Compares search terms against your ad group keywords using fuzzy matching to flag mismatches for review or exclusion, with an optional fully-automated negation mode.
by Charles Bannister & Gabriele Benedetti (shabba.io)·shabba.io/google_ads_scripts/keywords-vs-search-terms
Reviews search terms and adds negative keywords automatically using AI-evaluated rules you configure in a companion admin panel, with email summaries of every change.
Uses Claude AI to analyze your search terms and recommend negative keywords with clear reasoning. Each term is evaluated for business relevance, intent, and conversion potential — then exported to a Google Sheet with actionable suggestions.
Sheet-driven negative keyword automation: define rules in a spreadsheet and the script adds negatives and emails alerts about irrelevant search terms without manual monitoring. Supports MCC.
Finds brand search terms leaking into non-brand campaigns (inflating their stats and misleading Smart Bidding) and generic traffic riding in brand campaigns — with the misrouted spend quantified.
Measures how much of your exact-match spend actually flows through close variants and lists the zero-conversion variants — the most surgical negative keyword candidates in the account.
Reads negative keywords and placement exclusions from a Google Sheet and syncs them to multiple campaigns via shared negative keyword and placement lists. Optionally filters to campaigns with a specific label.
For each search term, computes a fuzzy-match similarity against the keywords in the ad group that triggered it; terms below a similarity threshold are flagged as potential negatives. Runs in review mode (writes to a sheet and emails) or full-automate mode (adds negatives directly). Use it to catch queries that only loosely relate to your keywords.
| Variable | Default |
|---|---|
SPREADSHEET_URL | PASTE_YOUR_SPREADSHEET_URL_HERE |
LOOKBACK_DAYS | 30 |
FUZZY_MATCH_THRESHOLD | 70 |
FULL_AUTOMATE_MODE | false |
MIN_CLICKS | 0 |
MIN_IMPRESSIONS | 1 |
MAX_CONVERSIONS | 0 |
NEGATIVE_MATCH_TYPE | EXACT |
EMAIL_RECIPIENTS | (empty) |
CAMPAIGN_NAME_CONTAINS | [] |
CAMPAIGN_NAME_NOT_CONTAINS | [] |
Analyzed 480 search terms over 30 days. Flagged 37 low-similarity terms (threshold 70) — written to Output tab. Report emailed to your recipients (or negatives added in full-automate mode).
Bundles the author's exclusion-logic helper into a single paste. Ships with a built-in test suite (RUN_TEST_SUITE) for validating the fuzzy matcher. Co-authored by Gabriele Benedetti. Start in review mode before enabling full automation.