Negation word rule

The word rule solves a problem the ordinary negation rule cannot: a single word that is dragging down dozens of different search terms, none of which has enough traffic on its own to be worth negating.

It is a mode, not a separate rule

You reach it from the negation rule editor. Negative mode at the top of the dialog decides which of the two you are configuring:

Individual words is selected. Everything below the toggle is the same editor as an ordinary negation rule.

Negative mode	What it negates
Search term	The full search term, when it meets the Criteria
Individual words	The search term is split, and a Negative Phrase is created for each word that meets the Criteria

What it does

It looks for a word that appears across several distinct search terms, checks the combined evidence, and creates a Negative Phrase for that word. The dialog says so at the foot of the Trigger section: create a Negative Phrase for each individual word that meets the Criteria and safeguards.

This rule creates phrase negatives exclusively. It is designed to block a broader search pattern rather than a single query, including related queries that have not yet appeared in the data.

Configuration

Setting	Default	What it does
Performance Criteria	—	Optional. Metrics are aggregated across the distinct search terms containing the candidate word
Minimum search term count	3	How many distinct terms must contain the word before it counts as a pattern. Range 2–100
Search term scope	Individual ASIN	Whether to aggregate evidence at ASIN or Product Group level
365-day orders safeguard	On	Skip words whose terms have converted in the last year

Minimum search term count is the whole rule

Set it to 2 and single coincidences start looking like patterns. Set it to 20 and you will only ever catch words so common you would have found them yourself. Three is a sensible default; raise it on a high-volume account.

The negative still lands locally

Widening the evidence does not widen the action. However many terms the word was found across, and whichever scope the evidence was read at, the Negative Phrase is created in the assigned ad groups where the word actually appeared — never at campaign level, and never in an ad group the rule is not assigned to.

Safeguards

The same protections apply as on a negation rule: the 365-day orders safeguard is on by default, and a term already covered by an existing negative is skipped rather than negated twice.

Use cases

Block a competitor’s brand across the board

Their brand name shows up inside forty different queries, each with two or three clicks. No single query justifies a negative; together they are a real amount of spend. One Negative Phrase ends it.

Kill a wrong-intent modifier

free, used, refurbished, for kids — words that reliably describe a shopper who will not buy your product at your price. They spread across many queries and never concentrate anywhere.

Cut off a compatible-product pattern

Shoppers searching for something your product fits but is not. The pattern lives in a word, so that is where the block belongs.

A worked example

Across a 90-day window, 11 distinct search terms contain the word refurbished. Individually they have 1–4 clicks each and no orders. Together: 26 clicks, $18.40 spent, nothing sold.

With minimum search term count = 3, the rule sees a pattern rather than eleven separate non-events, and creates one Negative Phrase: refurbished. The 365-day safeguard passes it through because none of those terms has produced an order in the last year.

Figures are an illustration, not measured performance.

Considerations

• A phrase negative is wide by design. Check the word is not also inside queries you want — case blocks phone case as well as case of 12.
• Words that are part of your own product name will show up in losing queries too. Blacklisting one of those is a slow, quiet way to remove your best traffic.
• Because the evidence is aggregated across terms, the scope setting matters as much here as anywhere: at Product Group scope a word is judged on the whole group’s queries.