What you'll learn in this article
- Why how to choose keywords for google ads is a filtering decision, not a collecting one
- The mistakes I watch people make on almost every new account
- The exact three questions a term has to pass before it goes into a campaign
- How I decide what stays out even when the volume looks tempting
- How my own search terms report tells me which choices actually held
The hardest part of how to choose keywords for google ads isn't finding terms, it's refusing them. Anyone can generate a thousand keyword ideas in ten minutes; the whole skill is deciding which of them deserve to spend money, and that decision is a filter, not a collection exercise. When I inherit an account that's bleeding budget, the problem is almost never a shortage of keywords. It's the opposite: a bloated list where nobody ever asked, term by term, whether the person typing it actually wanted what the ad was about to show them. So this article is less about where to find keywords and more about the criteria I use to let one in.
Everything I do here rests on a single idea: a keyword is a bet on intent I can't fully see until traffic arrives. Before launch, every term is an inference about what a searcher wants, and my job is to make that inference as disciplined as possible so I'm not paying to discover the obvious. The advertisers who waste the most are the ones who treat keyword selection as a brainstorm and let volume decide. The ones who win treat it as a series of small, defensible judgements about whether a query and their offer genuinely meet.
The mistakes I see people make constantly
The most common error, by a wide margin, is confusing relevance to the business with relevance to the searcher. Someone selling accounting software adds "accounting" because it describes what they do, never stopping to ask what a person typing that single word actually wants. Usually they want a definition, a job, or a free spreadsheet, not a €200-a-month subscription. The term is perfectly relevant to the advertiser and almost useless as a purchase signal, and that gap is where most wasted spend lives. Choosing the right keywords for Google AdWords starts by moving the question from "is this about my product" to "does someone searching this want to act right now."
The second mistake is worshipping volume. A big search number feels like opportunity, so people load up on head terms and quietly assume the traffic will sort itself out. It won't. A one-word query hides a dozen different intents behind an impressive figure, and you pay for all of them. I've watched accounts burn a month's budget on a single broad head term that looked like the biggest opportunity on the list and converted like the worst, precisely because its volume came from people who were never going to buy.
The third is skipping the exclusion side entirely. People obsess over which keywords to add and never think about which searches to keep out, so their carefully chosen terms get dragged into queries they'd never have picked on purpose. You can't choose keywords well in isolation from the searches you refuse; the two decisions are the same decision. That's why I treat building the negative keyword list as part of choosing keywords, not a chore for later, because the terms you block define the intent of the terms you keep just as much as the keywords themselves do.
My criteria: the three questions a term has to pass
When I decide whether a keyword goes into a campaign, I run it through three questions in order, and a failure on any one keeps it out regardless of how good it looks elsewhere. The first is intent: does the searcher want to buy or act, not merely learn? "Best CRM for small business" wants to act; "what is a CRM" wants to read. Both are relevant to a CRM vendor, but only one belongs in a conversion campaign, and pretending otherwise is how research traffic ends up charged to a sales budget.
The second question is capability: can I actually serve what this person is looking for? A locksmith who only does residential work has no business on "commercial safe cracking," however high the intent, because the click lands on a page that can't help. This sounds obvious and gets violated constantly, usually because a keyword tool suggested a lucrative-looking adjacent term and nobody checked whether the business could fulfil it. Relevance runs in both directions, and a term the searcher means and I can't answer is as wasteful as one I can answer and they don't mean.
The third question is readability: is the query specific enough that I can confidently guess the intent behind it? This is where I lean toward long-tail keywords for Google Ads, because a longer, more specific query tells me what someone wants clearly enough that I can write a matching ad and choose the right landing page. Google's own guidance is aligned here: its documentation on building a keyword list recommends specific keywords that directly relate to your ad's theme precisely so your ads don't show for irrelevant terms. The more specific the term, the less I'm guessing, and the less I'm guessing the more defensible the choice.
What goes in, and what stays out
Put those three questions together and the in-or-out decision gets refreshingly blunt. A term that clears intent, capability and readability goes in without hesitation, because I can predict who it reaches, write copy that matches, and point it at a page that delivers. A term that fails intent stays out even if I could serve it and even if it's specific, because I refuse to pay conversion prices for research traffic. And a head term that only passes on volume, failing readability, stays out of the initial build entirely, to be reconsidered later only if my own data earns it a place.
The stay-out list is where discipline actually shows. It's easy to add the term that clears every bar; the judgement is in refusing the one that's relevant to your business, has huge volume, and fails the intent test, because that's the exact term that feels like a mistake to leave out. I keep it out anyway and let the account prove me wrong with evidence, rather than letting a big number talk me into a bet I can't read. This is the same restraint I apply across keyword ideas generally: the tool's job is to propose, mine is to refuse, and a shortlist of terms I can defend beats a long list I merely assembled.
Match type is the last lever on the in-or-out call, because it decides how tightly a chosen term is allowed to behave. A high-intent term I trust can run on broader matching to catch variants; a general term I'm only half-sure of goes in on exact match, if at all, so it can't wander. Understanding the types of keywords in Google Ads is what lets me admit a slightly risky term under tight control instead of banning it outright, which keeps the list both defensible and complete without leaning on volume to make the decision for me.
Validating the choice with my own data
Everything above is inference, however careful, so the method only closes once real queries arrive. That evidence is the search terms report, and it doesn't exist until the campaign runs. After a week or two it shows me the exact searches my keywords actually matched, which is the first genuinely factual data in the whole process. Everything before launch was me guessing intent from the shape of a query; this is the ground truth that tells me which guesses held and which pulled traffic I never wanted.
The pattern is consistent enough that I plan around it. The specific, high-intent terms I chose confidently almost always match sensible queries, which keeps validating the intent-first order. The head terms I let in against my better judgement are where the junk shows up, and that report is where I catch it and cut it. I feed every correction back into the list, promoting terms that proved out and demoting the ones that didn't, so the account I choose keywords for at launch is never the account I'm running a month later.
Put the whole thing together and choosing keywords for Google Ads is a filter run in a strict order: intent decides whether a term is worth reaching, capability decides whether I can serve it, readability decides whether I can trust my read of it, and the search terms report decides, after launch, whether I was right. There's no perfect list waiting to be exported, only a set of defensible bets and the discipline to refuse the ones you can't defend. The advertisers who lose treat keyword selection as collecting; the ones who win treat it as deciding, term by term, what earns the right to spend.