The short version: good Google Ads keyword research isn’t about finding clever keywords, it’s about ruthlessly narrowing down to the handful of phrases that match buying intent, then spending the first two weeks watching your search terms report and cutting everything that doesn’t convert. Most of the “keyword research” advice out there stops at building a list. The real work starts after you launch.
Why Google Ads keyword research isn’t the same job as SEO keyword research
I get asked this a lot, usually by someone who’s spent three hours in a keyword tool building a spreadsheet with 400 rows and then wonders why their campaign is bleeding money. Here’s the difference nobody explains clearly enough: SEO keyword research is about relevance and volume. Google Ads keyword research is about intent and money, right now, this week.
A term like “how to fix a leaking tap” is a brilliant SEO keyword. It’s a terrible Google Ads keyword for a plumbing business, because the person typing it wants a free answer, not a callout fee. Meanwhile “emergency plumber Bristol” has less search volume and a higher cost per click, but every single person typing it has their credit card in hand. Volume is not the goal. Money is the goal.
If you’ve read my post on why most small businesses waste half their Google Ads budget, you’ll know this is where most of that wasted spend starts, at the keyword list stage, before a single ad has even gone live.
Start in Google Keyword Planner, but don’t trust it
Open Google Keyword Planner (it’s free inside your Google Ads account, under Tools and Settings, then Planning) and type in your main service. Say you run a small accountancy firm in Manchester. Type “accountant Manchester” and Keyword Planner will spit out fifty or so variations with search volume ranges and suggested bids.
Two things to know straight away:
- The search volume figures are ranges (like “1K to 10K”), not real numbers, because Google deliberately blurs them unless you’re actively spending on ads. Once you have a live campaign, the ranges tighten up.
- The suggested bid is Google’s guess at what will win an auction, not what you’ll pay. I’ve seen suggested bids of £8 turn into real average CPCs of £2.40 once Quality Score kicks in.
Use Keyword Planner for the raw list, but treat every number it gives you as a rough guide, not gospel. If you want a way to sanity check that list without paying for another tool, my guide on how to do keyword research for free without expensive tools walks through using autocomplete, Reddit threads and “People Also Ask” boxes to add real, human-phrased terms that Keyword Planner tends to miss.
The uncomfortable bit: your first list is a guess, and that’s fine
Here’s what almost nobody says out loud. No matter how good you are at this, your first keyword list for a new campaign is an educated guess. You do not know which of your thirty keywords will convert until real people type real searches and click your ad. Anyone who tells you they can build a perfect keyword list on day one, before a single click has happened, is selling you something.
The actual research happens after launch, inside your search terms report. That’s the list of exact phrases people typed before your ad showed up, and it’s where you find out you’ve been paying £3.20 a click for “accountant jobs Manchester” because you added “accountant” as a broad match keyword and Google decided that was close enough.
Match types don’t work the way most guides pretend they do
This is the part that trips up even experienced advertisers. Google’s official documentation still talks about exact match, phrase match and broad match as if they behave the way they did in 2015. They don’t. Since Google rolled out “close variant” matching across all match types, an exact match keyword like [accountant Manchester] can now trigger for “chartered accountants near Manchester” or “Manchester accountancy firms”. Exact match hasn’t meant exact for a long time.
What this means practically is that match type is now a dial, not a switch. Broad match will show your ad for a wider net of searches, including some rubbish, but Google’s machine learning is decent at broad match when you feed it strong conversion data. Exact match gives you more control but less reach, and it’s not as tight as the name suggests. I run most new campaigns with a mix, phrase match for the core terms I know convert, and a small broad match test budget (usually 15 to 20 percent of daily spend) to see what Google’s algorithm surfaces that I hadn’t thought of.
A real example from a client
Last year I worked with a small kitchen fitting company in Kent. Their previous agency had built a keyword list around “kitchen fitters”, “kitchen installation”, “bespoke kitchens”, all sensible-sounding, all broad match. Within three weeks their search terms report showed they’d spent £340 on clicks for “kitchen fitter jobs”, “kitchen fitting course” and “how to fit a kitchen yourself”. None of those people were buying a kitchen. They were job hunting or trying to save money by doing it themselves.
We didn’t add a single new keyword to fix it. We added negative keywords: “jobs”, “course”, “yourself”, “DIY”, “salary”, “apprenticeship”. Cost per lead dropped from £61 to £34 in the following month, with the same keyword list and the same budget. The keywords weren’t the problem. The lack of negative keywords was.
Negative keywords are the research most people skip
If I had to pick one habit that separates a Google Ads account that makes money from one that doesn’t, it’s this. Building a positive keyword list feels productive. Building a negative keyword list feels like admin. Most people do the fun bit and skip the boring bit, and then wonder why their cost per lead is double what it should be.
Before you launch, sit down and write a negative keyword list covering:
- Job-seeking terms: jobs, career, salary, apprenticeship, hiring, vacancy
- Free and DIY terms: free, DIY, how to, template, guide, tutorial
- Wrong location terms: any towns or cities you don’t cover, if you’re a local business
- Competitor brand names, unless you’re deliberately running a competitor conquesting campaign
- Irrelevant qualifiers: for a “kitchen fitter” business, that’s words like “doll”, “toy”, “cabinet” (as in filing cabinet, not kitchen cabinet)
Then, every single week for the first month, and monthly after that, go through the search terms report and add whatever’s wasting money. This is not a one-off task. It’s ongoing housekeeping, like clearing spam out of your inbox.
Step by step: the process I use
- List your services in plain English. Not “solutions”, not “offerings”. What does a customer type when they need this? Write ten to fifteen core phrases.
- Run each through Google Keyword Planner to get variations, search volume ranges and competition level.
- Check the intent of every term. Ask yourself: would someone typing this pay me today, this week, this month? If the honest answer is “maybe someday” or “no, they want free advice”, cut it.
- Look at what your competitors are saying. I regularly check the Facebook Ads Library for competitor research, not because it’s Google Ads, but because it shows you the exact offers and phrases competitors are testing right now, which often reveals angles and terms you hadn’t considered for your own Google campaigns.
- Build your negative keyword list first, before launch, using the categories above.
- Launch with a mix of phrase and broad match, roughly 80/20, with a modest daily budget (£15 to £30 a day is enough to gather useful data for most small local businesses within two weeks).
- Check the search terms report every 3 to 4 days for the first fortnight. Add negatives, pause anything with zero conversions after 15 to 20 clicks and a reasonable spend against it.
- Promote winning search terms to their own exact match keywords so you can control bids on them individually, rather than leaving them buried inside a broad match group.
- Repeat monthly. Search behaviour changes with seasons, news events and your own reputation growing, so a keyword list from January is stale by June.
Where AI earns its place in this process
I use AI tools to speed up two parts of this, never the whole thing. First, turning a rough service list into fifty phrased variations in seconds, which I then manually filter for intent. Second, scanning a messy search terms report and grouping similar phrases so I can spot patterns faster than scrolling a spreadsheet. What I don’t do is let an AI tool decide what’s negative and what isn’t, because it doesn’t know your margins, your capacity, or that “kitchen fitter” searches from your own town centre postcode are worth ten times more than the same phrase from forty miles away. If you want the wider context on using AI without it backfiring on your visibility, my piece on using AI for SEO without getting penalised by Google covers the same caution that applies here: AI speeds up the grunt work, it doesn’t replace judgement.
And if all of this sounds like more time than you have spare on top of running the actual business, that’s a fair reason to bring in help rather than a reason to skip it. An AI consultant for small businesses can set up the tracking, the negative keyword lists and the review cadence the first time, which usually pays for itself within the first month or two of reduced wasted spend.
The mistakes I see over and over
- Chasing volume over intent. A keyword with 50 searches a month and clear buying intent beats one with 5,000 searches and vague curiosity, every time.
- Never checking search terms after launch. Building the list once and walking away is the single most expensive habit in Google Ads.
- Treating exact match as a guarantee. It isn’t, and hasn’t been for a while. Close variants mean you still need negatives even on exact match campaigns.
- Copying competitor keyword lists wholesale. Their margins, capacity and customer value aren’t yours. A keyword that’s profitable for a competitor with a £2,000 average sale might lose you money on a £200 sale.
- Ignoring location and device data. A keyword can look average overall while converting brilliantly on mobile in three specific towns and doing nothing everywhere else.
Frequently asked questions
How many keywords should I start with in a new Google Ads campaign?
Start small, around 15 to 25 tightly related keywords grouped into two or three ad groups by service or intent. A bloated list of 200 keywords with tiny budgets spread across all of them just means you never get enough clicks per keyword to learn anything useful.
Is Google Keyword Planner enough on its own, or do I need a paid tool?
Keyword Planner is enough to get started and it’s free inside your Google Ads account. Paid tools like SEMrush or Ahrefs add competitor gap analysis and more precise volume data, which is useful once you’re spending several thousand pounds a month, but not essential when you’re starting out or working with a smaller budget.
How often should I update my keyword list once a campaign is running?
Check your search terms report every 3 to 4 days for the first two weeks of a new campaign, then weekly for the first two months, then monthly once things settle down. Search behaviour shifts with seasons and competitors, so a list left untouched for six months will always underperform a list that’s actively maintained.
Should I use broad match or exact match for Google Ads keywords?
Use a mix. Phrase match for terms you already know convert well, and a smaller portion of your budget, around 15 to 20 percent, on broad match to let Google’s algorithm surface new search terms you hadn’t thought of. Exact match no longer means exact, since Google applies close variant matching across all match types, so negative keywords matter regardless of which match type you choose.
