How to Structure Ad Groups: The Architecture of Relevance

If you want to know how to structure ad groups, here’s the direct answer: an ad group is a promise of sameness. Every keyword you put inside one shares the same ads and the same landing page, so the rule that governs everything is simple — group only the keywords that can honestly share one message. If two keywords would need different headlines to feel relevant, they belong in different ad groups. Campaigns hold budgets and settings; ad groups hold message themes. Get that division right, and almost everything else in your account — Quality Score, click-through rate, cost per click — gets easier to fix.

Okay, let’s be honest for a second. Ad group structure sounds like the boring, bureaucratic part of Google Ads — the filing-cabinet work you rush through so you can get to the “real” stuff like writing ads and setting bids. I used to think that too. Then I spent enough years inside enough accounts to notice a pattern: the accounts that struggled almost never had a bidding problem or a copywriting problem at the root. They had a structure problem wearing a bidding problem’s clothes. The ads were generic because the ad groups forced them to be generic. The landing pages felt off because one page was trying to answer twelve different intentions. Structure isn’t paperwork, friend. It’s the mechanism of relevance — and relevance is the thing the auction actually rewards. So let’s build yours properly, step by step, in plain language.

What is an ad group, really?

Here’s the mental model that changed how I build accounts: an ad group is not a folder. It’s a contract. When you place a keyword inside an ad group, you’re promising Google — and more importantly, the person searching — that the ads in that group are a fair, relevant answer to that keyword, and that the landing page behind those ads delivers on what the ad said. Keyword, ad, landing page: that’s the message-match chain, and the ad group is the link that holds it together.

The moment you understand that, the structural law writes itself. Keywords that can share one honest message can share an ad group. Keywords that can’t, can’t. “Emergency plumber” and “how to fix a dripping faucet” are both plumbing keywords, but one person wants a human at their door in an hour and the other wants a five-minute video. One ad cannot serve both without lying to somebody.

And here’s the part nobody tells you: this is also why structure matters to your costs. Google’s auction rewards relevance — ads that closely match what someone searched tend to earn better positions and better prices than generic ones. When your ad groups are tight, your ads can be specific; when your ads are specific, they’re more relevant; and relevance is the lever you actually control. If you’ve read my guide on how to lower cost per click, you already know that message match is one of the few CPC levers that doesn’t require spending less or bidding smaller. Structure is where message match is born.

Campaigns vs. ad groups: who does what

Before we go further, let’s settle the division of labor, because mixing these up causes half the messy accounts I see:

• Campaigns hold the business decisions: budget, location targeting, bid strategy, network settings. Split campaigns when you need different budgets or different settings — separate business lines, separate regions, brand vs. non-brand spend.
• Ad groups hold the message decisions: which keywords share which ads and which landing page. Split ad groups when keywords need different messages.

A useful gut check: if the reason you want to separate two things is money or geography, that’s a campaign split. If the reason is “these searches want to hear different things,” that’s an ad group split. If you’re still setting up the foundations, my walkthrough on how to set up your first Google Ads campaign covers the campaign layer in full — this article picks up where that one leaves off, at the level where relevance actually gets built.

What is the one-message test?

This is the whole system in one question, so I’m giving it its own section. Before any keyword goes into any ad group, ask: can one honest, specific ad serve every keyword in this group?

Not “could one vague ad technically apply to all of them” — any ad can be vague enough to apply to anything, and that’s exactly the trap. The test is whether one specific ad, with a headline that names what the searcher actually wants, would feel right to every person searching every keyword in the group. If you find yourself wanting to hedge the headline — making it broader so it sort of covers everything — the group is too broad. Split it.

Run the test in both directions:

• Adding a keyword? Read the group’s current ad. Would someone searching this new keyword feel like that ad was written for them? If not, the keyword needs a different home — maybe a new group.
• Writing an ad? Read every keyword in the group. Does your headline honestly answer all of them? If you’re stretching, the group is hiding two themes inside one container.

I promise this gets easier with practice — after a few weeks it becomes reflexive, the way a good editor hears a clunky sentence. But early on, say it out loud. “Can one ad serve ‘wedding photographer pricing’ and ‘wedding photographer portfolio’?” The answer is no — one searcher is budgeting, the other is judging style — and saying it out loud makes the no obvious.

Do single-keyword ad groups still work?

Time for some era honesty, because a lot of the advice floating around about how to structure ad groups is from a Google that no longer exists.

For years, the power-user dogma was SKAGs — single keyword ad groups. One keyword per group, so every ad could mirror its keyword exactly. In the era when match types were literal and exact match meant exact, that logic held up: maximum granularity meant maximum message match, and the bookkeeping cost was worth it.

That era is over. Google has steadily loosened how match types work — exact match now matches to searches with the same meaning or intent, not just the same words — and its bidding systems learn from data volume. (Match type behavior has changed several times over the years, so verify the current definitions in Google’s own documentation before building around them.) In this environment, single-keyword ad groups mostly create two problems: a maintenance burden that grows faster than your results, and dozens of tiny groups each too starved of data to teach the system anything. Three keyword variants that all mean the same thing no longer need three ad groups — Google treats them as one intent anyway, and splitting them just fragments your signal.

But — and this matters — the pendulum can swing too far the other way. “Consolidate” does not mean “dump fifty keywords into one group and let the algorithm sort it out.” A group that broad forces a generic ad, and generic ads lose the relevance game no matter how much data they accumulate. The goldilocks reasoning goes like this:

• Too granular: unmanageable account, starved data, hours of upkeep for themes Google already considers identical.
• Too broad: one vague ad trying to answer many intentions, weak message match, mediocre relevance everywhere.
• Just right: groups built around themes of close intent — clusters of keywords that genuinely want the same answer, big enough to gather data, tight enough to share one honest ad.

The modern structural unit isn’t the keyword. It’s the intent theme. Everything below builds on that.

How to structure ad groups step by step

Let’s make this concrete with a worked example. (Fictional example: “Maple & Main,” an invented home-cleaning company in a mid-size city, offering recurring house cleaning, one-time deep cleans, and move-out cleans. Any names, keywords, and structures below are illustrative — map the logic onto your own business, not the specifics.)

Step 1: Draw campaign lines where budgets and settings differ

Maple & Main starts with campaign boundaries, which are business boundaries:

• Campaign: Recurring Cleaning — the core service, biggest budget.
• Campaign: Deep Clean & Move-Out — project work, separate budget so it can’t eat the core.
• Campaign: Brand — searches for “Maple and Main” itself, tiny budget, own campaign.

Why does brand get its own campaign and not just its own ad group? Because brand searches behave completely differently — they’re from people who already know you, they convert at their own rate, and mixing them into non-brand campaigns makes everything else look better than it is while quietly distorting your budget decisions. Separating brand from non-brand is one of the first structural moves I make in any account, and it’s a campaign-level split because the budget logic differs, not just the message. (Whether and how to bid on your own brand terms is its own strategic question with real trade-offs — worth deciding deliberately rather than by default.)

Step 2: Theme ad groups by intent, inside each campaign

Now the craft part. Inside the Recurring Cleaning campaign, Maple & Main doesn’t make one big “cleaning keywords” group. It listens for distinct intentions:

• Ad group: House Cleaning Service — “house cleaning service,” “home cleaning service near me,” “residential cleaning company.” Theme: people who want the service, stated plainly. One ad can honestly serve all of these.
• Ad group: Weekly / Recurring Cleaning — “weekly house cleaning,” “biweekly cleaning service,” “recurring maid service.” Theme: people who’ve already decided on recurring — the ad should talk about consistency, the same cleaner each visit, easy scheduling.
• Ad group: Apartment Cleaning — “apartment cleaning service,” “condo cleaning.” Theme: same service, but these searchers named their home type, so the ad should too.
• Ad group: Cleaning Cost / Pricing — “house cleaning cost,” “how much is a cleaning service.” Theme: price-shoppers. They need an ad that talks about transparent pricing and a landing page that actually shows it. Putting these in the general group would waste them — they asked a specific question.

Notice what’s happening: each group is a message theme, not a word-pattern. “Apartment cleaning” could share a group with “house cleaning service” on pure semantics — but the person who typed “apartment” gave you a gift, a detail you can mirror back, and mirroring it is worth a group.

Step 3: Split buying intent from research intent

Here’s a split that pays for itself every time. “Hire a house cleaner this week” and “how often should you deep clean a house” are different humans in different moments. One is ready to buy; the other is learning. If Maple & Main bids on research-style keywords at all, they go in their own ad groups — because they need a softer message, a different landing page (a guide, not a booking form), honest expectations about conversion, and usually different bids. Mixing research keywords into buying-intent groups is how accounts end up with “bad” keywords that were really just mis-grouped ones.

This is also where your match-type choices interact with structure — how tightly each keyword is allowed to expand changes what belongs together. I go deep on that in how to use keyword match types in Google Ads, so here I’ll keep the structural rule light: keep a group’s match types coherent enough that you know roughly what traffic it will attract. A group whose phrase-match keywords pull in wildly varied queries isn’t one theme anymore, whatever the keyword list says.

Step 4: Name everything so Future You can find it

Unsexy, transformative. Six months from now, you (or whoever inherits the account) will navigate this structure by names alone. Pick a convention and never deviate. A pattern that has served me well:

Campaign: [Country/Region] | [Business Line] | [Brand/NonBrand]
Ad group: [Theme] | [Intent: Buy/Research] | [Match style if split]

So Maple & Main’s structure reads: Local | Recurring | NonBrand → Pricing | Buy. Anyone can open this account cold and understand it in ninety seconds. That’s not vanity — findable names are what make the maintenance rituals below actually happen, because nobody maintains what they can’t navigate.

How many keywords should an ad group have?

I know you want a number. I’m not going to give you one, because every “ideal keyword count” you’ve seen was made up, and the honest answer is more useful.

The real constraints are two-sided. On one side, a group needs enough traffic to generate data — modern automated bidding learns from outcomes, and a group that gets a handful of impressions a month teaches the system nothing and gives you nothing to judge. On the other side, a group must stay small enough to share one honest message — the one-message test from earlier is the ceiling.

So the working rule is: as many keywords as can truthfully share one specific ad, and no more. For some themes that’s three keywords; for others it’s fifteen. If you can’t write one honest ad for all of them, split. If a group is so narrow it never accumulates meaningful traffic, merge it into its closest sibling theme. The number falls out of the test; it isn’t an input to it.

One habit that helps: when you draft a new group, write the ad first, then ask which keywords that ad honestly answers. Building in that order makes it nearly impossible to create a junk-drawer group, because the message — the promise — came before the keyword list.

How do you maintain ad group structure over time?

Structure isn’t a thing you build once. Search behavior drifts, Google’s matching loosens and shifts, your business adds services — a structure that was right in January can be quietly wrong by June. Maintenance is three loops at three speeds.

The search-terms loop (monthly, or biweekly for active accounts)

Your search terms report shows the actual queries your keywords matched to — and with today’s looser matching, that list will surprise you. Read it with one question per query: does this search fit the message of the group that caught it? Three outcomes:

• Fits the group: great, leave it (add it as a keyword if it’s recurring and valuable).
• Valuable but doesn’t fit: it deserves its own message — spin up a new ad group where the ad can actually speak to it.
• Not valuable: add it as a negative keyword so it stops leaking budget. (Negative keywords are a discipline of their own, and they’re also what keeps your carefully separated groups from bleeding into each other.)

This loop is what makes structure a living thing. The queries are your customers telling you, in their own words, which themes exist. Your job is to keep the map matching the territory.

The consolidation check (quarterly)

Hunt for starved groups: ad groups that have run for months with barely any impressions or clicks. They feel harmless, but they’re not — they clutter the account, they never produce decision-grade data, and in an automated-bidding world, fragmented data is a real cost. (How much data the bidding system wants has shifted over time as Google’s guidance evolves, so verify current recommendations rather than trusting old thresholds.) For each starved group, merge it into the nearest sibling theme if the combined group still passes the one-message test. If it genuinely can’t share a message with anything, ask the harder question: should you be bidding on this theme at all right now?

The structure review (quarterly)

Once a quarter, look at the whole map instead of the parts:

• Orphan groups: paused or forgotten groups that no longer map to anything you sell. Archive them.
• Overlapping themes: two groups whose keywords have drifted toward each other. Merge or re-draw the boundary.
• Duplicate keywords across groups: the same keyword living in two places means you can’t cleanly control which ad serves, and your own groups end up competing for the same query. Audit for dupes and give each keyword exactly one home.
• Renovate vs. rebuild: for each problem, ask whether a targeted fix solves it or the whole wing needs re-architecting. Default to renovation — see below for why.

When should you restructure — and when should you leave it alone?

Here’s a caution I wish someone had given me earlier: performance history has value, and restructures spend it. When you tear down ad groups and rebuild them, you reset the accumulated learning attached to the old structure — automated bidding has to re-learn, and performance commonly wobbles while it does. A restructure is surgery, not tidying.

So hold a high bar. Renovate — split one overloaded group, merge two starved ones, fix a naming mess — whenever a targeted change solves the actual problem. Rebuild only when the structure is wrong at the map level: themes that don’t match how customers actually search, brand and non-brand tangled together, one everything-campaign straining to hold three business lines.

And when you do rebuild, stage it. Don’t nuke a working account on a Friday. Move one campaign’s worth at a time, keep the old structure paused (not deleted) so you can roll back, and where the stakes are high, run old and new in parallel briefly — new structure live, old structure paused but intact — so a bad week doesn’t become an irreversible one. Boring? Deeply. But boring migrations are the ones that work.

What are the biggest ad group structure mistakes?

Four patterns cause most of the structural pain I see. If you recognize yourself in one, no shame — every one of these is common precisely because it feels reasonable while you’re doing it.

• The everything-group. Fifty loosely related keywords, one generic ad. It feels efficient; it’s actually a relevance sinkhole, because no specific searcher sees a message written for them. The fix is the one-message test, applied ruthlessly.
• The thousand-micro-groups account. The opposite failure: granularity as a security blanket. Unmanageable, data-starved, and — in the modern matching era — mostly redundant, since Google already treats close variants as one intent. Consolidate to themes.
• Keyword duplication across groups. One keyword, two homes, and now your own account is confused about which ad should serve. Run a cross-group duplicate audit quarterly; every keyword gets exactly one address.
• Structure theater. The sneakiest one. Reorganizing ad groups feels like progress — it’s visible, it’s busy, it produces a satisfying before-and-after. But if the real problem is weak ads or a landing page that doesn’t deliver, no amount of restructuring fixes it. Before any reorganization, ask: is the structure actually the broken link in the keyword-ad-landing-page chain, or is it just the easiest one to touch?

Your ad group structure toolkit

Everything above, compressed into the four tools you’ll actually use.

The one-message test card

Ask	If yes	If no
Can one specific, honest ad serve every keyword in this group?	The group is sound.	Split along the message line.
Would a searcher of each keyword feel the ad was written for them?	Keep the keyword here.	Move it to (or create) its true home.
Does the landing page deliver what that ad promises?	Chain intact.	Fix the page before touching structure.

The structure map (the spine)

• Campaigns = budget and settings boundaries: business lines, regions, brand vs. non-brand.
• Ad groups = intent themes: one honest message per group, buying intent separated from research intent.
• Keywords = one home each, grouped by what the searcher wants to hear, not by shared words.

The naming template

Campaign: [Region] | [Business Line] | [Brand/NonBrand] — Ad group: [Theme] | [Intent]. Pick your version, write it down, never deviate.

The quarterly structure checklist

• Search terms reviewed; misfit-but-valuable queries promoted to their own groups; junk queries added as negatives.
• Starved groups merged into sibling themes (one-message test still passing).
• Cross-group duplicate keywords found and resolved — one home per keyword.
• Orphan and overlapping groups archived or merged.
• Naming convention still holding everywhere.
• Any restructure urge interrogated: renovate first, rebuild only with staging and a rollback plan.

FAQ: how to structure ad groups

What’s the difference between a campaign and an ad group?

Campaigns hold business-level decisions: budget, location targeting, bid strategy, and settings. Ad groups hold message-level decisions: which keywords share which ads and which landing page. Split campaigns when budgets or settings need to differ; split ad groups when keywords need different messages.

How many keywords should I put in one ad group?

There’s no magic number, and any specific figure you’ve read was invented. The honest rule: include as many keywords as can truthfully share one specific ad, and no more. If you can’t write one honest ad for all of them, split the group; if a group is too narrow to ever gather meaningful data, merge it into its closest theme.

Are single keyword ad groups (SKAGs) still worth using?

Generally no. SKAGs made sense when match types were literal, but Google now matches keywords to searches by meaning and intent, so near-identical variants behave as one theme anyway. Splitting them mostly adds maintenance burden and fragments the data that automated bidding learns from. Build around themes of close intent instead — and verify current match-type behavior in Google’s documentation, since it has changed over the years.

Should brand and non-brand keywords be in separate ad groups?

Separate campaigns, ideally — not just separate ad groups. Brand searches come from people who already know you, so they convert differently and need their own budget and message. Mixing them with non-brand keywords inflates your blended results and distorts budget decisions.

Will restructuring my ad groups hurt performance?

It can, temporarily. Restructures reset the performance history attached to the old structure, and automated bidding has to re-learn, which often causes a wobble. Renovate (targeted splits and merges) before you rebuild, and when a rebuild is genuinely needed, stage it one campaign at a time with the old structure paused — not deleted — so you can roll back.