Search Term Harvester

Scans the search terms report for high-converting queries and adds them as exact match keywords in the matching ad group. Runs in dry-run mode by default so you can review before applying.

keywords

search terms

negatives

Overview

Search Term Harvester scans the last 30 days of search query data across all enabled campaigns and surfaces queries that have driven at least the minimum number of conversions. In dry-run mode it logs candidates so you can review; once you're confident, switch ADD_KEYWORDS to true and re-run to write the new exact-match keywords directly into the matching ad groups.

Setup

1Create a new script in Google Ads Scripts (Tools → Bulk Actions → Scripts → +).
2Paste the code and run it once in preview mode to review candidate keywords in the logs.
3Tune MIN_CONVERSIONS to a threshold that makes sense for your account volume.
4Optionally set MAX_CPA to skip high-CPA queries (0 = ignore CPA).
5When satisfied with the candidates, set ADD_KEYWORDS = true and run again to apply.

Ready to Use This Script?

Configuration Variables

Variable	Default	Description
`ADD_KEYWORDS`	false	Set to true to write exact-match keywords. false = dry-run (log only).
`MIN_CONVERSIONS`	1	Minimum conversions a search term must have to be considered.
`MAX_CPA`	0	Maximum CPA allowed for a term. Set to 0 to disable the CPA filter.
`DATE_RANGE`	LAST_30_DAYS	Google Ads date range string to pull search term data from.
`SIMILARITY_THRESHOLD`	0.85	Skip a candidate if it is at least this similar (0-1) to an existing keyword in the same ad group. 1.0 = skip only exact duplicates.

Expected Output

=== Search Term Harvester ===
2 search term(s) meet the conversion threshold.
CANDIDATE: [winter running shoes]  Conv: 3  CPA: $18.50  AdGroup: Running Footwear
SKIP (similar): [trail runner]  AdGroup: Trail Running

Summary: 1 candidate(s)  |  1 skipped
Set ADD_KEYWORDS = true to apply changes.

Notes

Before adding a term the script compares it against every existing keyword in the target ad group using a Levenshtein-distance similarity ratio. Any candidate scoring at or above SIMILARITY_THRESHOLD is skipped, so near-duplicates (e.g. singular/plural or minor spelling variants) are not added. Comparison is case-insensitive and ignores match-type punctuation.