Keyword Research in 2026: A Complete Guide for Businesses That Want Leads, Not Traffic

Most keyword research produces a spreadsheet nobody uses. Someone exports two thousand phrases from a tool, sorts by search volume, highlights the biggest numbers in yellow, and hands it over. Six months later the business has ranked for a handful of high-volume terms that bring visitors who were never going to buy anything, and the phrases that would have brought actual customers are still untouched because they only get ninety searches a month.

Keyword research done properly is not a volume exercise. It is a demand-mapping exercise: working out what your future customers type at each stage of deciding to spend money, which of those queries you can realistically compete for given your current authority, and what kind of page each one requires. This guide covers the full process, the free and paid tools, and the judgement calls that separate a useful keyword plan from a decorative one.

Why Search Volume Is the Wrong Starting Point

Consider two phrases for a dental clinic in Toronto. “Dental care” gets tens of thousands of searches a month globally. “Emergency dentist near me open now” gets a fraction of that. The first is a research term typed by students, patients in other countries, and people idly reading; the second is typed by somebody in pain with a credit card who will book within the hour.

The second keyword is worth more than the first by an enormous margin, and it is also far easier to rank for. This pattern holds in every industry. High-volume head terms are dominated by publishers, aggregators and national brands; the money sits in the specific, unglamorous, lower-volume phrases underneath them.

So the first mental shift is this: you are not looking for the biggest numbers. You are looking for the highest expected value, which is roughly volume multiplied by intent multiplied by your realistic chance of ranking. A term with 90 searches a month, buying intent, and a plausible path to page one is worth more than a term with 40,000 searches you will never rank for.

The Four Types of Search Intent

Every query carries an intent, and Google now matches intent before it matches keywords. Producing the wrong page type for a query is the most common reason well-written content fails to rank at all.

• Informational — the searcher wants to learn something. “What is technical SEO”, “how much does a website cost”, “why do my gums bleed”. Requires a guide, explainer or tutorial. Long consideration cycle, but this is where you build trust and topical authority.
• Navigational — the searcher wants a specific site or brand. “KhatooTech pricing”, “Shopify login”. You can only win these for your own brand.
• Commercial investigation — the searcher is comparing options before buying. “Best dental marketing agency”, “Shopify vs WooCommerce”, “dental marketing company vs agency”. Requires comparisons, reviews, buyer guides. Extremely high value and consistently under-served.
• Transactional — the searcher is ready to act. “Dentist near me”, “hire SEO consultant”, “emergency plumber Brampton”. Requires a service or landing page, not a blog post.

The practical test costs thirty seconds: search the term and look at what Google is already ranking. If the first page is entirely blog posts, Google has decided this query wants information, and your service page will not rank no matter how well optimised it is. If the first page is service pages and local listings, a blog post will not rank either. Match the format that is already winning.

Step 1: Build Your Seed List Before You Open Any Tool

Start with what you already know. Tools are good at expanding a list and terrible at generating one from nothing, because they do not know your business.

• Your services, written the way customers say them. You might call it “conversion rate optimisation”; your customer types “why is nobody buying from my website”.
• Questions you answer on sales calls. Every recurring question is a keyword. Ask whoever handles enquiries to write down the ten questions they answer most often.
• Objections and comparisons. “Is it worth it”, “how much”, “how long does it take”, “X vs Y”, “alternatives to X”.
• Problems, not solutions. Customers search their symptom before they know the name of the cure. Someone needing local SEO searches “my business does not show up on Google Maps”.
• Locations you actually serve, at the granularity people use — neighbourhood and suburb names, not just the city.
• Your competitors’ names, which will feed comparison and alternative-to content later.

Twenty to forty seeds is plenty. This list is the input to everything that follows.

Step 2: Expand With Tools

Free tools that are genuinely enough to start

• Google Search Console. The single best keyword tool you own, and the only one with real data about your site. The Performance report shows every query you already receive impressions for — including hundreds you never targeted. Sort by impressions with low clicks and you have a list of things Google already thinks you are relevant for.
• Google autocomplete. Type your seed and read the suggestions. Add a letter of the alphabet after it to force more. These are real queries, ordered roughly by popularity.
• People Also Ask. The expanding question boxes in results are a live map of related intent. Each expansion loads more. They are also excellent raw material for the FAQ sections that AI engines quote.
• Related searches at the bottom of the results page.
• Google Keyword Planner. Free with a Google Ads account. Volumes are bucketed rather than precise unless you are spending, but the discovery function is solid and the commercial-intent signal (suggested bid) is genuinely useful — a high suggested bid means advertisers are making money from that term.
• Google Trends. Direction and seasonality rather than absolute volume. Essential for deciding whether a topic is growing or dying.
• Bing Webmaster Tools. Free keyword research with its own data set, and Bing’s index feeds several AI assistants.
• Reddit, Quora and industry forums. Unfiltered phrasing from real people, which is exactly the phrasing that matters for conversational and AI search.

Paid tools and when they earn their fee

Ahrefs, Semrush and Moz cost real money and are worth it once you are publishing regularly or managing client work. What you are actually paying for is not keyword volume — it is competitor visibility. Being able to see every keyword a competitor ranks for, and which of their pages earn links, collapses weeks of guessing into an afternoon.

If budget is tight, subscribe to one for a single month, export everything you need, and cancel. A well-run month of research produces a plan you can execute for a year.

Step 3: Judge Difficulty Honestly

Every tool reports a keyword difficulty score. Treat it as a rough prior, not a verdict, because these scores are computed almost entirely from the backlink profiles of the current top ten and know nothing about your site.

A more honest assessment takes two minutes per keyword:

1. Search the term and read the first page. Who is actually there? If it is Wikipedia, Forbes, government sites and three national brands, you are not ranking there this year. If it is small businesses, thin blog posts, forum threads and a couple of directories, there is room.
2. Check whether the results genuinely answer the query. Weak, off-target results on page one are the clearest opportunity signal that exists.
3. Compare domain strength. If every result comes from a site with hundreds of referring domains and you have twelve, adjust expectations regardless of what the difficulty score says.
4. Count the SERP features. If ads, a map pack, a video carousel and an AI Overview occupy the screen, organic position three may receive almost no clicks. Volume without visible real estate is not opportunity.

The uncomfortable arithmetic that most keyword guides skip: on a new or low-authority domain, almost nothing competitive is winnable in the first year, however good the content. If your pages consistently sit around position 60-80 for terms you have covered thoroughly, the constraint is not keyword selection — it is domain authority, and the fix is external links and citations rather than another article. Choose keywords accordingly: go long-tail and specific until the authority exists to compete higher.

Step 4: Group Keywords Into Topics, Not Pages

A single page should target one intent, not one keyword. Google understands synonyms and variations well; “seo audit checklist”, “checklist for seo audit”, “how to audit a website for seo” and “seo audit steps” are the same intent and belong on one page. Splitting them across four thin pages guarantees all four compete with each other and none ranks.

Group your expanded list into clusters where every phrase would be satisfied by the same page. Then decide, per cluster:

• Page type — service page, pillar guide, supporting article, comparison, location page, FAQ.
• Primary keyword — the phrase the page is built around, used in the title, H1 and URL.
• Secondary keywords — variations used naturally in H2s and body copy.
• Funnel stage — awareness, consideration or decision, which determines the call to action.

Then arrange the clusters into hub-and-spoke structures: one pillar page per major topic, several supporting articles linking up to it and across to each other. This is the architecture that builds topical authority, and it is why a site with thirty well-linked pages on one subject routinely outranks a site with three hundred scattered ones.

Step 5: Prioritise Ruthlessly

You cannot execute two hundred keywords. Score each cluster on four factors and work top-down:

• Commercial value — how close is this to someone paying you? A “pricing” or “near me” query outranks a “what is” query every time on this axis.
• Winnability — can you realistically reach page one within six to twelve months given your current authority?
• Volume — the least important of the four, and the one everyone over-weights.
• Strategic fit — does this reinforce a cluster you are already building, or start an unrelated one? Depth beats breadth.

A useful sequencing rule: publish the transactional and commercial-investigation pages first even though they are harder, because they are the ones that generate revenue when they land. Fill in informational content around them to build the topical support the commercial pages need. Doing it the other way round — a year of blog posts, then service pages — means a year without leads.

Step 6: Keyword Research for Local Businesses

Local search follows different rules and is far more winnable for most small businesses.

• “Near me” queries are location-implicit. You do not optimise for the literal words “near me”; you optimise your Google Business Profile, your proximity signals and your local citations. Google supplies the location.
• City and neighbourhood modifiers genuinely matter for pages: “dental marketing agency Vancouver”, “web design Brampton”. These are far less competitive than the national head term and convert better.
• The map pack is a separate ranking system. Proximity, prominence and relevance drive it, and no amount of on-page keyword work substitutes for a complete, verified, actively reviewed Google Business Profile.
• Do not build hundreds of near-identical location pages. Google treats spun city pages as doorway pages. Build them only for places you genuinely serve, with genuinely distinct content — local examples, local pricing, local context.

Our local SEO checklist covers the map-pack side of this in detail.

Step 7: Keyword Research for eCommerce

Product search splits cleanly by page type:

• Category pages target broad, high-volume terms: “running shoes for men”, “cotton kurta sets”.
• Product pages target specific model and attribute terms, plus long-tail buying phrases.
• Blog content targets the informational layer that feeds both: “how to choose running shoes”, “cotton vs linen in summer”.

Modifiers do the heavy lifting: “best”, “cheap”, “under 5000”, “for beginners”, “reviews”, “size guide”, “vs”. Each signals a different stage and deserves its own page. The full treatment is in our eCommerce SEO guide.

Step 8: Keyword Research for AI Search

A growing share of searches now end in a generated answer rather than a click. That changes what you research for, in three concrete ways.

Queries get longer and more conversational. People type full sentences to an assistant: “which dental marketing agency is best for a small clinic in Canada” rather than “dental marketing agency”. Your keyword list should include natural-language questions, not just noun phrases.

Being quotable matters more than being ranked. Models extract self-contained passages. A section that opens with a direct, complete answer gets cited; one that builds to its point over four paragraphs does not. Structure content so each heading is a question and each first sentence is the answer.

Comparison and specificity win. Assistants are constantly asked to compare and recommend. Content with real numbers, named alternatives, explicit trade-offs and honest limitations is what gets pulled into answers. Vague marketing copy is never quoted by anything.

Practically, add a research pass where you ask ChatGPT, Perplexity and Gemini the questions your customers ask, and note which sources they cite and what shape those pages take. That is the new SERP analysis. We go deeper in our AEO and GEO guide and the technical SEO guide.

A Worked Example: Keyword Research for a Dental Clinic

Abstract advice is easy to nod along to and hard to apply, so here is the same process run end to end for a single-location dental clinic in Brampton, Ontario.

Seeds come from the services on the price list and the questions the front desk answers daily: dental implants, Invisalign, teeth whitening, root canal, emergency appointments, kids’ dentistry, cost, insurance, “does it hurt”, “how long does it take”.

Expansion through Search Console, autocomplete and People Also Ask produces a few hundred phrases. Grouped by intent they fall into recognisable buckets:

• Transactional, local: “dentist Brampton”, “emergency dentist Brampton”, “dental implants Brampton”, “Invisalign near me”. These belong on service pages and the Google Business Profile. Highest value, hardest to win without local authority.
• Commercial investigation: “dental implants cost Ontario”, “Invisalign vs braces”, “best dentist in Brampton”, “is teeth whitening worth it”. These belong on cost pages, comparison pages and honest buyer guides. Under-served almost everywhere, and they convert.
• Informational: “why do my gums bleed”, “how long does a root canal take”, “what to do if a crown falls out”. These belong in the blog. They rarely convert on the first visit but they are how a clinic becomes the name people already recognise when they finally need treatment.

Prioritisation then flips the obvious order. The clinic publishes the cost and comparison pages first, because those searchers are weeks from booking. The local service pages get built alongside, but everyone accepts they will take six to twelve months and a serious Google Business Profile effort to land. The informational blog fills in behind, each post linking up to the relevant service page.

What does not happen: nobody writes a 3,000-word post targeting “dental care”, and nobody builds forty near-identical pages for every suburb within fifty kilometres. Both are volume plays that produce nothing.

Mapping Keywords to the Buyer Journey

Every keyword sits somewhere on the path from “unaware of the problem” to “ready to pay”. Mapping them explicitly stops you writing decision-stage copy for awareness-stage visitors, which is the most common conversion leak on business websites.

A useful sanity check: if a page ranks well and gets traffic but generates no enquiries, the usual cause is not weak copy — it is a decision-stage call to action bolted onto an awareness-stage page. Match the ask to the stage.

Branded vs Non-Branded Keywords

Branded searches — your company name, or your name plus a service — convert extraordinarily well, because the searcher already knows who you are. They are also almost entirely a consequence of other marketing rather than of SEO: ads, social, word of mouth, offline presence.

The distinction matters when you are judging whether SEO is working. A site whose traffic is ninety percent branded is not being found by new customers; it is being looked up by people who already heard the name elsewhere. Segment the two in Search Console (filter queries containing your brand, then invert the filter) and track non-branded impressions and clicks as the real measure of search growth.

The healthy pattern over a year is non-branded impressions rising steadily, non-branded clicks following a few months later, and branded searches rising slowly as overall awareness grows.

Seasonality and Trend Timing

Some demand is flat all year; much of it is not. Tax services, gifting, admissions, travel, fitness and most retail categories swing enormously by month, and content published the week demand peaks is already too late — Google needs time to crawl, evaluate and build confidence in a new page.

Check Google Trends for every significant cluster and work backwards: publish and internally link seasonal content roughly two to three months before the curve rises, then refresh it annually rather than writing a new page each year. An updated page that already has history and links will outrank a fresh one every time.

Trends also tells you when a topic is structurally declining. If a term you were planning a pillar around has been falling for three years, that is a good reason to spend the effort somewhere else.

Common Keyword Research Mistakes

• Chasing volume you cannot rank for. The single most common and most expensive error.
• Ignoring intent. Publishing a service page for an informational query, or vice versa. It will not rank, and no amount of optimisation fixes a format mismatch.
• One page per keyword variation. Creates cannibalisation and thin content simultaneously.
• Trusting difficulty scores blindly. They do not know your site, your niche or your links.
• Never revisiting the list. Search behaviour shifts; a plan built two years ago is describing a market that has moved.
• Optimising for keywords rather than people. Copy written to hit a phrase count reads badly, converts badly, and has not ranked better for over a decade.
• Skipping your own Search Console data. Free, specific to you, and consistently more actionable than any third-party estimate.

Competitor Keyword Analysis

The fastest way to find keywords worth targeting is to look at what is already working for someone else. Competitor analysis is not copying — it is using their years of trial and error as free research.

Start by identifying the right competitors, which are frequently not the businesses you think of as rivals. Your search competitors are whoever occupies page one for your target terms: that might be a directory, a national aggregator, a publisher, or a small operator in another city who simply writes better content. Search five of your priority keywords and list who appears repeatedly.

Then, for each, look for three specific things:

• Keywords they rank for that you do not. This is the content-gap analysis every major SEO tool offers, and it is the single most useful paid feature available. Filter for terms with commercial intent and realistic difficulty.
• Which of their pages earn the most traffic. Usually a small handful carries the site. Understanding what those pages do — format, depth, angle — tells you what the market rewards.
• Which of their pages earn links. Different from their traffic pages, and more instructive. Whatever attracted links for them will usually attract links for a better version.

What to do with the output: do not simply write your own version of their top page. Write the version that fixes what theirs gets wrong — the outdated figures, the missing local context, the question it raises and never answers, the pricing it refuses to discuss. Being genuinely more useful on a specific dimension beats being a slightly longer duplicate.

One caution: if a competitor ranks for a term purely because they have ten times your authority, adding it to your plan is setting money on fire. Filter their keyword list through your own winnability assessment before committing to anything.

Turning the List Into a Content Plan

Research is worthless until it becomes a schedule. Convert your prioritised clusters into a simple table with, for each row: the target URL, the primary keyword, the intent, the page type, the internal links it should receive and give, and a publish date. Assign roughly seventy percent of your capacity to clusters you are already building and thirty percent to new ground.

Then measure the right thing. Rankings are a vanity metric on their own; what matters is impressions growth for the target cluster, then click growth, then enquiries. Impressions rising while clicks stay flat is normal and healthy for the first few months — it means Google is testing you at low positions. Impressions flat after three months of publishing means the cluster is not being taken seriously, and the answer is usually consolidation or authority, not more posts.

Review quarterly. Pull the Search Console query list, find phrases where you rank between eleven and thirty with meaningful impressions, and improve those pages first. Moving a page from position fourteen to position six is almost always cheaper and more profitable than starting a new one from zero.

Frequently Asked Questions

What is keyword research?

Keyword research is the process of finding the phrases your potential customers type into search engines, judging how competitive each one is, understanding what kind of page the searcher expects, and prioritising which ones your site should target. Done properly it is a demand-mapping exercise, not a list of high-volume phrases.

What are the best free keyword research tools?

Google Search Console is the most valuable because it shows the queries your own site already receives impressions for. Beyond that: Google Keyword Planner for discovery and commercial intent, Google autocomplete and People Also Ask for real phrasing, Google Trends for direction and seasonality, and Bing Webmaster Tools for a second data set. These are genuinely enough to build a first content plan.

How many keywords should one page target?

One intent, not one keyword. A single page should target a primary keyword plus every close variation that would be satisfied by the same content – often ten to thirty phrases. Splitting near-identical phrases across separate pages causes cannibalisation, where your own pages compete and none of them ranks well.

What is a good keyword difficulty score to target?

It depends entirely on your domain’s authority, which is why the score alone is misleading. A new site with few backlinks should focus on long-tail terms with difficulty under roughly 20 and clear buying intent. An established site can attack 40 and above. The more reliable test is to search the term and look at whether the current page-one results are sites comparable to yours.

Should I target high-volume or long-tail keywords?

Long-tail, in almost every case, especially early on. Specific phrases have clearer intent, far less competition and much higher conversion rates. Ten long-tail terms bringing 50 visitors each will usually generate more enquiries than one head term bringing 500, because the visitors are further along in deciding to buy.

How do I do keyword research for a local business?

Combine service terms with the locations you genuinely serve, at the granularity locals use – neighbourhood and suburb names, not just the city. Remember that ‘near me’ searches are location-implicit: you win them through a complete, verified, well-reviewed Google Business Profile and local citations rather than by putting the words on a page. Build location pages only for real service areas, with genuinely distinct content.

How is keyword research different for AI search?

Queries become longer and more conversational, so your list should include full natural-language questions rather than only noun phrases. Being quotable matters as much as ranking, so each section should open with a direct, self-contained answer under a question-shaped heading. Comparison content and specific figures get cited by AI engines far more often than general marketing copy.

How often should I redo keyword research?

A full refresh once or twice a year, plus a quarterly review of your Search Console data. The quarterly pass is the high-value one: find queries where you rank between eleven and thirty with real impressions, and improve those existing pages. That is consistently cheaper and faster than starting new pages from zero.