How to do keyword research. Step by step, the way we actually run it.

What keyword research is, and why this order matters

Keyword research is the work of finding the searches your customers actually run, then deciding which ones are worth building pages around. The output is a keyword map: one row per phrase, columns for volume, intent, difficulty, target URL, and current rank. The keyword map is the upstream document every other SEO decision flows out of.

The order of the six steps below is not arbitrary. Most agencies run keyword research in the wrong order, starting with volume data and only checking intent at the end. That is how you end up with content that ranks for nothing. The right order is: relevance first, intent second, difficulty third, volume last. Start by working out what the business actually sells and what its customers actually search. Volume comes in at the end as a filter, not a starting point.

How keyword research works in practice

At a mechanical level, keyword research is a series of additions and subtractions on a spreadsheet. You add candidate phrases from multiple sources, then subtract the ones that fail each filter, until you are left with the shortlist you intend to write pages for.

Each filter throws out roughly half the candidates. A typical Perth small-business research session looks like this:

• Seed list: 30 to 60 root phrases.
• After tool-based expansion: 400 to 1,200 candidate keywords.
• After relevance filter: 200 to 500.
• After intent scoring: 120 to 250.
• After difficulty filter: 80 to 150.
• After clustering and URL mapping: 15 to 30 target URLs.

If your funnel does not narrow at every stage, you are skipping a filter. The whole point is that 95 percent of the phrases you start with should not survive to the final plan.

Step 1. Define the jobs the business sells

Before opening a keyword tool, write down what the business actually sells. Not what you wish it sold. Not what is on the menu page. What customers genuinely pay for, in the language they themselves use.

The cleanest way to do this is to read the last 20 sales conversations or enquiry emails. Pull out the verbatim phrasing. A customer might say "We need someone to look after our Google rankings". You might call that "SEO retainer service". The customer's phrasing is what they will type into Google later. Yours is not.

For a Perth plumber, the job list might look like:

• Burst pipes, emergency call-outs
• Hot water system repairs and replacements
• Blocked drains
• Kitchen and bathroom plumbing for renovations
• Gas plumbing and certifications
• Commercial plumbing for strata and property managers

For a Perth lawyer, the job list might be totally different: separation, divorce, parenting orders, property settlement, consent orders, mediation. The two businesses live in two different keyword universes. Most agency keyword research goes wrong here because the agency never sat with the founder long enough to learn what the business actually does.

Step 2. Build the seed list

The seed list is your initial set of root phrases. Aim for 30 to 60 seeds across all the jobs you listed. Three sources to pull from:

Source A: Your own brain

For each job, write down five to ten ways a customer might phrase it. "Burst pipe" might generate seeds like "burst pipe", "emergency plumber", "pipe leaking water", "water flooding kitchen", "after hours plumber". Do not try to be exhaustive at this stage; the keyword tool will do that for you in step 3.

Source B: Google Search Console

If the site has any history, GSC is the most valuable source on this entire list. Open the Performance report, set the date range to the last 12 months, and export the top 1,000 queries. You will find dozens of phrases you never would have guessed, because they are what real Australian searchers are typing. Many will be queries the site shows up for at position 20 to 40, currently leaking impressions you can convert into clicks. We cover the broader GSC workflow in the technical SEO audit guide.

Source C: Competitor SERPs

Search your top three competitors' brand names plus "service" or your category. Look at the autocomplete suggestions. Open their top-ranking pages. The H1, H2 and FAQ headings on competitor pages are seed candidates: they have already done some of the research for you. The competitor keyword gap analysis chapter goes deeper on this when you are ready.

Step 3. Expand the seeds with a keyword tool

Now feed every seed into a keyword research tool to generate related terms, questions, long-tail variants, and Australian search volumes. You do not need a paid tool for this on the first pass.

Our recommended free stack:

1. Google Keyword Planner. Set location to Australia. For each seed, pull the related keyword list and the monthly search volumes. Planner under-reports low-volume terms by lumping them into ranges (10-100, 100-1k), but it is accurate for relative scale.
2. Google autocomplete and People Also Ask. Type your seed into Google.com.au and write down every autocomplete suggestion. Scroll to the People Also Ask box and click each question to expand it; new questions appear as you click. This is free, fast, and tells you exactly what Australians are searching this week.
3. Ahrefs Free or SEMrush Free. Limited free searches give you difficulty scores and the top-ranking URLs for each phrase. Enough to make difficulty decisions in step 5.

The output of step 3 is a spreadsheet with 400 to 1,200 candidate keywords, with columns for the seed they came from, monthly search volume, and (where available) a difficulty score. Do not filter yet. The job here is breadth.

For deeper reviews of each tool we have actually used at agency scale, see free keyword research tools, reviewed.

Step 4. Read the SERP and score search intent

This is the step beginners skip and pros never skip. For every shortlisted keyword, you open Google.com.au, type the phrase, and read what is actually ranking. The SERP tells you what kind of page Google currently thinks satisfies the searcher. If the SERP is full of how-to articles, an informational query. If it is full of service pages, a commercial query. If it is full of product pages with Add to Cart buttons, a transactional query. If it is full of brand homepages, a navigational query.

The four search intent categories we tag:

• Informational. User wants to learn. "What is SEO", "how long does SEO take". Ranking pages are guides, articles, definitions.
• Commercial investigation. User is comparing options before buying. "Best SEO agency Perth", "Ahrefs vs SEMrush". Ranking pages are comparisons, reviews, listicles.
• Transactional. User is ready to buy or convert. "Buy running shoes online Australia", "SEO services Perth pricing". Ranking pages are product, service or pricing pages.
• Navigational. User wants a specific brand or page. "Bunnings", "Google Search Console login". You almost never write content for these unless it is your brand.

Tag every keyword on the list with one of the four. When in doubt, the SERP wins. If your gut says "informational" but the top ten are all service pages, the keyword is commercial and Google has decided. Argue with Google and you will rank for nothing.

For the full intent treatment, including the edge cases and the SERP feature shifts that have moved through 2025 and 2026, see understanding search intent. For how intent feeds into the actual page you write, the content depth chapter in the On-Page pillar shows the matching content patterns.

Step 5. Filter by difficulty and revenue fit

Now apply two filters in sequence.

The difficulty filter

For each keyword, check the difficulty score from your tool and pair it with a glance at the live top ten. Difficulty scores from Ahrefs, SEMrush and Moz are useful but not absolute. They measure different things and disagree often. Use them as a relative ranking inside one tool, not a universal score. The full breakdown is at keyword difficulty explained.

Our rough cut-offs by site age and current ranking footprint:

• Site under 12 months old, fewer than 50 ranking keywords: cut anything above KD 30.
• Site 1 to 3 years old, decent topical footprint: cut anything above KD 50.
• Established site, hundreds of ranking keywords, strong backlink profile: anything is fair game, but expect 6 to 12 months for KD 70 plus.

These are guidelines, not laws. A small Perth site can beat a KD 65 keyword if the SERP is full of weak, off-topic results. Always sanity-check by reading the actual ranking pages, not just the score.

The revenue-fit filter

For each surviving keyword, ask the brutal question: if this keyword ranked tomorrow, would the traffic it brings in actually convert?

A 5,000-search-per-month informational keyword that brings students or curious tourists to a B2B legal site is worse than a 10-search-per-month commercial keyword that brings business owners ready to engage. Cut keywords where the traffic would be the wrong audience, even if the volume looks tempting.

The revenue-fit filter is usually where the list halves again. After step 5, you should be left with 80 to 150 keywords for a typical small business.

Step 6. Cluster and map to URLs

The last step turns your filtered list of keywords into an actual content plan.

Cluster keywords that share the same underlying intent and could be answered by the same page. "How long does SEO take", "SEO timeline for small business" and "how soon will I see SEO results" are three keywords but one cluster, because one page can answer all three. The keyword clustering chapter shows the manual and automated approaches in detail.

For each cluster:

1. Pick the primary keyword: the one with the highest volume and the cleanest intent match. This becomes the focus of the page title and H1.
2. List the variants as supporting keywords. These get woven into H2s, body copy and FAQ questions.
3. Map the cluster to exactly one URL. Either an existing URL that needs a rewrite or a new URL to be created.
4. Tag the URL with intent (so writers know the content pattern), priority (which to write first), and any existing rank position from Search Console.

The output of step 6 is a content plan: typically 15 to 30 URLs for a small business, each with a clear primary keyword, intent, supporting variants and current status. That is what you write the next quarter's content briefs from. The internal linking strategy chapter then explains how to weave those URLs into a hub-and-spoke that builds topical authority.

Common mistakes

Tools and checklist

The free stack

1. Google Search Console. Your own query data. Free, indispensable.
2. Google Keyword Planner. AU search volumes inside Google Ads. Free with an Ads account.
3. Google autocomplete + People Also Ask. Instant question discovery. Free.
4. Ahrefs Free or SEMrush Free. Difficulty scores and SERP overviews. Free with limits.
5. The live SERP. Free. Ignored by 90 percent of agencies.
6. A spreadsheet. Sheets or Excel. One row per keyword.
7. Our free SEO audit tool. Pulls your top queries and ranking pages alongside the technical signals.

The 10-point quality check

1. Did the seed list come from at least three sources?
2. Is every shortlisted keyword intent-tagged from the live SERP?
3. Have head-term keywords above the realistic 12-month ceiling been demoted?
4. Is each cluster mapped to exactly one URL?
5. Has the primary keyword per URL been picked?
6. Are local modifiers applied where the business serves a defined area?
7. Has the SERP feature mix (AI Overview, local pack, video) been noted for priority keywords?
8. Have existing pages been audited as rewrite candidates first?
9. Has a quarterly refresh date been set?
10. Is the keyword map tied to conversion measurement so you know which pages earn money?

Perth and WA context

Three real-world Perth patterns we see and the fixes we apply:

Trades businesses targeting "Perth" only. A roofer or electrician targets "roofing Perth" and "electrician Perth" and nothing else. Volume is reasonable, difficulty is brutal, and the SERP is dominated by directory aggregators like Oneflare and HiPages. The fix is to add suburb-level keywords: "roofing Fremantle", "electrician Joondalup", "roofing Mandurah". Difficulty drops by half. The directories rarely have dedicated suburb pages. See SEO Fremantle, SEO Joondalup, SEO Mandurah and SEO Cockburn for how we structure the suburb pattern, and the trades SEO guide for the broader category.

Mining and resources B2B. WA mining services firms often inherit a generic global keyword list from a metro agency. The list misses the regional terminology entirely. Words like "FIFO", "shutdown", "ROM", "Karratha" and "Bunbury logistics hub" are commercial gold but only show up if you pull seeds from the actual industry. See SEO Karratha and SEO Kalgoorlie for the regional pattern, and the mining SEO guide for the category.

Healthcare and legal firms with regulated language. Medical clinics and law firms have to be careful with terminology, and that flows into keyword choice. "Best knee surgeon Perth" is an AHPRA risk for a clinic. "Knee replacement surgery Perth" or "knee surgeon Subiaco" is safe and still commercial. The keyword research step has to filter for the regulatory boundary. See healthcare SEO and legal SEO for the compliance-aware patterns.

For the wider Perth context, the why SEO matters for Australian businesses chapter covers the local pack mechanics that sit underneath all of this, and Local SEO Perth is where most of this work lands as service delivery.