Keyword research in 2026: sources, metrics and a step-by-step method

What keyword research is (and is not)

Keyword research is demand discovery. You start from what you sell or explain, find out how people actually phrase their needs in search, and end with a prioritised list that tells you which pages to create, which to improve and which queries to leave alone.

It is not a list of phrases to repeat in headings. Google's helpful content guidance describes the goal plainly: “People-first content means content that's created primarily for people, and not to manipulate search engine rankings.” Keywords tell you what people need and what words they use. The page still has to answer that need better than the alternatives.

This lesson covers the research itself: sources, metrics and the method. Two steps that come right after it have their own lessons. Search intent and SERP analysis explains how to read what kind of result each query expects. Clustering and the content map explains how to group keywords into topics and assign each group to one URL. Here you will tag intent and topic roughly, so the list is ready for those steps.

What Google officially says

Google does not publish a keyword research guide for SEO, but its help pages for Keyword Planner, Google Trends and Search Console explain exactly what each tool's numbers mean. These are the sentences worth remembering.

The helpful content page also lists warning signs that apply directly to how you choose keywords. Two of them:

• “Are you writing about things simply because they seem trending and not because you'd write about them otherwise for your existing audience?”
• “Did you decide to enter some niche topic area without any real expertise, but instead mainly because you thought you'd get search traffic?”

In other words: a keyword with demand is only an opportunity if it fits your audience and you can genuinely help with it.

Where keyword ideas come from

No tool sees all searches. Combine several sources, and know what each one is good and bad at.

Search Console queries

Start with your own data. The Performance report lets you, in Google's words, “Learn which queries are bringing traffic to your site, and use this to improve your SEO efforts.” These are real queries for which Google already shows your pages, with clicks, impressions and average position. Useful filters: queries with many impressions and few clicks, queries where your average position is roughly 8–20, and question queries found with a regex such as ^(how|what|why|can|does|is) .

Two limits. “The table can display a maximum of 1,000 rows, so some rare or long-tail rows might be omitted from the table, but included in the chart total.” And “To protect user privacy, the Performance report does not show all data. For example, it omits some queries that are searched a very small number of times.” Export through the API or filter by page to see more rows; the lesson Performance report: filters, regex, comparisons covers this in detail.

Google Keyword Planner

Keyword Planner is built for advertisers, but it is Google's own free source of search volume estimates. You can start from words or from a page: “Start with a website: Enter any website and Google will look for keywords related to the content on that site.” Combining both usually works best, and Google's best-practice advice is to “Describe how your customers think of your product or service category.” Start with hiking boots, not best hiking boots for wide feet; the Planner will suggest the longer variants itself.

Practical limits: “You must complete your account setup by entering your billing information to access basic features like "Get ideas for new keywords".” You do not have to run ads, but you do need a complete account. And “Keywords with very low search volumes, or keywords that are considered sensitive aren't discoverable or forecastable.” Health and finance topics are often affected: “Google’s personalized advertising and data privacy policies often prevent results for keywords in sensitive categories like health or finance.”

Google Trends

Trends shows relative interest over time and by region, not search volume. “Google Trends normalizes search data to make comparisons between terms easier.” Use it to see seasonality, to check whether a topic is growing or fading, and to compare wording (for example trainers vs sneakers) or regions. Compare terms in the same request, because each chart is scaled on its own.

Autocomplete

Type a seed into Google and note the predictions, then add letters or question words (trail shoes a…, how trail shoes…). Google explains that “The predictions come from real searches that happen on Google and show common and trending ones relevant to the characters that are entered and also related to your location and previous searches.” It also warns: “Autocomplete should not be taken as always reflecting the most popular search terms related to a topic.” Use it for wording and ideas, not for ranking keywords by size. A private browser window reduces the effect of your own history.

People also ask and related searches

The question boxes and the related searches on the results page show what else people look for around your seed. They are a good source of sub-questions for a page and of neighbouring topics. They also show you how Google interprets the query, which is the starting point of the Search intent and SERP analysis lesson.

Competitor pages

Look at the pages that rank for your seeds. Note the topics, questions and product types they cover that you do not. Keyword Planner can do part of this: enter a competitor's page as the website. Third-party SEO tools can list the keywords a domain ranks for, but those lists are modelled estimates, so treat them as an idea source, not as facts about the competitor's traffic.

Customers: support, sales, reviews, site search

The most valuable phrases are often the ones no tool shows. Support tickets, live chat logs, sales call notes, product reviews and your on-site search log contain the exact words customers use, including problems and objections that never appear in a keyword tool. Ask the support and sales teams for their ten most common questions; this alone often produces several page ideas.

Forums and communities

Reddit, Quora, niche forums and video comments show how people describe problems in their own words, and which questions get asked again and again. Many of these queries show little or no volume in tools, but the repetition is evidence of real demand.

AI assistants as idea generators

ChatGPT, Claude, Gemini and similar assistants are useful for brainstorming: seed terms, synonyms, questions, sub-topics, regional vocabulary and angles you had not thought of. Their numbers are not data. If an assistant gives you a search volume, a difficulty score or a trend, it is a guess or a recollection, not a measurement. Take the ideas, then check every metric in Keyword Planner, Trends or Search Console.

Keyword metrics and how to read them

Search volume

In Keyword Planner, average monthly searches means “The average number of times people have searched for a keyword and its close variants based on the month range as well as the location and Search Network settings you selected.” Three details follow from that definition:

• It is an average. “By default, the number of searches for the term (regardless of language) is averaged over a 12-month period.” A seasonal keyword with a big December and quiet summer gets one smoothed number.
• Close variants are grouped. running shoe and running shoes can show the same figure. Do not add them up; that counts the same searches twice.
• It is rounded. Google says the statistics are rounded, so volumes for several locations “might not add up as you'd expect.”

Accounts with little or no ad spend usually see ranges such as 1K–10K instead of a single figure. Ranges are less precise, but they are enough for what volume is really for: comparing keywords and sorting them into demand bands. Third-party tools show exact-looking numbers, but they are modelled estimates and different tools often disagree. Keep one source per column and note which source you used.

Volume is also less reliable for small areas: “You can expect results to be less accurate if you're targeting a small geographic location, simply because there is less data on which to base predictions.”

Seasonality and trends

Many keywords are seasonal: gifts, travel, tax, sports, heating. Keyword Planner shows monthly figures, and “The date range feature lets you adjust the date range in order to view search volume trends over specific time periods, and get suggestions for keywords that are more relevant during certain times of the year.” Google Trends adds the longer view: whether interest is growing year on year or slowly fading. Record the peak months in your spreadsheet and publish or update seasonal pages before the peak, not during it.

CPC and competition

Keyword Planner's bid columns show what advertisers have paid: the low range is “an approximation of the 20th percentile” and the high range “an approximation of the 80th percentile”. For SEO this is a signal of commercial value. Advertisers pay more for queries that lead to sales or leads, so a high top-of-page bid usually means a valuable query.

The Competition column is often misread. It is “The number of advertisers that showed on each keyword relative to all keywords across Google.” It says nothing about how hard it is to rank organically.

Keyword difficulty

Keyword difficulty (KD) scores come from third-party SEO tools, not from Google. Each tool calculates its score differently, usually from the links and authority of the pages currently ranking. Treat KD as a rough estimate: compare scores only within one tool, and confirm by looking at the actual results page. Who ranks there (big brands, forums, small sites like yours) and how well they answer the query tell you more than any score.

Your own position and impressions

For queries you already appear for, Search Console is ground truth. Impressions show real demand for your site in that market; average position shows how far you are from page one. A query with many impressions at position 12 is usually easier to win than a new query where you do not appear at all. Remember that “Search results are specific to the time, place, device, and recent history of the person searching”, so do not judge rankings from a single search on your own computer.

Long tail and zero-volume keywords

Long-tail keywords are longer, more specific queries: trail running shoes for flat feet rather than running shoes. Each one is searched rarely, but together they make up a large share of all searches. They usually have a clearer intent, they are easier to rank for, and the people searching them are often closer to a decision.

Many real queries show 0 or no data in tools. Keyword Planner does not show very low-volume terms, Google Trends shows them as “0”, and Search Console omits the rarest queries for privacy. Zero volume therefore means too small to report, not nobody searches this.

When is a zero-volume term worth covering? When there is other evidence: customers ask it in support or sales, it appears in autocomplete or People also ask, or it is a precise variant of a topic you already serve. Most of these terms belong as a section or FAQ on a broader page, not as a page of their own. Sum long-tail terms by topic when you judge demand: ten queries of “no data” on the same subject can add up to a worthwhile page.

Language and location

Search behaviour changes by country, language and device, so research each market separately:

• Set location and language in every tool. Keyword Planner and Trends let you choose them; the same keyword can have very different demand in the UK, the US and Australia.
• Regional vocabulary differs within one language. UK shoppers search for trainers, US shoppers for sneakers; holiday and vacation work the same way.
• Never translate a keyword list. Translate the meaning, then research again in the target language, ideally with a native speaker. A literal translation is often a phrase nobody searches.
• Local modifiers change the query. City names and near me searches signal local intent and often need a location page rather than an article.
• Devices differ. Keyword Planner can break volume down by computers, mobile devices and tablets, which helps when mobile and desktop searchers phrase things differently.

The technical side of multi-language sites is covered in the hreflang lesson. Keyword research decides what each language version should be about.

Step-by-step method

1. Define goals and seed topics. List what the business sells or explains, who it serves and what counts as a conversion. Turn this into 10–30 short seed terms in customer language, such as trail running shoes or shoe repair.
2. Harvest Search Console. Export queries for the last 12 months (or the longest period available), with clicks, impressions, position and page.
3. Expand with Keyword Planner. Enter seeds and your own pages, set the right location and language, and export the ideas with monthly searches and bid ranges.
4. Add SERP and suggestion data. Autocomplete, People also ask and related searches for your main seeds. Use an AI assistant to brainstorm missing synonyms and questions.
5. Add customer language. Support tickets, sales notes, reviews, site search, forums.
6. Check competitors. List topics and page types they have that you do not.
7. Put everything in one sheet and add metrics: volume (with source), seasonality from Trends or Planner, CPC, difficulty (with tool name), your current position and URL.
8. Clean the list. Remove duplicates and off-topic terms, merge close variants, drop keywords for products you do not offer.
9. Tag topic and rough intent. Is the searcher learning, comparing, buying or looking for a specific site? Check intent properly later in the SERP, as explained in Search intent and SERP analysis, and group keywords with Clustering and the content map.
10. Score and prioritise with value × demand × feasibility (below).
11. Map to URLs. For each topic, choose an existing page to improve or a new page to create. One topic, one main URL; two pages for the same keyword compete with each other, as the Keyword cannibalisation lesson explains.
12. Review regularly. Add new Search Console queries, re-check seasonal terms before each season and update scores as rankings change.

The research spreadsheet

A spreadsheet is enough for most sites. One row per keyword, one column per fact, and a note of where each number came from. Recommended columns:

How to prioritise: value × demand × feasibility

Volume alone is a poor way to choose keywords. A big term you cannot rank for, or one that never leads to a sale, is worth less than a smaller term that converts and is within reach. Score each keyword, or better each topic, from 1 to 3 on three factors:

• Value (V): how close the query is to your business goal. 3 = people ready to buy or enquire about something you offer; 2 = comparing or researching options you serve; 1 = related interest with no direct link to a conversion. CPC helps you judge this.
• Demand (D): volume band, trend and seasonality together. For long-tail groups, judge the topic as a whole, not each keyword separately.
• Feasibility (F): how likely you are to rank with reasonable effort. Consider your current position, the difficulty estimate, who ranks today, your site's authority on the topic, and whether you already have a page to improve.

Priority = V × D × F, from 1 to 27. Multiplying, rather than adding, means a score of 1 on any factor pulls the keyword down hard. In the example sheet, trail running shoes for flat feet scores 3 × 2 × 3 = 18 and comes first: it sells, it has demand, and the page already ranks 9th. The head term trail running shoes scores 3 × 3 × 1 = 9 because ranking for it is a long project.

Two practical rules. First, score topics after clustering when you can, because one page will target the whole group. Second, revisit feasibility as you grow: today's strategic bet becomes a quick win once you rank on page two for it.

Common mistakes

• ✕Choosing by volume aloneA big head term you cannot rank for, or that never converts, wastes months.
• ✕Treating estimates as countsPlanner volumes are rounded averages; third-party numbers are models. Use them to compare, not to forecast traffic precisely.
• ✕Adding up close variantsPlanner groups singular, plural and similar spellings; summing them double-counts.
• ✕Reading Planner Competition as SEO difficultyIt measures advertisers, not organic competition.
• ✕Skipping Search ConsoleYour best opportunities are often queries you already rank for on page two.
• ✕Trusting AI-generated metricsAssistants are for ideas; their volumes and difficulty scores are not data.
• ✕Discarding zero-volume terms automaticallyCustomer questions with “no data” are often the most convertible long tail.
• ✕Translating keyword listsResearch each language and country on its own.
• ✕Ignoring seasonalityPublishing a seasonal page at the peak is too late for it to rank in time.
• ✕One page per keywordClose variants belong on one page; separate pages compete with each other.
• ✕Chasing topics outside your expertiseGoogle lists this as a warning sign; demand alone does not make a topic yours.

FAQ

Sources

• Google Ads Help — Use Keyword Planner
• Google Ads Help — About Keyword Planner forecasts
• Google Ads Help — Best practices for finding new keywords
• Google Ads Help — Refine your new keywords in Keyword Planner
• Google Trends Help — FAQ about Google Trends data
• Search Console Help — Performance report: overview and basic setup
• Search Console Help — Performance report: troubleshooting data discrepancies
• Google Search Central — Creating helpful, reliable, people-first content

Quotes checked against the official pages on 30 October 2026.