Keyword Research for Google Ads: What to Pay For

Keyword versus search term, and the four tiers you sort every list into.

Most people arrive at Google Ads with an SEO habit: find the biggest numbers, write them into a spreadsheet, start.

That habit is expensive here.

In organic search a term you cannot convert costs you some writing time. In paid search it costs money on every click. So the job inverts: a high-volume keyword you cannot convert is not an opportunity, it is a line item.

This post is the paid half of keyword research. It sits under our Google Ads guide and deliberately does not repeat the organic process — that lives in our keyword research guide, written for ranking, not buying. Read that one to build an SEO map; read this one before you spend.

How ads keyword research differs from SEO keyword research

Both start with a list of search terms. They diverge immediately on the question being asked.

That last row is the one that catches people. SEO research is front-loaded. Ads research is continuous, and the account itself is the data source.

Keywords and search terms are not the same thing

A keyword is what you bid on. A search term is what a person typed. Those two used to be close to identical. They are not any more, and understanding the gap is the single most useful thing on this page.

Google's own definitions are blunt about it. A search term is "a word or set of words a person enters when searching on Google", while a keyword is "a word or set of words that Google Ads advertisers can add to a given ad group" — and the same help page notes that "depending on your keyword matching options, the search terms listed might be different from your keyword list" (Google Ads Help, About the search terms report — the page prints no last-updated date; read 8 October 2026).

Why they drift is a matching question. Google documents broad match as showing ads on searches "that are related to your keyword", which "can include searches that don't contain the direct meaning of your keywords"; phrase match on searches "that include the meaning of your keyword"; and exact match on searches "that have the same meaning or same intent as the keyword" (Google Ads Help, About keyword matching options — no printed date; read 8 October 2026).

Read those three definitions again and notice that none of them promises the words you chose. Even exact match is matched on meaning. So your keyword list is a request, and the search terms report is the receipt.

The mechanics live elsewhere: our match types post covers when each type earns its place and how it interacts with your bidding strategy.

The search terms report is the real research tool

Keyword Planner tells you what the world might search. The search terms report tells you what your money actually bought. One is a forecast; the other is evidence.

Google describes the report as showing "how your ads performed for the searches people actually made", listing terms "that many people used and that triggered impressions and clicks", and showing which of your keywords matched each search (same help page, read 8 October 2026). That last column is where you see a keyword you were proud of pulling a query you would never have chosen.

How to work it, once you have a week or two of data:

1. Sort by cost, descending. Not by clicks, not by impressions. You are looking for where the money went, and the expensive terms are rarely the ones you planned.
2. Read the matched-keyword column. Every surprise is a match-type decision to revisit.
3. Split the list three ways. Promote to a keyword, leave alone, or block. Terms you promote get their own keyword so you control the bid and the ad; terms you block become negatives.
4. Check which ad group the term landed in. A good term in the wrong ad group gets the wrong ad — a structure problem, not a keyword problem.
5. Repeat weekly at first, then fortnightly. New phrasings appear constantly; the report never finishes.

Sort your raw list into four intent tiers

A raw keyword list is not a plan. Before anything gets uploaded, sort every term into one of four intent tiers: buy-now, compare, research, and not-us. The tier decides the bid, and in a small account it decides whether the term gets any budget at all.

What you do with the tiers depends entirely on how much you have to spend.

• Small budget. Buy-now and brand only. Nothing else. A tight list on a small budget beats a broad list every time, because a broad list spends the whole day before the good terms get a turn.
• Mid budget. Add the compare tier in its own campaign with its own budget, so it cannot borrow from buy-now.
• Larger budget with measurement in place. Research terms become arguable — but only once conversion tracking can show you what a top-of-funnel click eventually did. Without that, a research tier is a donation.

Long, specific phrases usually land in buy-now or research, rarely in between — our long-tail keywords post explains why.

Negative keyword research is half the work, not an afterthought

Treat the negative list as a research output with the same status as the bid list. You build one before launch and then harvest it forever.

The pre-launch list. Written before a single rupee moves, from what you already know about your market:

• Price-blockers: free, cheap, lowest price, discount — unless those words describe your actual offer.
• Career queries: jobs, salary, vacancy, internship, resume, course.
• Self-service intent: DIY, at home, template, sample, how to make.
• Learning intent on a transactional campaign: "how to", "what is", "meaning", "full form", "pdf".
• Competitor brand names, where you have decided not to buy them. Deciding is the point; drifting into them is not a decision.
• Adjacent categories that share your words but not your product — the ones a broad match will find for you within a week.

The ongoing harvest. Every search terms review ends with new negatives. That is the loop working, not a sign that something is broken.

Three documented behaviours stop a negative list from doing what people assume it does. Google states that "negative keywords won't match to close variants or other expansions"; that for Search "you'll need to add synonyms and singular or plural versions if you want to exclude them"; and that your ad may still show when the negative keyword appears after the sixteenth word of a search. Google also warns that "if you use too many negative keywords, your ads might reach fewer customers" (Google Ads Help, About negative keywords — no printed date; read 8 October 2026).

Which means a negative list is not a rule you write once. It is a list of literal strings, in singular and plural, with the misspellings you actually see. The mechanics of where to add them — account, campaign or ad group — and which match type to use are covered in our negative keywords post; this page is only about generating them.

What Keyword Planner can and cannot tell you

Keyword Planner is a planning aid, not a measurement tool, and Google says so in plain language. Its own description of the product is "view estimates on the number of searches a keyword gets each month" (Google Ads Help, About Keyword Planner — no printed date; read 8 October 2026).

The limits worth knowing before you build a plan on it, all from Google's forecasts documentation (Google Ads Help, About Keyword Planner forecasts — no printed date; read 8 October 2026):

• Volume is a rounded twelve-month average. "By default, the number of searches for the term (regardless of language) is averaged over a 12-month period", and the figures are rounded — which is why totals across locations may not add up. Treat what you see as a band.
• Volume and forecast are measured differently. Search volume reflects exact keyword matches only, while forecast impressions reflect the match types you select. Those two columns are not describing the same thing.
• Forecasts run on one week of data. They are "based on one week of data, and are averaged to provide daily forecasts", refreshed daily on the last seven to ten days and adjusted for seasonality.
• A new account gets other people's history. Google states that for new accounts "your forecasts are based on historical average data for all advertisers" — so your first forecast is not about your business at all.
• Small geographies are less accurate. "You can expect results to be less accurate if you're targeting a small geographic location" — worth remembering before you plan a single-city India campaign off planner numbers.

None of that makes the tool useless. It is good at one job — finding phrasings you had not thought of. Use it for discovery, never for the business case.

One related status is worth knowing. Google gives a keyword "low search volume" when it has "very little to no search history on Google", judged worldwide over the past twelve months, and says the status is "unrelated to quality score, bid, or creatives" with "no negative impact" on the account (Google Ads Help, Low search volume — no printed date; read 8 October 2026). So keep genuinely specific terms; they switch on when demand appears. Relevance is a quality score question, not a volume one.

The India layer: one intent, three spellings

In India the same buying intent arrives in several written forms, and a list built only in English will miss paying traffic while a careless broad match hoovers up the rest.

Three things to handle deliberately:

• English and transliterated Hindi for the same intent. People type Hindi words in Roman script constantly — "kitna", "kaise", "sasta", "kaha milega" — often mixed into an otherwise English query. These are real commercial searches when attached to a product word, and they will appear in your search terms report whether or not they are in your keyword list.
• Brand-name misspellings, yours and your competitors'. Google explicitly lists "significantly misspelled" terms among those that get low search volume status, so some of these will sit inactive and switch on later. Add them anyway; defending your own name cheaply is one of the few free wins in a Search account.
• City and locality names in their local spelling. Suburb and locality names are written several ways in practice. Check the search terms report for the variants your market actually uses rather than guessing from a map.

The honest method is not a translation exercise. Launch in English, keep match types tight enough to stay readable, and let the report tell you which transliterated forms your buyers actually use.

Where to go from here

If you take one habit from this page, make it the weekly search terms review. It is the only keyword research that uses your own money as the dataset.

Set expectations honestly on timing: paid search gives you data within days, but it stops the moment the spend stops. Organic work on the same keyword map takes roughly three to six months to show meaningful movement, then compounds. Most accounts need both, doing different jobs.