# How to plan new Google Ads keywords with an AI agent

> Tier: deep. Full contract: quantified decision rules and evidence requirements.

New keyword candidates with historical volumes, forecast cost, and theme grouping, filtered against what the account already covers.

**Intent.** Deliver a keyword plan the account can actually absorb: net-new terms, sized, grouped, costed.

## When to use

- Entering a new market, product line or language.
- Search impression share high with budget left: growth must come from coverage.
- Quarterly planning.

## When not to use

- Accounts failing on existing keywords: expansion multiplies a problem; run account-audit first.
- Without a landing page for the new theme: keywords without destinations burn Quality Score.

## Sources

Required: Google Ads
Optional: Search Console

## Preconditions

- The seed: products, pages or themes to expand from.
- A budget envelope for the forecast to size against.

## Inputs

- `seeds` (required): Seed terms, URLs or themes.
- `market` (required): Country and language.

## Procedure

1. Generate candidates from the seeds. — tools: `google_ads_generate_keyword_ideas`
2. Attach real historical volumes and competition. — tools: `google_ads_generate_keyword_historical_metrics`
3. Forecast clicks and cost for the shortlist at the intended bids. — tools: `google_ads_generate_keyword_forecast_metrics`
4. Remove what the account already covers, from live keywords and search terms. — tools: `google_ads_get_keyword_performance`, `google_ads_get_search_terms`
5. If Search Console is connected, mark candidates where the site already ranks top 3 organically. — tools: `gsc_query_search_analytics`
6. Group the survivors into ad-group-sized themes. — tools: `google_ads_generate_ad_group_themes`

## Decision rules

- **When** A candidate duplicates an existing keyword or a converting search term → **conclude** Drop it; expansion means net-new coverage, not re-buying the same auction.
- **When** A candidate ranks top 3 organically → **conclude** Flag for the paid-organic-overlap discussion rather than auto-including; incremental value is the question.
- **When** Forecast cost of the shortlist exceeds the envelope → **conclude** Cut from the bottom of volume x relevance, and show what was cut.
- **When** A theme has volume but no plausible landing page → **conclude** Park it with the page requirement stated; do not launch keywords onto the homepage.

## Evidence requirements

- Every kept candidate carries volume, competition, forecast cost and its theme.
- The deduplication states how many candidates were removed and why.

## Human approval boundary

- The plan is delivered; creating campaigns or ad groups from it is human work, and any campaign the gateway creates via google_ads_create_campaign is created paused, through Safe Writes.

## Failure modes

- Planner volumes are ranges and averages; treat them as sizing, not truth.
- Forecasts assume the bid landscape holds; a competitor entering mid-quarter voids them silently.
- Language mixing in one market splits volumes across variants; group per language before judging size.

## Output

- **The plan**: Themes, keywords, volumes, forecast cost, landing pages.
- **Removed**: Duplicates and covered terms, so the net-new claim is checkable.
- **Parked**: Themes waiting on pages or budget.

## Related

- Skill: [paid-organic-overlap](https://www.getmcpads.com/skills/paid-organic-overlap.md)
- Skill: [google-search-waste](https://www.getmcpads.com/skills/google-search-waste.md)
- Guide: [wasted-search-spend](https://www.getmcpads.com/guides/wasted-search-spend.md)