Negative keywords are a maintenance habit, not a setup task
Negative keywords tend to be treated as configuration: build the list at launch, tick the box, move on. Then six months later a search terms report shows a third of spend going to phrases nobody would have approved.
The list decays because the thing it filters — how people phrase searches, and how Google interprets them — does not hold still.
The three tiers worth keeping
Account-level negatives. Things you never want, anywhere: "free", "jobs", "salary", "how to", competitor brand names if that is your policy, and locations you do not serve. Small, stable, rarely changed.
Campaign-level negatives. Intent that belongs elsewhere. If you run separate campaigns for repair and installation, each is a negative for the other. This is the tier that prevents campaigns cannibalising each other.
Ad group-level negatives. Sculpting within a theme. Use sparingly — over-sculpting fragments data and modern matching makes it less effective than it used to be.
The weekly routine
Fifteen minutes, sorted by cost:
- Open search terms for the last 7–14 days.
- Sort by cost descending.
- Read the top fifty. For each: would I pay for this again?
- Negatives for the ones you would not, at the right level.
- New keywords for the ones converting that are not already in the account.
Step five is the one people skip, and it is where the value is. Negatives stop waste; promoting proven search terms into their own keywords is what grows the account.
Match types on negatives
Negatives use their own matching rules and they are stricter than positive matching. A broad match negative does not expand to synonyms — it blocks queries containing all the words, in any order. It does not block plurals or misspellings automatically.
That last point causes most negative keyword surprises: blocking "job" does not block "jobs".
The list that is too aggressive
It is possible to over-negative. Signs: impression share falling while position holds, or a campaign that stops finding new converting terms entirely.
A negative list should remove waste, not narrow the account to only what you already thought of. If it has been years since a genuinely new converting search term appeared, the filter is doing too much.