Look, if you've ever inherited one of those "spray and pray" Google Ads accounts, you already know what I'm talking about.
500 to 1,500 keywords per campaign. All three match types for basically the same thing. 20% of your spend disappearing into keywords that never convert.
On paper? Looks super sophisticated. In practice? It's like trying to tune a race car while wearing oven mitts. You can see all the problems, but you're not "allowed" to fix them because "they worked back in 2019."
And then comes the question I hear on literally every client call:
"How many keywords should we actually be using?"
Everyone wants a magic number. "Is it 10 per ad group? 50? 200?" But here's the thing - in 2026, the real answer has less to do with the count and more to do with your structure, your intent, and your budget.
This isn't me theorizing. This is a practical framework you can use on a messy account tomorrow morning.
Why "more keywords = better" stopped working
Here's how everyone used to think about it:
More keywords = more coverage
More coverage = more impressions
More impressions = more opportunities
That logic kind of made sense when match types were strict, CPCs were cheaper, and you had real control with manual bidding. But now? With broad match, close variants, and smart bidding doing whatever they want, cramming 1,000 keywords into a campaign doesn't give you more control. It just fragments your data and hides what's actually working.
You end up with the same pattern every time: 20% of spend going to queries that never convert, keywords spending $5 over 30 days month after month, and the exact same intent duplicated across broad, phrase, and exact. You're not running a sophisticated account. You're running three versions of the same idea and starving all of them.
So instead of chasing some magic number, you need to answer three actual questions: What intent are we even targeting? Do we have enough budget per keyword to learn anything? And at what point do extra keywords stop helping and start diluting everything?
Step 1: Start with intent buckets, not a keyword quota
Here's the truth - keywords are just handles for intent. That's all they are in 2026.
Nobody wakes up thinking "I need 300 keywords today." They wake up thinking "I need more qualified plumbing leads in Portland" or "I need more demo requests from payroll managers."
So start simple. Break it down like this:
Top of funnel - research and "what is" searches
Middle of funnel - comparison shopping and looking at options
Bottom of funnel - high intent, "ready to buy" or "ready to book" terms
For most lead-gen and local service campaigns, the bottom-of-funnel bucket is where all your profit lives. That's where you want most of your budget and your cleanest structure.
Once you're clear on intent, the question becomes:
"How many distinct intents does this campaign actually need?"
Not how many keyword lines you can cram into Ads Editor. How many actual intents.
Step 2: Work backwards from budget and CPC
Okay, this is the uncomfortable part most people skip.
You can't decide how many keywords you "need" without looking at your budget and your CPC. If a keyword can only spend a few bucks over 30 days, you're never going to collect enough data to know if it's any good.
That's not testing. That's just clutter.
Here's a practical way to think about it: Figure out your average CPC for a theme, decide how many clicks per month you need to actually make a decision, then multiply those together - that's your minimum monthly budget for that keyword.
If your average CPC is $5 and you want at least 20 clicks per month, that keyword needs about $100/month behind it. Now multiply that by the number of keywords in your ad group and compare it to your actual daily budget.
Suddenly "we have 300 keywords" stops sounding impressive and starts sounding like "we're giving each keyword coffee money and expecting it to buy us dinner."
Step 3: Tight themes, 10-20 workhorse keywords per ad group
Once you align on intent and budget, the keyword count kind of sorts itself out.
In most healthy search campaigns, here's what you'll see: Each ad group is built around one clear intent. You've got maybe 10 to 20 high-intent keywords that are closely related. And those keywords all share language you can actually mirror in your ad copy and landing pages.
You don't need 75 different ways to say "emergency plumber near me." You need a tight cluster of high-intent variants that all point to the same promise: fast, local, emergency help.
When your ad groups are built this way, your RSAs don't have to do gymnastics to stay relevant, your landing pages can actually echo the query language, and your search terms report becomes readable instead of a wall of random noise.
The test isn't "do we have enough keywords?" The test is "could a normal person look at this ad group and instantly see what problem it's trying to solve?"
Step 4: Stop duplicating match types
One of the quiet killers in these bloated accounts? Match-type duplication.
The exact same idea gets added as broad, phrase, and exact across multiple ad groups. On paper that looks like more coverage. In reality, you're bidding against yourself, you're scattering data across three versions of the same intent, and you're making it way harder to see what's actually pulling its weight.
In 2026, a more sane pattern looks like this: Pick your core, bottom-of-funnel themes. Use broad or phrase to sweep in variants and long-tails. Use negatives and structure to keep that traffic in the right intent bucket.
You can still layer in exact where it truly matters, but you're not building a museum of slightly different keyword spellings just to feel like you're in control. The control comes from your intent design, your negatives, and your conversion definitions - not from having six near-identical lines in the interface.
Step 5: Use a tier system instead of hoarding "just in case" keywords
Look, most of the anxiety in these conversations isn't about structure. It's emotional.
Clients are protective of old campaigns because they built them. They're afraid if you pause keywords, impressions will die. You feel like you're talking to a wall every time you suggest pruning anything.
The cleanest way through this? Don't argue about "too many" or "too few." Introduce tiers instead.
Tier 1 - High-intent, proven keywords that consistently drive conversions. These get the bulk of budget and the cleanest structure.
Tier 2 - Promising ideas with some volume or early signals. These live in clearly labeled test campaigns with modest budgets.
Tier 3 - Long-tails and speculative ideas that sound nice but have zero data. These don't live in your main conversion campaign. They're parked, documented, and only promoted if there's budget and a reason.
Now you're not saying "we're deleting your hard work." You're saying "we're promoting the best performers to Tier 1 and parking Tier 3 ideas so they don't steal budget from what's already working."
The keyword count drops naturally once you stop pretending Tier 3 deserves the same budget access as Tier 1.
So what's the actual number?
If you absolutely need a range, here it is for a focused lead-gen or local service account:
Per ad group: roughly 10 to 20 tightly themed, high-intent keywords around a single problem or offer
Per campaign: enough ad groups to cover your real intent buckets - usually a handful of core themes, not every possible edge case from Keyword Planner
But honestly? The more useful answer is this:
You need as many keywords as you can actually fund with real testing budget, inside clean intent buckets, without fragmenting your data.
If you've got the budget and volume to run 200 high-intent keywords properly? Great. Do it. But if you don't, forcing 200 keywords into the interface doesn't make your campaign advanced. It just hides the truth longer.
In a world of broad match, smart bidding, and AI-driven auctions, your value isn't in memorizing some magic keyword number. It's in designing campaigns around intent, stripping out ego and nostalgia, and making sure every dollar has a real chance to come back as revenue.
The question isn't "how many keywords are you using?"
The question is "how many of your keywords are actually allowed to do their job?"