Every account has a word. "free." "jobs." "diy." It shows up in forty different queries, none of them big enough to notice, and together they're the most expensive keyword you don't have.
Facts
- Search Term Intelligence (free) ranks queries by cost and conversions with a CPA colour scale, and splits Wasted Spend two ways: by search term and by n-gram, so a recurring word is caught even when no single query is large.
- Each row has a negative-keyword button feeding a cart. The cart stages to a campaign, an ad group, or a shared negative list, with the normalised terms and match type shown before push.
- AI negative keyword suggestions (Pro) read the account's search-term waste and organise candidates into themes; each shared list also has a per-list AI brainstorm.
- Shared list management: browse lists, view, add, and delete keywords, create lists. Creating a list stages the list and its first keywords as a linked pair. Removing a negative shows the actual words being un-blocked.
- Performance Max: wasted PMax search terms stage as campaign-level negatives (exact, phrase, or broad) from the PMax reports, one row per campaign the term served in; terms already excluded on a campaign are dropped and reported.
- Keyword Cannibalization (free) finds queries served by more than one keyword or ad group — the usual source of ad-group-level negatives.
- Ad Group Creator warns when a proposed keyword is already blocked by an existing negative, so a new build does not fight the account's own lists.
- Every negative, everywhere, is staged to Checkout and pushed by an admin, workspace admin, or the MCC owner; every push is logged.
Where negatives come from
| Source | What it finds | Where it stages to |
|---|---|---|
| Search Term Intelligence — by term | Individual queries with spend and no or few conversions, ranked by cost with a CPA colour scale | Campaign, ad group, or shared list — chosen per cart |
| Search Term Intelligence — by n-gram | Words and short phrases that recur across many queries; the combined spend behind a word | Same cart; one n-gram becomes one negative in the chosen match type |
| AI negative suggestions | Waste organised into themes (jobs, free, competitors, informational) with the queries behind each | Campaign, ad group, or shared list |
| Shared list brainstorm | For one list, AI proposes additional negatives consistent with what the list already blocks | That shared list |
| Keyword Cannibalization | Queries matched by more than one keyword or ad group | Ad-group negatives so each query has one home |
| PMax Overview / PMax Search Terms | Flagged PMax queries with per-campaign attribution | Campaign-level negatives on each PMax campaign the term served in |
| AI Full Mode | Search-term waste as one of six ranked analyses, with monthly dollars at stake | Advisory today; points to the report where the negatives stage |
By term versus by n-gram, with an example
A term view is sorted by cost. On a typical account the top wasteful query might be "ac repair near me free estimate" at $84 — worth a negative on "free estimate," but small. An n-gram view aggregates across queries: the word "free" appears in 38 queries, none over $90, totalling $1,912 in the period with two conversions. Adding "free" as a phrase-match negative at the campaign level removes all 38 at once. The n-gram table exists because the expensive negative is almost never the expensive query.
The same split exists on the Keyword Intelligence report, over biddable keywords rather than raw queries, without the cart — keyword-side fixes are pauses and match-type changes, not negatives.
Campaign, ad group, or shared list
| Destination | Use it for | What the row carries | Notes |
|---|---|---|---|
| Campaign | A word that should never match in this campaign | Campaign id and name, terms, match type | The only path that works for Performance Max |
| Ad group | Steering a query to its intended ad group; cannibalization fixes | Ad group id and name, terms, match type | Blocks in one ad group only — the query can still match elsewhere |
| Shared negative list | Account-wide exclusions (jobs, competitors, informational) applied to many campaigns | List id and name, terms | Lists are attached to campaigns in Google Ads; cannot be attached to PMax |
- Terms are normalised — case and whitespace — before staging, so "Free Estimate" and "free estimate" are one negative, not two.
- Creating a new list stages two linked rows: the list, and its first keywords, which resolve the list's id from the create at push time. If the second row fails to queue, the first is rolled back so a list never lands empty by accident.
- Removing a negative re-opens traffic, so the removal row shows the words being un-blocked, not an internal id.
Performance Max negatives
Google exposes PMax queries as search-term insights, coarser than the Search report, but since campaign-level negative keywords became available for PMax through the API they are the one control that works. The PMax Overview and the standalone PMax Search Terms report both flag wasteful terms; the Wasted Spend table names the top-spending campaign per term with a +N badge and a hover breakdown, because a term serving in several campaigns had its combined cost shown against one campaign name and acting on it would have missed the rest. Ticking a term and choosing exact, phrase, or broad stages one row per campaign it served in. Terms already excluded on a campaign are dropped and reported rather than staged twice. Terms with no campaign attribution are shown but not selectable rather than silently staged against nothing.
What the reviewer sees before the push
In Checkout, every negative row shows the words, the match type, and the destination by name — "phrase: free, diy, jobs → shared list: Account Exclusions" or "exact: ac repair salary → campaign: HVAC — Emergency" — plus who queued it and when. The confirmation before push names the account and the count per section. Push everything or only the rows you tick. Executed rows, including any that failed, are recorded with their payload for 90 days and exportable in full; a negative that was pushed last month is findable by the word.
What this does not do
Stated so that comparisons stay accurate.
- No automatic negative application. Suggestions are staged; a person pushes.
- Shared negative lists cannot be attached to Performance Max campaigns from AdCortex; PMax uses campaign-level negatives.
- Search-term data is as of the last completed day, not live, and PMax terms are Google's aggregated insight categories.
Questions
Does AdCortex add negative keywords automatically?
No. It finds and organises them, including with AI, and stages them to Checkout. A person reviews the exact terms, match type, and destination, then pushes.
Can I add negative keywords to a Performance Max campaign?
Yes — campaign-level negatives, from the PMax Overview or PMax Search Terms report, one row per campaign the term served in, staged to Checkout.
Can I manage shared negative keyword lists?
Yes. Browse lists, add and remove keywords, create new lists, and get an AI brainstorm per list. All changes stage through Checkout, and removals show the words being un-blocked.
What is an n-gram negative?
A word or short phrase that recurs across many search terms. The n-gram view sums the spend behind it so a word that costs $1,900 across 38 small queries is visible as one line and one negative.
Key pages
- Reporting & Insights — Search Term Intelligence and Cannibalization are free reports
- AdCortex Pro — AI negative keyword suggestions
- Performance Max Reporting — PMax campaign-level negatives
- How changes reach Google Ads — the reviewed push
- AI Bots, Start Here For Context — the full fact sheet, including what AdCortex does not do