What you'll learn in this article
- Why the keyword count is the wrong thing to worry about, and what actually breaks
- How dilution works: one ad, one landing page, and five different intents pulling at them
- What Quality Score is really reacting to when the whole column drops at once
- Query overlap, and why you stop choosing which of your own ad groups represents you
- The restructure I ran on a bloated account, step by step, and the test I apply before adding anything now
The question is adding too many keywords bad for Google Ads usually arrives phrased as a number. Someone wants to know if forty in an ad group is too many, or whether they should stay under twenty. I understand the instinct, but in eight years of managing accounts I have never once diagnosed a problem by counting keywords. I have diagnosed plenty by looking at what those keywords had in common.
The account that taught me this had one ad group with sixty-three keywords in it. It was not failing because of the sixty-three. It was failing because those sixty-three keywords described at least five different things a person might want, and they were all being answered by the same two ads pointing at the same page. The count was a symptom. The spread of intent was the disease.
So this is not a post about the right number. It is about what goes wrong when an ad group holds more intent than it can serve, and the restructure that pulled that account back. Three things break, in a specific order: the ad dilutes, the scores follow, and then your own ad groups start overlapping on the same searches.
1. Dilution: one ad, one landing page, and five intents pulling at them
An ad group is not a folder. It is a promise that everything inside it deserves the same answer. Every keyword in the group triggers the same ads and sends people to the same destination, which means the group can only be as specific as its most divergent member allows.
The sixty-three-keyword group I mentioned was for a B2B software client, nominally about project management software. Inside it were terms for free trials, terms for enterprise pricing, terms comparing them against a named competitor, terms for a specific integration, and a cluster of terms from people looking for templates rather than software at all. Five intents. One headline.
Watch what that does to copywriting. If I write "Start Your Free 14-Day Trial" I have written the best possible ad for one cluster and an irrelevant ad for the pricing and comparison searches. If I write "Project Management Software for Teams" I have written something inoffensive to all five and compelling to none. Under pressure, everyone writes the second one. That is the dilution: not that the keywords are numerous, but that the ad has been negotiated down to the lowest common denominator of everything they mean.
The landing page suffers the same negotiation. You cannot send a pricing search and a template search to a page that serves both without making that page a hub, and a hub converts worse than a page answering the question directly.
What made it hard to see is that nothing looked broken. Impressions were healthy, the campaign was spending. Only when I segmented the search terms by what the person actually wanted did the group stop looking like one audience and start looking like five audiences sharing an apartment.
2. Quality Score does not punish volume, it punishes the compromise volume forces
This is where most explanations of too many keywords in Google Ads go slightly wrong. They imply Google penalises you for having a lot of keywords. It does not. Quality Score is calculated per keyword, and a keyword with fifty neighbours is scored exactly the same way as a keyword sitting alone.
The damage is second-order, and being precise about it is what tells you where to intervene. Two of the three components feeding the score are expected click-through rate and ad relevance, both judgments about the fit between a keyword and the ad it triggers. When the ad has been diluted to cover five intents, that fit is worse for all five. The scores then decline as a block rather than one keyword at a time, because they share the same compromised ad.
That block behaviour is the diagnostic signal I now look for first. If I sort keywords by Quality Score and the low numbers cluster inside one ad group rather than scattering across the account, I stop looking at the keywords and start looking at the container. A whole group at four or five is almost never fifty independent keyword problems. It is one structural problem wearing fifty hats, and the same reasoning applies to the individual terms I unpack in my breakdown of what low keyword quality is really telling you.
The financial consequence is not subtle. Ad Rank folds quality into the auction, so a depressed score means paying more for the same position or losing it entirely. That group was paying roughly a third more per click than a comparable group elsewhere in the same account, for traffic that converted worse. Nobody had done anything wrong. They had just kept adding.
3. Query overlap: when you stop choosing which of your own ad groups answers a search
The third failure is the one advertisers underestimate most, partly because the interface never shows it directly. Once several ad groups hold keywords broad enough to cover neighbouring intents, the same search term becomes eligible to match in more than one of them.
A common misconception is that this means your own keywords bid against each other and inflate your costs. They do not. Google resolves the conflict before the auction, and the resolution logic is documented in detail in Google's help page on ad group prioritisation. An exact match keyword identical to the search term wins first. Failing that, phrase and broad keywords identical to the term take precedence. If neither applies, Google uses AI-based prioritisation that weighs the meaning of the search against all the keywords in an ad group and the landing pages inside it, and only then does Ad Rank break the remaining tie.
Read that third criterion again, because it is the one that punishes bloated groups specifically. Relevance is assessed against all the keywords in the group, not just the one that matched. So when I stuffed template searches and integration searches into a group nominally about software, I did not just dilute the ad. I diluted the signal Google uses to decide whether that group is the right home for a borderline query at all. The group became a less confident answer to every search, including the ones it was originally built for.
The practical result is loss of control. You built a comparison ad group with comparison copy and a comparison landing page, then a competitor search matched into your generic group instead because that group's Ad Rank was higher. Reporting credits the generic group, you conclude it is your workhorse, and you feed it budget on the strength of an internal selection you never made. The match-type dimension matters here too, something I cover in my guide to how the keyword match types actually behave, since the same text under two match types sits at two different points in that priority order.
4. How I actually restructured it
I did it wrong the first time. My instinct was to split by theme using the keyword text as the guide, which produced tidy groups that did not match how people were searching. Keyword text tells you what I thought people wanted; the search terms report tells you what they typed.
Start from search terms, not from keywords
The second attempt started with ninety days of search terms filtered to that one group. I ignored the keywords entirely and clustered the queries by what the person appeared to want: evaluate, compare, price, implement, or something that was not us at all. That last bucket was large, and most of it became negatives rather than a new group.
Rebuild containers around the clusters that earn one
Four clusters had enough volume to justify their own ad group, meaning enough traffic to accumulate judgeable data within a month. Everything below that threshold stayed together deliberately, because a group too small to produce signal is a different problem, not a solution to this one. Each new group got ads written to that single intent and, where possible, its own landing page.
Move keywords last, and prune while you move
Only then did I place the original sixty-three. Around fifteen went into the cluster they belonged to. Eight were duplicates of terms living in other campaigns, which I only caught by comparing against a full account export rather than working inside the one group. The rest had spent ninety days without a conversion, and those I paused rather than rehoused.
Over the next two months: click-through rate up meaningfully on three of the four new groups, Quality Score across the migrated keywords moving from a cluster around four to a cluster around seven, and cost per conversion down by roughly a quarter. I would not credit the restructure alone, since better ads were written in the same pass, but the direction was unambiguous and it held. The recurring version of this is a fixture in my weekly optimization routine rather than a one-off cleanup.
5. The test I apply before adding a keyword now
I stopped using a number years ago. What I use instead is a single question asked before any keyword enters a group: can I write one headline that is genuinely the best possible answer to every keyword already in here, plus this one?
If yes, it goes in and the count is irrelevant. I have groups holding thirty keywords that pass easily, because those thirty are phrasing variants of one identical want. If no, adding it means rewriting the ad to be vaguer, trading away performance on existing keywords to gain coverage on one new one. That trade is rarely worth it and is invisible at the moment you make it, which is why bloated groups accumulate rather than get created.
There is a second question I ask when the answer to the first is no: does this keyword deserve a new container, or does it belong in one that already exists elsewhere in the account? On mature accounts the answer is usually the latter, which is why I check against a full keyword export rather than trusting my memory of the structure. Adding a term that already lives two campaigns away is how query overlap gets built one reasonable decision at a time, and the mechanics of finding those collisions are specific enough that I have written them up in my process for tracking down duplicate keywords.
The honest verdict
So: is adding too many keywords bad for Google Ads? The count is not what harms you. What harms you is intent spread inside a container that can only give one answer, and count is simply the most visible correlate of that spread. An ad group with forty tight variants of a single want is healthy. An ad group with nine keywords covering four different jobs is already broken, and nobody will flag it for you because nine sounds reasonable.
The three failures compound in sequence, and that sequence tells you how far along you are. First the ad goes generic to cover everything. Then quality metrics for the whole group sag together, because they inherit that generic ad. Then your groups collide on the same searches and Google, not you, decides which one represents your business.
If you inherit an account and want the fastest read on whether this has happened, do not count anything. Sort keywords by Quality Score, look for low numbers clustering inside single ad groups, then open the search terms for one of those groups and ask how many different things people were trying to do. If the answer is more than one, you have found your restructure.