What you need to know
- Give the assistant business context before asking it to judge query relevance.
- Separate irrelevant queries from relevant queries that have not converted yet.
- Review each proposed negative’s text, match type and scope before applying it.
Start with a defined report and business brief
A search-term review asks which reported queries deserve attention. It is narrower than a full account audit: you are not simultaneously rewriting campaigns, diagnosing every conversion issue and replacing the bidding strategy.
Google distinguishes the query a person searched from the keyword that matched it. Preserve both fields alongside campaign, ad group, clicks, cost, conversions and conversion value where available. The report may not expose every query, so avoid presenting the visible rows as the entire search landscape.
Choose a date range and identify the account currency. With PlugMind, request the reporting period and collection limits explicitly. If a result is capped, narrow the scope or obtain a fuller export before claiming the review is exhaustive.
Explain what counts as a qualified search
Describe the product, service area, intended customer, excluded offers and important exceptions. An assistant cannot infer whether “cheap” is bad for your business or whether a competitor name is an intentional acquisition target.
| Business fact | How it changes the review |
|---|---|
| You sell subscriptions, not equipment | Equipment-repair queries may be irrelevant. |
| You serve only specified locations | Out-of-area searches need a geography check. |
| You offer a free consultation | Do not blanket-exclude the word free. |
| A sale usually takes several weeks | Recent non-converting queries may still be undecided. |
These are illustrative examples, not universal negative-keyword rules. Ask the assistant to explain its reasoning using your brief and the actual query rather than producing a large unexplained exclusion list.
Classify queries before recommending negatives
Keep a separate list of promising queries. A useful review can discover language to test in a keyword or landing-page proposal, not just terms to block. Preserve an uncertain category when intent is ambiguous or the evidence is thin.
For example, an expensive query with zero conversions could reflect poor relevance, delayed conversions, a weak landing page or a measurement problem. The number alone does not tell you which explanation is correct.
Review exclusions and record the decision
Google’s negative-keyword behavior differs from positive matching. Review the selected match type and the intended ad-group or campaign scope, and check that the proposal will not exclude searches you still want.
Google’s interface and PlugMind’s enabled write tools do not necessarily expose identical scopes. Apply only a supported, explicitly reviewed proposal through PlugMind; use the platform interface when the requested scope is not available.
- Keep the original query and reason for every approved exclusion.
- Record the date and account scope so later reports can be interpreted correctly.
- Review uncertain suggestions manually rather than approving the whole list.
- After a change, check subsequent relevant traffic and lead quality.
Frequently asked questions
Can ChatGPT find every bad keyword automatically?
No. It can help classify reported queries, but relevance depends on your offer and intent. Missing report rows and incorrect business assumptions limit the result.
Should every query with zero conversions become a negative?
No. Consider relevance, sample size, conversion delay and measurement before excluding a query.
Do I need a live connection?
You can review an authorized export or use a supported live connection. Exports are snapshots; connected reports still need explicit account and date scope.
Sources and verification
Platform references are listed below. PlugMind tool descriptions were checked against its implementation. Example prompts are suggested workflows, not claims of results from a live advertising account.