A keyword is a word or phrase that people type into a search engine, and that anyone working on a website uses as a label to group together all the searches that are similar to each other.
The distinction matters straight away. The keyword written in a spreadsheet almost never matches the phrase someone actually typed: it is an approximation, and you choose it by looking at Google's first page for that search, not at the volume figure next to it.
Keywords and queries are two different things
The query is what the user types. The keyword is the name we give to a family of queries that want the same thing. Tools show you the second; Search Console shows you fragments of the first.
Google sees billions of searches a day, and 15% of them it has never seen before. The figure comes from Pandu Nayak, then head of Search, in the BERT announcement of 25 October 2019. No keyword list can contain them, because when you compile it those searches do not exist yet. It is one of the reasons volumes remain rounded estimates, and why keyword research done only with tools leaves part of the demand out.
In the same text Google uses a word worth borrowing: keyword-ese. It is the dialect people speak when they think they have to write like a machine, "vegan restaurant Manchester city centre" instead of "where can I get vegan food in Manchester tonight". Google was already asking people to stop speaking it back then.
What Google does with them
The documentation on how Search works is explicit: the most basic signal of relevance is that the content contains the same words as the query. Two lines later comes the limit, with Google's own example: someone searching for "dogs" does not want a page with the word "dogs" repeated hundreds of times, and the systems assess whether there is something else around it, images, videos, a list of breeds.
Since 2019 exact matching has counted for even less. BERT changed how one in ten English-language searches in the US were interpreted, and the example Google gave is "2019 brazil traveler to usa need a visa": the preposition to decides the direction of travel, and it used to be ignored.
The practical consequence is that counting occurrences is pointless. Keyword density is not a Google metric, and overdoing it has a name and a dedicated policy: keyword stuffing.
In the UK, "keyword" and "SEO keywords" are not the same search
On 26 September 2026 we queried Google UK on desktop for five variants of the same concept and counted how many URLs in the organic top 10 repeat. Between "keyword" and "seo keywords" there is not a single one.
| Search | Average monthly volume in the UK | URLs shared with "keyword" | What the first page shows |
|---|---|---|---|
| keyword | 5,400 | — | Twinkl primary school word lists, a dictionary, a few SEO blogs, a children's services procedures manual |
| key words | 5,400 | 1 of 8 | Twinkl reading resources and UK agency guides to keywords |
| keywords meaning | 210 | 0 of 8 | dictionaries (Cambridge, Dictionary.com) and marketing glossaries |
| what are keywords | 320 | 0 of 8 | SEO explainers from Moz, Semrush, Yoast and Mailchimp |
| seo keywords | 720 | 0 of 9 | keyword tools (Semrush, Keyword Tool, Wordtracker, Keyword Planner) and Google's SEO Starter Guide |
Anyone searching for "keyword" in the UK gets a first page where half the results are not about search at all: Twinkl's Key Stage 1 word lists and a matching activity for children, a dictionary entry and a local authority procedures manual. Anyone searching for "seo keywords" gets Semrush, Keyword Tool, Wordtracker and Google's own starter guide. They are two different audiences, and one page does not serve both. Note also the volume column: Keyword Planner gives "keyword" and "key words" the same 5,400 searches, yet their first pages share one URL out of eight. The volume merges what the SERP keeps apart.
The count works the other way too: when two lists of URLs overlap almost entirely, you have a single query and risk two pages competing with each other. The full procedure, with thresholds, is in the entry on keyword clustering.
The only classification that changes the work
There are plenty of labels around: short tail and long tail, branded and non-branded, local, seasonal, related. The one that really decides what you write is search intent: informational, navigational, commercial, transactional. A category page aimed at an informational search will not rank even if it contains the right word everywhere, and you can see that by opening the SERP before writing a single line.
Volume, on the other hand, tells you less than it seems. For Macropix, an Italian manufacturer of custom LED screens, the query "monitor pubblicitario" (advertising monitor) gets 390 searches a month in Italy: it went from position 88 in 2020 to position 2 in 2025, and on the LED wall cluster the domain now has a 25.55% share of voice against Amazon's 18.08%. The full figures are in the manufacturing case study.
In Google Ads the word means something else
In SEO a keyword is an analysis label: there is no field to enter it in and nobody declares it to Google. In Google Ads it is a technical object with documented syntax. Every keyword has a match type that decides which searches can trigger the ad: broad (the default, written without symbols), phrase (in quotation marks) and exact (in square brackets). There are also negative keywords, which only serve to exclude. The documentation on match types is in the Google Ads Help Centre.
| Keyword in SEO | Keyword in Google Ads | Meta keywords | |
|---|---|---|---|
| What it is | a label for grouping similar searches | a targeting criterion you attach a bid to | an HTML tag in the page head |
| Syntax | none | broad, "phrase", [exact] | meta name="keywords" |
| Where it is declared | nowhere | in the ad group | in the page code |
| Effect on results | indirect, through the content | decides whether the ad enters the auction | none since 2009 |
People coming from SEO tend to carry the vocabulary from one world to the other and end up talking about "exact keywords" where none exist. Cost per click also lives only in the second world, even though SEO tools show it next to every row.
Meta keywords have not counted since 2009
The code of many websites still contains a meta name="keywords" tag filled in with care. Google wrote on 21 September 2009, in a post signed by Matt Cutts, that its web search ignores it completely, and it has not changed its position since. The same post clears up a point that still confuses people: the meta description is sometimes used to build the snippet in the results, but it plays no part in ranking.
Where to put a keyword on a page
Title, H1, first paragraph and URL remain the places where matching the words helps, because they are also what the user reads in the results before clicking. The rest of the work is no longer lexical: it is about covering what the SERP shows and competitors leave out. The list of things to fix is in the on-page SEO guide; if instead you already have a keyword list and a site that will not move, that is the kind of problem we work on in our SEO projects.
Keywords FAQs
The query is the exact phrase a person types into the search engine. The keyword is the label the people working on the site use to group all the queries that want the same thing. Google states that 15% of the searches it sees every day are new to it, so part of the queries do not appear in any list: tools show a summary of demand, not demand itself.
One main keyword, plus all the variants the SERP shows to be the same search. Counting occurrences in the text is not a Google metric: the official documentation cites the presence of the same words as the most basic relevance signal, and in the same place explains that repeating the term hundreds of times does not help.
They matter, but they are no longer enough. Inside an assistant people type whole questions, not two words. The practical method is to keep the keyword as the planning unit and add the question versions of the same search, the ones a customer would use when speaking. None of these tools returns the phrases people typed, so checking remains indirect.
Keyword Planner gives them the same UK volume, but Google does not treat them as the same search. In our check of 26 September 2026 on Google UK, the top ten for keyword and for key words had only one URL in common out of eight, and keyword shared none with seo keywords, whose first page is made of tools and Google's SEO Starter Guide. If you write for a professional audience, the more specific phrase is the one to target.
No. Volume is a rounded estimate and new searches do not have one yet. In B2B projects low-volume technical searches are often the ones that generate enquiries: for Macropix, the Italian query monitor pubblicitario (advertising monitor) gets 390 searches a month and went from position 88 in 2020 to position 2 in 2025.
One thing the industry rarely says: the best keyword lists that cross our desks do not come from tools. They come from the questions the sales team hears on the phone, written down the way the customer says them and only then run through a tool to see which have volume. It is the quickest way out of keyword-ese, which is what Google was asking for in 2019 and what conversational assistants have made compulsory.
