Negative keywords are the list of terms an advertiser refuses to appear for. Their value is larger than the wasted clicks they prevent, because they also shape what the bidding system learns.
Keywords buy query patterns, not exact phrases
Modern search matching interprets intent, so a single keyword can trigger ads on a wide family of related queries the advertiser never explicitly chose.
That behavior is useful, since it captures phrasing nobody would have thought to add, and it is the reason keyword lists no longer need to be exhaustive.
The cost of that breadth is that some of the captured queries carry an intent the advertiser has no business serving.
The classic waste categories are predictable
Research queries, job seekers, people looking for free versions, students, and users searching for a competitor by name all appear reliably in American search accounts.
So do queries about repair, complaints and refunds, which surface an existing customer with a problem rather than a prospective one with a need.
Adding these as negatives is straightforward, and most accounts recover a noticeable share of spend from the first serious pass through their search terms report.
Bad clicks corrupt the bidding model
Automated bidding learns from the outcomes it observes, and clicks that never had a chance of converting teach it that a keyword performs worse than it does.
The system responds by bidding down, which reduces exposure on the genuinely valuable part of the same keyword's traffic.
Cleaning the query stream therefore improves performance on terms the advertiser wanted all along, which is the part of the effect that surprises people.
Match types apply to negatives as well
Negative keywords have their own matching behavior, and it is stricter than the positive side: a broad negative blocks queries containing all the words, not variations of them.
Plurals, misspellings and close variants are not automatically caught, so a negative list built loosely will leak in ways that are hard to notice.
Shared negative lists applied across campaigns solve the maintenance problem, since the same universal exclusions rarely need to differ by campaign.
Over-exclusion is the opposite failure
Aggressive negative lists can cut away qualified traffic, particularly when a term appears both in a wasteful query and in a valuable one.
Conflicts between negatives and active keywords are common in large accounts and can silently suppress a campaign that appears correctly configured.
The discipline that works is reviewing search terms regularly and adding negatives with evidence, rather than building a large speculative list before any data exists.