Skip to content
search-engine-optimizedby Netsolution

Keyword research

Looking for keywords means looking for real demands, not flattering volumes. Tool estimates rest on methods those tools do not publish; Search Console, by contrast, reports the queries that actually showed your pages. So you start from what exists before inventing what might exist.

What exactly are you looking for when you “look for keywords”?

You are looking for demands. Not words, not phrases to sprinkle through a text, but moments when somebody had a problem and typed that problem into a search box. The difference sounds rhetorical; it decides everything that follows, because a list of words gets handled by scattering them through a page, whereas a list of demands forces you to decide which one you can handle better than anyone else.

That confusion has a historical trace, and it is documented:

The meta keywords tag is not used by Google Search.

officialSource : Google Search Central, “SEO Starter Guide”

The tag in question was designed precisely to declare your keywords to the engine. It is no longer read. What survives from that era is the reflex of thinking in words to declare rather than questions to handle.

Why are the volumes shown by tools not measurements?

Every keyword tool displays a number of monthly searches. That number looks like a measurement: it is precise, it is stable from month to month, it can be compared.

The search volumes displayed by third-party tools are estimates built by methods none of them describes publicly.

hypothesis

The “hypothesis” status is not decorative caution. None of these tools publishes its estimation method, and it takes opening two competing suites on the same phrase to get two different values. They are useful hints for comparing two demands against each other, and worthless figures the moment you take them for readings. This site will publish none of them: a volume displayed with neither method nor date is exactly the kind of data that ages badly.

Two views of the same lexicographical archive room: from floor to ceiling, hand-labelled boxes holding the vocabulary slips of a regional dictionary.
A vocabulary built slip by slip, from what people actually said. That is the material being looked for here: recorded phrasings, not lists produced by a tool that estimates.J. Lunau — CC BY-SA 2.5 · Wikimedia Commons, relevé le

Where do you find real queries rather than estimates?

There is one source that does not estimate, because it reports what happened.

The Search Console performance report provides four metrics per query: clicks, impressions, click-through rate and average position.

officialSource : Google Search Console, “Performance report (Search results)”

Those four metrics are enough to work with for months. Impressions say a page was offered for a query; clicks say whether the offer was taken; their ratio says whether the title and description keep their promise; average position places the context. On a site that is already live, that list of real queries is worth more than any market exploration, for one simple reason: the engine has already decided your pages had something to do with those demands.

Starting from the queries Search Console already recorded

hypothesis

Effort
low
Time to effect
weeks
Risk
none
Durability
lasting

An important reservation comes with that source, and it is written in its own documentation:

A query present in your data may not show your site if you run it yourself: the time, the place, the device and the person's history change the result.

officialSource : Google Search Console, “Performance report (Search results)”

In other words, checking a query from the list “by hand” proves nothing, in either direction. What you see from your own browser is one results page among many. It is a reflex to lose early, because it produces false conclusions with a great deal of conviction attached.

How do you decide that a query deserves a page?

Three questions, in this order, and the first one eliminates the most candidates.

Can you answer better than what already exists? Not “as well”, not “with more words”. Better, on an identifiable point: experience you have and others do not, data you can verify, a concrete case nobody tells. If the answer is no, the page will be one more duplicate, and producing it will cost time without returning any.

Is the demand a demand, or a rephrasing? Two phrasings expressing the same need belong on the same page. Splitting them into two pages produces two thin texts competing with each other, and that is not caution: it is a choice that divides the work for nothing.

Does the expected answer look like something you can publish? That is the subject of search intent, which deserves its own guide. An enormous demand whose expected answer is a price comparison engine is not handled with an article, however carefully written.

What does not work, or no longer works

Two very widespread methods are contradicted by the documentation itself.

Aim for a keyword density around three percent, and write articles of at least two thousand words: long content ranks better.

Length on its own plays no part in ranking, and the official documentation is explicit about it. As for density, it is a writing instruction aimed at a machine rather than a reader: it produces texts that repeat a phrase instead of handling a question.

The point about length deserves quoting directly, given how widely it is contradicted:

Content length on its own plays no part in ranking.

officialSource : Google Search Central, “SEO Starter Guide”

What that does not say is that a short page equals a long one. A subject treated seriously often needs room; the room is simply a consequence of the treatment, never a target to hit. The difference shows up within three paragraphs of reading.

That leaves the last reflex, more expensive than it looks since it is paid for when the domain is bought:

Keywords placed in the domain name have almost no effect.

officialSource : Google Search Central, “SEO Starter Guide”

A domain name is kept for ten years. Loading it with commercial terms for an effect documented as negligible means paying for a long time for a decision made in a day.

Sources

  1. Performance report (Search results): Overview and basic setup — Google Search Consoleaccessed on
  2. SEO Starter Guide: The Basics — Google Search Centralaccessed on
  3. In-depth guide to how Google Search works — Google Search Centralaccessed on
  4. Creating helpful, reliable, people-first content — Google Search Centralaccessed on

Block revalidated on

Frequently asked questions

Do you need a paid tool to do keyword research?
Not at the start. Search Console gives you the queries that actually showed your pages, which no third-party tool can know on your behalf. Paid suites are mostly useful for exploring a market you do not touch yet, and that is a different job entirely.
What are the search volumes shown by tools worth?
They give orders of magnitude and let you compare two demands against each other. They measure nothing checkable: no calculation method is published, and two tools regularly display different values for the same phrasing. Treat them as hints, not readings.
How many keywords should a page target?
The better question is different: how many distinct demands does this page really handle? A page answers one demand and its close rephrasings well. Two genuinely different demands almost always end up deserving two separate pages instead of one crowded page.
Should the keyword go in the domain name?
No, and the official documentation says so bluntly: keywords in the domain name have almost no effect. A short, memorable, pronounceable name will serve you longer than one stuffed with commercial terms that nobody manages to remember or repeat.