Check negative keyword conflicts before blocking useful searches

Match behaviour and scope both matter

Start with the exact candidate, match type and scope. The same text at ad-group, campaign or shared-list level can affect different services. Record where the exclusion would apply and which other campaigns use that list. A local problem should not become an account-wide rule merely because a shared list is convenient. Scope is part of the hypothesis, not an administrative detail to add later.

Google explains that negative broad match blocks searches containing all its terms, negative phrase requires the specified sequence and negative exact targets the complete query. Close variants are not handled like positive keywords. Use those definitions to test a candidate, and verify current documentation for the campaign type involved. Do not replace the matching rules with an AI guess that two expressions probably mean the same thing.

Educational example: a business rents tools and also repairs customers' tools. Someone proposes the single-word negative 'repair' for the rental campaign after unrelated searches. The observed query 'rent drill while mine is in repair' may be valuable even though it contains that word. Applying the exclusion to a list shared with the repair campaign would reach still further. An exact exclusion of one confirmed irrelevant query and a broad exclusion of a service concept are different decisions.

Create two query sets: observed useful searches and plausible useful searches the team wants to preserve. The first has historical evidence; the second expresses the intended offer and must be labelled hypothetical. Check the candidate against both. A conflict with a plausible query does not prove a financial loss, but it is a reason to narrow the candidate or request stronger evidence before proceeding.

The three rows of a conflict packet

• A conflict row contains the proposed negative, intended scope, useful query, predicted blocking result and the business reason for preserving that query.
• An evidence row contains the original search's spend, mature outcomes and relevance judgement. It prevents 'zero conversions yesterday' from becoming the sole justification.
• A decision row records retain, narrow, investigate or approve for the authorised process. It also states who will verify the result and when the review expires.

Review the exclusion without executing it

1. Export or otherwise inspect the current exclusions and their inheritance. Confirm you are reviewing the actual state, rather than an old list supplied during a handoff.
2. Collect the candidate queries and enough surrounding context to understand the requested task. Keep useful converting searches in the sample as well as waste candidates.
3. Prepare a preserve set with sales or product staff. Mark observed queries separately from invented examples, including relevant mixed-intent searches.
4. Apply the documented match behaviour on paper. List each predicted conflict and the scope that would be affected; do not submit a write to test your understanding.
5. Compare narrower alternatives. An exact query exclusion, a different campaign scope or an offer change may solve the problem with less collateral restriction.
6. Attach the final evidence to the decision record. After any separately authorised implementation, check the resulting configuration and later outcome mix, including demand that could have disappeared.

Passing the check is not proof of zero risk

Sources and further reading

• Google Ads: search terms report coverage
• Google Ads: negative keyword match types