ADS & LANDING PAGES

Negative keyword mining from a search terms report

Use when a search terms export is full of waste and you need a structured negative list.

negative-keyword-mining-from-search-terms.md

Download .md

You are a PPC analyst cleaning up wasted spend.

Here is my search terms data as rows of: search term, impressions, clicks, cost, conversions.
{{SEARCH_TERMS_DATA}}

What I sell: {{WHAT_I_SELL}}
Who I do not want: {{DISQUALIFIED_AUDIENCE}}
Current campaign structure: {{CAMPAIGN_STRUCTURE}}

Do this:
1. Group the terms into intent buckets: buying intent, research intent, job seeking, DIY or free seeking, competitor, wrong product, wrong location, unclear.
2. For each bucket give total cost and total conversions from the data I supplied. Do not estimate numbers I did not give you.
3. Recommend negatives in a table: Negative term, Match type (exact / phrase), Level (campaign / ad group / shared list), Cost saved, Reason.
4. Sort by cost saved, highest first.
5. Separately list terms you would NOT add as negatives even though they look irrelevant, and why.
6. List up to 10 terms that look like new keyword opportunities rather than waste.

Rules:
- Only recommend a phrase negative when it cannot block a converting term already in the data. Say which term you checked it against.
- If a term is ambiguous, put it in "unclear" and say what you would need to decide.

Fill in before running

Replace each placeholder with your own detail. The more specific you are, the less the model invents.

{{SEARCH_TERMS_DATA}}
{{WHAT_I_SELL}}
{{DISQUALIFIED_AUDIENCE}}
{{CAMPAIGN_STRUCTURE}}

Getting a better result

Include at least 60 days of terms so single-click noise does not drive the list.
Check every phrase negative against your converting terms before you upload.
Keep the "do not add" list - it stops the same terms being re-flagged next month.

Questions about this prompt

When is this better than sorting the report by cost and eyeballing it?

When the export is too long to read line by line and you want waste grouped by why it is waste. Sorting by cost finds the obvious junk and misses the pattern underneath. The intent buckets, from job seeking through to wrong location, are what let you add one phrase negative instead of forty exact ones.

What do I need before running it?

A search terms export carrying cost and conversions on every row, at least sixty days of it so single-click noise does not drive the list, plus {{DISQUALIFIED_AUDIENCE}} and your campaign structure. Without the structure it cannot tell you whether a negative belongs at campaign, ad group or shared list level, and that level is most of the decision.

What does it hand back?

Intent buckets with cost and conversions totalled from your own rows, a negatives table sorted by cost saved, a list of terms it deliberately would not block, and up to ten keyword opportunities. The would not block list is the part that pays off next month, because it records why a term that looks irrelevant was kept.

What is the expensive mistake?

Uploading the phrase negatives without checking them. The prompt names the converting term it checked each one against, so audit that claim rather than the negative itself. A single phrase negative can quietly cut a query that converts, and unlike a bad bid change the damage shows up only as traffic that stops arriving.