Free keyword research when you have no tool

Keyword research is usually described as finding high-volume, low-difficulty terms, which is a description built around two numbers you cannot obtain without paying for them. Strip those out and something more useful remains: you are trying to learn what words your customers use for the thing you sell, and which questions they ask on the way to buying it. That is a question about language and intent, and most of it is answerable for free.

The reason this reframing matters is that the paid version tempts you into optimising for a number rather than for a customer. A term with an impressive volume figure attached is not necessarily a term your customers use, and the phrase your customers actually type may have a volume too low to appear in any tool while still being the phrase that brings people who buy. Working without volume data forces you to reason about intent, which is the part that determines whether a visit is worth having. What follows are five free sources, what each genuinely tells you, and where each one misleads.

Type the beginning of a query into a search box and the suggestions that appear are drawn from things people actually search for. That is a real signal about language, and it is free and immediate. It is particularly good at revealing the words customers use that you do not — the colloquial term, the local term, the misspelling that is more common than the correct spelling. Work through the alphabet after your core term, adding each letter in turn, and you will collect a substantial list of genuine phrasings in twenty minutes.

The related searches at the foot of a results page do a similar job at a different level, showing adjacent concepts rather than completions. Two limits are important. Suggestions are personalised and localised, so what you see is shaped by your own location and history, and it is worth checking in a private window to reduce that. And ordering is not volume: a suggestion appearing first does not mean it is searched more than the one below it, and treating the list as a ranked volume table is the most common misreading. Autocomplete tells you a phrase is used by real people. It does not tell you how often, and it never will.

The People Also Ask box lists questions related to your query and expands with more as you open them, which makes it the richest free source of question phrasing available. Its value is not the questions themselves so much as what they expose about where customers actually are in their thinking. A term you assumed was a purchase query turns out to generate questions about whether the procedure hurts, how long it takes, and whether insurance covers it — which tells you the person searching is nowhere near booking and that a page consisting of a price and a button will not serve them.

Used properly this shapes what a page contains rather than just which phrase it targets. The questions cluster, and the clusters correspond to stages: what is this, is it right for me, what does it cost, who should I use. A single page trying to serve all four usually serves none. Expanding the box repeatedly generates a lot of material, so treat it as a source to mine and then edit down to the questions your customers actually ask, rather than a checklist to answer exhaustively — the tail of an expanded PAA tree drifts steadily further from your business.

The strongest free source is the one that only you can access. For a verified property, Search Console's performance report lists the queries that produced impressions and clicks for your pages — actual queries typed by actual people who saw your site, with counts. This is not an estimate from anybody's model. It is your own data, and it is free, gated only on the owner verifying the property and connecting access.

What makes it uniquely valuable is a specific pattern: queries where you already receive impressions but few clicks, which means you appear and are not chosen. Those name pages where the search exists, your relevance is established, and something about how you present or answer it is failing — the most actionable finding available anywhere, and it costs nothing. The limits are real too. It only covers queries where you already appeared, so it cannot reveal demand you have never touched. It samples and omits low-frequency queries. And the position figure is an average across impressions, which makes it a weak basis for any precise claim. Read it for the language people used and the gap between impressions and clicks, not as a ranking dashboard.

The most underused source is not a search tool. It is the record of what customers have already asked you: support emails, WhatsApp threads, the questions the person on the front desk answers eight times a day, the objection that comes up in every quotation call. These are your customers' own words about your own product, which is precisely what keyword research is trying to approximate at a distance.

The advantage over any search-derived source is that these come with context a query string cannot carry: who asked, what they were worried about, what they went on to buy. Read a month of enquiries and write down the recurring phrases verbatim, resisting the urge to translate them into industry vocabulary — the translation is exactly the error, because customers search in their words and businesses write in theirs. The gap between the two is usually the most valuable thing this exercise finds. The limit is obvious and worth stating: these are people who already found you, which makes them unrepresentative of the wider market. Use the inbox for language and for questions worth answering, and do not mistake it for a picture of demand.

None of the five sources gives you a volume figure, and no free source will, because search engines do not publish volume to anyone who asks. Every volume number quoted anywhere is a company's estimate produced from its own data and sold accordingly. So the practical question is how to prioritise without it, and the answer is to substitute evidence you do have. Commercial value: does this phrase describe something you actually sell at a margin you want. Specificity: does the phrasing indicate someone close to acting rather than beginning to read. Corroboration: did this phrase turn up in more than one source — autocomplete and the inbox and PAA — which is weak evidence of real use and considerably better than nothing. Existing traction: does Search Console already show impressions for it, meaning the demand is demonstrated for your site specifically.

That ordering is not as precise as a volume column and it is honest about what it is. What it should not become is a set of numbers you invented to fill the gap. Assigning your own estimated volumes to a spreadsheet produces a document that looks like data and contains none, and a decision made from it is a guess wearing a table's clothing. If a figure matters enough to a decision, that is the point at which buying a vendor's estimate becomes a reasonable purchase — with the understanding that you are buying an estimate, from a named source, not a fact from Google.