Most wasted ad spend is decided before the campaign launches — wrong keywords, overlapping ad groups, no negatives, no audience plan. This is the exact pipeline we run to make sure that never happens.
No guessed keywords, no AI hallucinations — every stage works from live Google Ads data and validates its output before it can reach your account.
We pull live data from the Google Ads API for your niche and location: monthly search volume, competition, top-of-page bid ranges, 12-month trend direction, and even the peak months for seasonal services.
Duplicate phrasings of the same query are collapsed by token signature ("plumber dubai" = "dubai plumber"), then keywords are grouped into candidate themes by volume tier. Works in any language — Arabic and Urdu keywords included.
Every keyword is classified — transactional, commercial, informational, or question — and flagged for local intent ("near me"), voice search, and urgency ("emergency", "same day"). Buying-intent keywords are boosted; question keywords are routed to SEO content instead of ads.
Keywords are organised into 1–7 tightly-themed ad groups — the count comes from the data, never padded. Each group gets negative-keyword siloing: the other groups' core terms as negatives, so two of your own ads can never compete for the same search.
A ready-to-paste Google Ads Script that runs hourly in your account and auto-negatives junk search terms — jobs, DIY, courses, wrong cities, wrong trades — using a 10-rule engine built from your own keyword data, not a generic blocklist.
From a master list of 3,500+ real Google Ads segments (In-Market, Affinity, Topics), AI picks who to observe, who to bid up, and who to exclude — job seekers, students, DIYers. Every segment name is validated against the master list, so a hallucinated audience can never reach your account.
Not every keyword with volume deserves a bid. The scoring stage weights each keyword by what the searcher is actually trying to do.
| Intent | Weight | Where it goes |
|---|---|---|
| Transactional — "hire", "book", "price" | ×1.25 | Ads — top priority |
| Commercial — "best", service + city | ×1.10 | Ads |
| Informational — "how to", "guide" | ×0.60 | Blog / SEO content |
| Question — "why does my…" | ×0.50 | FAQ / problem-solution pages |
On top of intent, each keyword carries flags that shape the campaign:
Hand-written negative lists break accounts — they ban "price" for a business with a Pricing ad group. Ours is built from your own keyword data, so it can't.
Any search term that has ever converted is protected — the script can never negative it, no matter what other rules say.
AI proposes the niche junk list (other trades, DIY, jobs, courses) — then code removes anything that collides with your actual bid keywords. The #1 false-positive source in hand-written scripts, eliminated.
Whole-token fuzzy matching keeps misspelled buyers ("plamber", "carpanter") and never does substring matching — so "place" can never accidentally block "replacement".
Education, jobs, and DIY queries are caught in multiple languages — including Roman-Urdu and Hindi transliterations common in GCC markets.
Negatives are added as exact-match [terms], never broad phrase bans. Repeat offenders are logged as phrase-negative suggestions for human review.
The script starts in DRY_RUN mode — it logs everything it would block for your review before it's allowed to touch the account.
Want the script itself? It's part of our free Google Ads Scripts library — or use the AI Search Term Cleaner on your existing search terms report right now.
The blueprint is not a report that sits in a drawer — every output feeds the next part of the system.
Each ad group theme becomes one AI landing page with a dynamic headline matched to that group's keywords — perfect message match, higher Quality Score.
Ad groups, keywords, negatives, and the full audience plan export as Google Ads Editor import files — a whole account structure applied in minutes, no manual entry.
The informational and question keywords excluded from ads become the blog and problem-solution page plan — so the research pays for organic traffic too.