Negative keyword lists are treated like a setup task. Someone builds one in week one, feels productive, and never opens it again. Six months later the account is paying for job seekers, students, and people looking for free templates, and the report that would show this has not been opened since launch. Negatives are not a task. They are a habit, and the habit is what separates accounts that get cheaper over time from accounts that get more expensive.
The query mix underneath your keywords is not stable. It moves because search behavior moves, because your competitors change their bidding, because Google keeps widening how loosely a keyword can be matched, and because you added a new service page that pulled in a new neighborhood of intent.
Broad match and automated query expansion accelerate all of this. The set of searches you can be shown for today is not the set you were shown for in January. A list built once is a snapshot of a market that no longer exists.
Most accounts I audit have one messy list applied everywhere, or negatives scattered at ad group level with no logic. Use three layers instead.
Match types matter here and people get them backwards. Broad negative blocks any query containing all those words in any order, and it is the blunt instrument that causes accidents. Phrase negative blocks the words in order. Exact negative blocks only that precise query. When in doubt, go narrower. A negative you regret is invisible, because the traffic it blocked never shows up in any report.
The reason negative lists rot is that the recommended process is too heavy for a business owner with a real job. Here is a version that survives a busy month.
The weekly ten minutes is the one that matters. If you only do one, do that one. Consistency beats depth here by a wide margin.
Negatives are not free. Every one of them narrows the space the bidding system can explore, and I have seen accounts strangled by over-enthusiastic blocking.
The common failures: adding a broad negative for a single word that also appears in your best converting queries. Blocking a whole city because one suburb performed badly. Negativing terms after two weeks and forty clicks, which is not enough data to conclude anything. Copying a competitor negative list wholesale, which imports someone else business logic into yours.
There is also a real argument against aggressive negatives in low-volume accounts. If you get a handful of conversions a month, Smart Bidding is already short of signal. Cutting the query space further can push a campaign into a state where it cannot learn at all. In those accounts I stay tight on obvious junk and leave the ambiguous middle alone.
Put a recurring ten minute slot in your calendar, same day every week, titled search terms. Open the report, sort by cost, read the top thirty, add what is obviously wrong, and close it. That is the entire discipline. It is unglamorous and it compounds. Accounts that do it get quietly cheaper every quarter while accounts that do not pay a slowly rising tax on their own inattention.
One more thing. Before you add a single negative, look at your conversion tracking. Blocking traffic based on a conversion count that does not reflect real inquiries will remove the wrong terms with total confidence.