How Many Keywords Per Ad Group? Structure That Works

Two bad answers dominate this question. The old one says one keyword per ad group. The new one says throw everything in a pot and let the machine sort it out. The first is a maintenance tax you pay forever. The second is how a plumbing shop ends up showing a drain-cleaning ad to somebody searching for a water heater replacement.

The real rule is not a number. It is a set of constraints: one landing page, one ad set, one margin band per ad group. Get those right and the keyword count picks itself.

How many keywords should one ad group actually hold?

Five to fifteen tightly related keywords is the working range — and the test is simple: if you cannot write one ad that honestly matches every keyword in the group, you have too many.

Google publishes no number. Its guidance in Organize your account with ad groups is that for each ad group you should 'pick a narrow theme and create ad groups around that theme,' and that you should have at least 3 ads in each ad group. That is the whole official instruction. Everything else is practitioner convention.

Stuffing an ad group with near-identical variants buys you nothing. Google states that near-duplicate broad match keywords such as 'red car' and 'car red' are recognized as duplicates, and only the one with the higher Ad Rank is used — you get one bid in the auction either way. Twelve spellings of 'ac repair near me' is not twelve chances to win. It is one keyword and eleven rows of reporting noise.

• Same intent, same page, same ad — keep them together (ac repair, ac not cooling, air conditioner repair).
• Different intent, same service — separate ad groups (ac repair vs ac replacement: one is a $300 call, the other is a $7,000 job).
• Different service — separate ad group, always, because the landing page changes.
• Near-duplicate word orders — pick one. Google dedupes them anyway.

Are single-keyword ad groups (SKAGs) dead?

Yes — and three specific Google mechanics killed them, not opinion. Anyone still selling you a 200-SKAG rebuild in 2026 is selling billable hours.

Close variants collapsed the container. Google's keyword matching documentation now defines exact match as ads that 'may show on searches that have the same meaning or same intent as the keyword.' The bracket is a meaning, not a string. You cannot isolate a single query in its own ad group any more, because the platform no longer works in single queries.

The Quality Score argument is factually wrong. The classic SKAG pitch was that isolating a keyword protects its Quality Score. Google's Quality Score documentation states plainly that Quality Score 'is based on historical impressions for exact searches of your keyword, therefore changing keyword match types will not impact Quality Score.' It also says Quality Score is a diagnostic tool and not an input in the ad auction. Splitting keywords into their own boxes does not manufacture a better score. If Quality Score is your worry, read what actually moves it — it is usually the landing page.

A one-keyword ad group starves the matching model. Google says broad match takes into account 'other keywords in an ad group to better understand keyword intent.' An ad group with one keyword in it gives the system no context to read. And Google's account limits page explicitly lists 'generating a large number of ad groups or campaigns with few creatives or keywords in each' as a pattern that may undermine the stability of its systems — with throttling as the stated remedy. Mass SKAGs are not a clever hack. They are a documented anti-pattern.

When should you split an ad group by landing page?

The moment the honest landing page changes, the ad group splits. One ad group, one page — that is the hard line, and it is the one rule that survives every Google update.

This is not aesthetics. Google's ad group prioritization documentation says relevance 'is determined by the meaning of the search term, all the keywords in an ad group, and landing pages within the ad group,' and it closes with a direct instruction: 'Structure your ad groups thematically to clarify which ad group should match for a given search.' The landing page is an input to which ad group gets picked. Mixed pages, muddy signal.

The second splitter is margin. An emergency call-out and an annual maintenance plan cannot share an ad, because they cannot share a promise. Emergency copy leads with 'on site in 90 minutes' and can carry a $180 cost per lead. Maintenance-plan copy leads with price and cannot. Put them together and the automation will happily buy the cheap clicks with the money you meant for the expensive job. Pair each split with a landing page built for that intent, not a recycled services page.

Should you split campaigns by service, by city, or by margin?

Split campaigns only when a campaign-level setting has to differ — budget, location targeting, ad schedule, or bid target. That is all the campaign layer controls, so it is all the campaign layer should decide.

Google's guidance is the same: 'When you need to use a different set of campaign settings (ex: budget or location targeting), create a separate campaign.' If you want all your ads to share a budget and target the same locations, one campaign is fine. That means service and margin are usually good campaign splits — a $7,000 AC install deserves ring-fenced budget — while city is usually a bad one.

City-splitting is where trade accounts quietly bleed. Three cities times five services is fifteen campaigns, and a $3,000/month budget spread across fifteen campaigns is fifteen budget-starved campaigns. Google warns that a keyword in a budget-restricted campaign 'won't always be able to trigger an ad even if it otherwise could' — so your own structure starts suppressing your best keywords.

There is a measurement floor too. Google's Target CPA documentation recommends evaluating the strategy over the last 30 days 'including at least 30 conversions.' (Note that it also says you can start Target CPA with no conversion history — 30 conversions is the bar for judging it, not for launching it.) If a city split leaves each campaign with four leads a month, you have not gained control. You have gained noise you cannot read.

• Split by city ONLY if: that city has its own landing page and tracked phone number, or it needs its own budget, or its cost per lead is far enough off the others to need a different target.
• Otherwise: keep the cities as location targets inside the service campaign and read city performance in the locations report.
• Never split so far that a campaign cannot reach ~30 conversions in 30 days — you lose the ability to evaluate it.
• Add city names as keywords only where the search volume is real; otherwise let broad and phrase match plus negative keywords do the containment.

What does a five-service, three-city trade account look like?

Five campaigns, not fifteen — one per service, with all three cities as location targets inside each. Here is the structure we would build for a plumbing and HVAC shop covering three suburbs of one metro.

That is 5 campaigns, roughly 15 ad groups, 5 to 15 keywords each, three RSAs per ad group (Google's limit is three enabled responsive search ads per ad group anyway), and one dedicated landing page per ad group. A shop doing $3,000 to $8,000/month in ad spend can run this without any campaign starving.

Cities live as location targets, not as campaigns. You still get city-level reporting, you still get city-level bid adjustments, and you keep the conversion volume concentrated where the bidding model can use it. Split a city out the day it earns its own landing page — not before. This is the structure we build on paid ads engagements for plumbing and HVAC shops.

How does consolidating ad groups help smart bidding learn?

It helps less than the consolidation crowd claims, and the honest reason to consolidate is different: keywords across your own ad groups do not bid against each other anyway, so the fragmentation buys you nothing while costing you signal.

Google is explicit: 'If multiple keywords, targeting the same domain, are eligible to match the same search term, they don't compete with each other in the auction.' One of them is picked. The picking order is identical exact match first, then identical phrase or broad match, then AI-based relevance across ad groups, then Ad Rank. You are not gaining reach by holding the same idea in four ad groups. You are just splitting its history four ways.

Where consolidation genuinely matters is the target. Google says setting individual Target CPA targets per ad group 'isn't recommended as it can restrict Smart Bidding.' Set the target at the campaign, let the campaign carry enough conversion volume to be judged, and keep your structural control in the ad groups and the landing pages. Structure and match type are one decision, not two — how loose you go on broad, phrase, and exact match determines how tight your ad group themes have to be.

What structure mistakes waste the most money?

Four mistakes account for most of the waste in trade accounts, and three of them are structural rather than creative.

• One mega-campaign called 'Plumbing.' The $7,000 repipe and the $99 drain snake share a budget, and the cheap clicks win it every time.
• One landing page for everything. A services page that lists five services is relevant to none of them — and landing pages inside the ad group are an input to which ad group Google picks.
• Splitting by city before you have the volume. Fifteen starved campaigns, none with enough conversions to evaluate, all of them suppressing keywords because they are budget-restricted.
• Rebuilding the account every quarter. Google warns that reorganizing after you have created several campaigns and ad groups 'wipes out all the valuable data you've accumulated, possibly affecting how your ads perform.' Fix forward with negatives and new ad groups instead.

One more, quieter than the rest: no negative keyword discipline between ad groups. If 'ac repair' and 'ac replacement' live in the same campaign, the repair ad group will happily catch replacement searches unless you cross-negate. That is a ten-minute job that protects your highest-margin ad group.

If your account is already built wrong, do not blow it up. Add the missing landing pages, cross-negate between ad groups, and ring-fence the high-margin service into its own campaign. That sequence keeps your history and fixes the leak.

If you want an outside read on how your account is actually structured — which ad groups are cannibalizing which, and where your best service is being starved — that is what our paid ads work starts with. Get my free audit.