you'll ever get, because it's your own data.
- Forums and communities like Reddit, Quora and niche Facebook groups. Look at the exact wording people use when they ask for help.
Dump everything into one sheet. Don't filter yet. You'll usually end up with a few hundred rows.
Step 3: Add the metrics (and know what they mean)
Next, pull metrics for every row. Each tool calculates them a little differently, so pick one tool for the whole list and stay consistent.
| Metric | What it tells you | What it doesn't tell you |
|---|---|---|
| Search volume | Estimated average monthly searches in a chosen country | How many clicks you'll get. AI Overviews, ads and other features can take a large share |
| Keyword difficulty (KD) | A tool's estimate of how hard it is to reach page one, mostly based on the links of ranking pages | How hard it is for your site. It's a generic score |
| CPC | What advertisers pay per click; a rough signal of commercial value | Whether organic visitors will convert on your page |
| Trend | Whether interest is rising, falling or seasonal | Absolute numbers. Google Trends scales data from 0 to 100 relative to all searches |
| Search intent label | The tool's guess at informational, commercial, transactional or navigational | The exact page format Google prefers. You confirm that in Step 4 |
Two cautions. First, search volume is an estimate. Keyword Planner itself describes its monthly numbers as estimates, and accounts with low ad spend may see limited data. Second, low-volume terms are not worthless. A query with "10" searches a month can still bring a buyer every week.
Step 4: Look at the live search results
This is the step most people skip, and it's the one that decides what you'll build.
Search each shortlisted keyword in a private window (and in the country you target). Note:
- Page type. Are the top results blog guides, product pages, category pages, tools or videos?
- SERP features. Is there an AI Overview, a video block, "People also ask", a forum section, a local pack?
- Who ranks. Big brands only, or smaller sites like yours?
- Freshness. Are the ranking pages recent or years old?
If Google shows product listings and you planned a 2,000-word guide, you've picked the wrong format. Matching the dominant format is the core of search intent, which I explain properly in my guide to search intent.
Step 5: Judge keyword difficulty against your own site
A tool's KD score is a starting filter. The real question is: can my page beat these pages?
I check three things on the top results:
- Authority gap. Compare their domain and page-level link metrics with yours. If every result has far more referring domains than your whole site, park the keyword for now.
- Content gap. Are the ranking pages thin, outdated or missing something obvious? Weak content on page one is an opening, even on a "hard" keyword.
- Topical fit. Does your site already have related pages? A site with ten solid pages on one topic usually ranks faster for the eleventh than a site writing about it for the first time.
For a new site, I usually start with keywords where at least one or two smaller sites already rank on page one. That tells me it's possible.
If you want a deeper method for picking apart who ranks and why, see my SEO competitor analysis guide.
Step 6: Group keywords into topics
Now cluster the list. Keywords belong on the same page when Google shows mostly the same results for them. "How to do keyword research" and "keyword research steps" usually share results. "Keyword research tools" usually doesn't; it wants a tool list.
A simple manual method:
- Sort the list by your main seeds.
- For each group, search the two or three biggest terms.
- If the top 10 results overlap heavily, keep them together. If they don't, split them.
Each group becomes one page, with one primary keyword and several supporting ones. Specific, lower-volume phrases often sit inside these groups as supporting terms; I cover how to find and use them in my post on long tail keywords.
Skipping this step is how sites end up with two or three pages competing for the same query. If that has already happened on your site, my guide to keyword cannibalization shows how to find and fix it.
Step 7: Prioritise
You can't write everything at once. I score each topic group on four questions, 1 to 3 points each:
| Question | 1 point | 3 points |
|---|---|---|
| Business value | Loosely related to what I sell | Directly leads to a sale or enquiry |
| Winnability | Page one is all major brands | Smaller sites rank; content is weak |
| Demand | Very low or falling | Steady or rising |
| Effort | Needs original data, tools or design work | I can write it well this week |
Add up the points and work from the top. This keeps the plan honest: a high-volume keyword with no business value and no chance of ranking drops to the bottom on its own.
Step 8: Map keywords to pages
Write a simple keyword map: one row per URL.
| URL | Primary keyword | Supporting keywords | Page type | Status |
|---|---|---|---|---|
| /existing-page/ | ... | ... | Guide | Update |
| /new-page/ | ... | ... | Category | To write |
Check existing pages first. Often the best move is to improve a page that already gets impressions instead of writing a new one. Then put the primary keyword where it counts: the URL, the title tag, the H1 and the opening lines. After that, write for the reader. My content writing strategy covers how I turn a mapped keyword into a brief and a draft.
Where keyword research fits into your wider plan (goals, technical fixes, links, reporting) is a separate job; my SEO strategy template covers that.
Step 9: Publish, measure in Search Console, repeat
Keyword research doesn't end at publishing. Four to eight weeks after a page goes live, open Search Console:
- Filter the Performance report by the page.
- Look at the queries it gets impressions for.
- Find queries you didn't target but that keep appearing. Add a section or an FAQ for them, or plan a new page if they need a different format.
Note that Search Console leaves out rare queries to protect user privacy, so the totals in the query table won't always match the page totals. That's normal.
Free vs paid keyword research tools
You can do everything above with free tools. Paid tools mainly save time and add competitor data.
| Tool | Free or paid | Best for | Limitation |
|---|---|---|---|
| Google autocomplete | Free | Real phrasing, question ideas | No metrics |
| Google Keyword Planner | Free (needs a Google Ads account) | Volume estimates, seed expansion | Built for ads; data can be shown as ranges |
| Google Search Console | Free | Queries your site already appears for | Only works once you have impressions |
| Google Trends | Free | Seasonality, comparing terms | Relative 0–100 scale, no volumes |
| Ubersuggest, SE Ranking, Semrush, Ahrefs | Paid, most with a limited free tier or trial | Difficulty scores, SERP data, competitor keywords | Cost; each tool's numbers differ |
I'm not listing prices here, because they change often. Check each vendor's pricing page on the day you decide.
Which tool is best? The one you'll open every week. For most small sites I suggest Search Console plus one paid tool. Pick on database coverage for your country, how it shows the SERP, and whether it fits your budget. All-in-one platforms make sense once you also need rank tracking, audits and backlink data in one place. I compare the AI-driven options in best AI SEO tools.
Using ChatGPT, Claude and other AI tools for keyword research
Can AI do keyword research? Partly.
What AI chatbots are good at:
- Brainstorming seed keywords from a product description.
- Suggesting the questions a customer might ask at each stage of buying.
- Grouping a long list into rough themes so you can check them faster.
- Rewriting jargon into the words a beginner would use.
What they can't do reliably:
- Give you real search volumes or difficulty. Unless a chatbot is connected to a keyword data source, any number it gives is a guess.
- Tell you what Google ranks today. You still need to look at the SERP.
My workflow: use ChatGPT or Claude to widen the seed list in Step 1 and to sort a messy list in Step 6. Then check every idea against a real data source and the live results. Treat AI output as a draft, not as data.
A worked example: a small online candle shop

Illustrative example. The business is hypothetical and no tool metrics are quoted; it shows the decisions, not real data.
Say you run a small online shop selling hand-poured soy candles.
Step 1 – Seeds. soy candles, scented candles, candle gift sets, how to make candles last longer, candle care.
Step 2 – Expand. Autocomplete adds "soy candles vs paraffin", "candle tunnelling fix", "best candle gift for her", "non toxic candles". A keyword tool adds more "for" and "vs" variations.
Step 3 – Metrics. "Scented candles" has the most volume and the highest difficulty. "Candle tunnelling fix" has far less volume and a low score. "Candle gift sets" has a higher CPC, a hint of buying intent.
Step 4 – SERP check.
- "Scented candles": product grids from large retailers. That needs a category page, not a blog post.
- "Soy candles vs paraffin": blog comparisons and a forum thread. A guide fits.
- "Candle tunnelling fix": short how-to pages and videos. A short guide with a clear fix, plus a video if you can film one.
Step 5 – Difficulty. Big retailers own "scented candles". A new shop parks it as a long-term category target. The how-to and comparison terms show smaller blogs on page one, so they're winnable now.
Step 6 – Groups. "Candle tunnelling", "how to fix candle tunnelling" and "why does my candle tunnel" show the same results, so they become one page.
Step 7 – Priority. "Candle gift sets" scores highest on business value. The tunnelling guide scores highest on winnability. Both go first.
Step 8 – Map. /candle-gift-sets/ (category page), /candle-tunnelling/ (guide that links to the shop), /soy-vs-paraffin-candles/ (guide).
Step 9 – Measure. After a few weeks, Search Console shows the tunnelling guide getting impressions for "candle wick too short". That becomes a new section, not a new page.
The same steps apply to a SaaS blog, a local service business or an SEO site like this one. Only the seeds change.
Common keyword research mistakes I see
- Chasing the biggest number. High volume with no chance of ranking is wasted effort.
- Trusting KD blindly. Two tools can score the same keyword very differently. Look at the SERP.
- One keyword per page, no groups. It leads to thin, overlapping pages.
- Ignoring your own Search Console data. It's free and specific to your site.
- Researching once. Language and SERPs change. I revisit core topics at least every few months.
Frequently asked questions
Is keyword research still important with AI Overviews and AI search?
Yes. AI answers still respond to what people ask, and you still need pages that match those questions. What changes is how you judge value: some queries are now mostly answered on the results page, so weigh business value and the chance of being cited, not just clicks.
Which keyword tool is the most accurate?
None is exact. Third-party tools model their numbers from different data, so they disagree. For your own site, Search Console impressions are the closest thing to real data. For new topics, use one tool consistently and compare keywords against each other rather than trusting a single figure.
Is keyword research done before or after creating content?
Before, to decide what to write and in which format. Then again after publishing, using Search Console to find the queries a page actually attracts and to expand it.
What is the best keyword research tool for a small e-commerce website?
Start with Search Console and Keyword Planner, then add one paid tool that has good data for your country and shows product-heavy SERPs clearly. Shops benefit most from mapping keywords to category pages, so pick a tool that makes grouping easy.
Should I choose a keyword research tool or an all-in-one SEO platform for my blog?
If you only publish content, a dedicated keyword tool plus Search Console is enough. Move to an all-in-one platform when you also need rank tracking, site audits and backlink data, and you'd use them every month.
Are there keyword research tools that integrate with Google Analytics?
Google Analytics 4 doesn't collect search queries itself. Once you link Search Console, GA4's Queries report shows Search Console query data next to your analytics. Several paid SEO tools also connect to GA4 and Search Console; check each vendor's integration list.
Sources
- Google Search Central – SEO Starter Guide
- Google – Keyword Planner and Use Keyword Planner (Google Ads Help)
- Search Console Help – Performance report and About Search Console data
- Google Trends Help – FAQ about Google Trends data
- Google Analytics Help – GA4 Queries report
- Search Engine Land – Google throttling Keyword Planner data for low-spending accounts
Connect with Akif on LinkedIn, Instagram and Facebook.
Driven by advanced SEO expertise, deep marketing analytics, high-impact content strategy
With 2+ years of hands-on SEO experience, I specialize in holistic search strategies that don’t just rank—they drive real, measurable business growth. I’ve worked across industries including healthcare, hospitality, legal, e-commerce, and professional services, helping brands dominate their target markets. My approach bridges the gap between raw data and creative execution. Every strategy I build is rooted in rigorous market analysis, structured SEO frameworks, and tailored content ecosystems—no templates, no shortcuts. Whether you’re a single-location brand or scaling across multiple cities, I create data-driven marketing systems designed to compound results and grow with you.
Want a Keyword Plan Built for Your Site?
Get a keyword map that groups topics, matches search intent and assigns every keyword to the right page.
Book a Free SEO Audit →