Amazon Negative Keywords: How to Find and Apply Them

Last updated October 23, 2026 · First published November 5, 2026 · By Chad Norris

Disclosure: the author runs an Amazon advertising agency (Tide Flow) and builds Amazon software (AdSpark), which can automate parts of this workflow. The method below works in a spreadsheet and the Amazon console. See the disclosure.

A negative keyword tells Amazon not to show your ad for a query, a phrase, a product or a brand. The short answer on method: pull the search term report, mark every term that has spent more than a sale is worth with no orders, add confirmed irrelevant queries as negative exact, add never-right words as negative phrase, and review again in two weeks. The mistakes are almost all about doing this too fast or too broadly.

What Amazon says negatives do

Amazon’s targeting guide (read October 2026) describes negative targeting as a way to exclude keywords, products or brands you do not want your ads associated with, in both automatic and manual Sponsored Products campaigns. It names two keyword types:

The same guide advises evaluating a negative once it has received at least 20 clicks and letting changes run two weeks or more. Amazon’s keyword targeting guide says negatives can be applied at campaign or ad group level and mentions other match forms; we could not confirm every label from the pages we could read, so check which options your console shows before building a list around a match type it does not offer.

Step 1: Pull the right report

The search term report lists the shopping terms that produced at least one click (per Amazon’s best-practices guide). Pull at least 30 days, longer for slow-moving products, and export it rather than reading it in the console. Keep these columns: campaign, ad group, targeting, match type, customer search term, impressions, clicks, spend, orders and sales.

Use mature dates. Orders are credited to the click that preceded them within the attribution window, and recent days under-report sales. Third-party documentation puts the Sponsored Products seller window at 7 days; we could not read Amazon’s own statement of it, so confirm for your account. In practice, leave the most recent week out of any decision to negate.

Step 2: Set the threshold from your own margin

A fixed rule such as “10 clicks and no sale” ignores the product. Use the break-even work from the ACoS guide and set a spend threshold instead.

Allowable ad cost per sale = price x target ACoS. A term that has spent more than that with zero orders has cost you more than one sale was ever allowed to cost.

Product B shows why a click count fails: at a $0.80 click price, two clicks already exceed what the item can afford, while Product A can take over a dozen. Use a multiple of the allowable cost (1.5x to 2x) as the trigger, because conversion is noisy: a 10% converter will go ten or more clicks without a sale about one time in three. That figure is simple probability, (0.9)^10 is roughly 0.35, not an Amazon statistic, and it is the reason a premature negative can cut a term that would have sold.

Step 3: Sort flagged terms into four buckets

The fourth bucket is where most waste is actually mismanaged. A relevant term with no orders is a listing, price or bid problem, not a query problem. Negating it removes the evidence. The rule from our own method notes is to negate for irrelevance, and to fix everything else.

Competitor terms

Competitor brand queries often convert poorly, but they are also a deliberate strategy for some sellers. Decide per campaign. In a discovery campaign, a negative on a competitor’s brand name stops paying for clicks that rarely sell. In a conquest campaign built for that purpose, the same term is the target. The mistake is accidental conquest: broad or auto campaigns drifting into competitor names nobody chose.

Step 4: Apply at the right level

• Campaign level when the word is wrong for the product wherever it appears.
• Ad group level when it is wrong for one product in a shared campaign.
• Cross-campaign isolation when you have promoted a converting term to an exact campaign: add it as negative exact in the source campaign so the two do not bid against each other. Only do this where the source campaign is the weaker performer. If the source campaign converts well on that term, taking the term away to give it to a campaign with no history can lose sales.
• Product targeting: add negative ASINs for competitor pages that spend without selling, or for your own products you do not want cannibalized.

Order matters in a simple way: add negatives to automatic and broad campaigns first, because those are the ones that produce noise, and go lightly in exact campaigns where you chose the keywords yourself. For structure context, see how Sponsored Products campaigns are laid out.

Step 5: Record and review

Keep a log of every negative with date, campaign, match type and reason. Two weeks later (Amazon’s suggested minimum), check three things:

1. Did spend on the affected campaign fall by roughly what the removed terms cost?
2. Did orders hold, or did sales fall with the spend?
3. Did the campaign pick up new waste in the space the old terms left?

If orders fell with spend, the negative was wrong. Remove or relax it. We could not verify from Amazon’s documentation whether negatives are deleted or archived in the console, so keep your own log as the record rather than relying on the console to show history.

A worked example, with the threshold applied

Illustrative search term rows for Product A ($40, 30% target, flag threshold about $18):

Two of the five rows are not negatives at all: one is a bid problem and one is simply too early. That ratio is typical of a first pass.

Common mistakes

• Negating on too few clicks. Under Amazon’s 20-click guidance, a term has not had a fair chance.
• Using negative phrase for a generic word. Blocking “small” cuts every query containing it, including good ones. Prefer exact for single ambiguous words.
• Ignoring plurals and variants. Amazon says close variations are included, so you do not need to list every plural, but check the report for variants that slip through.
• Negating a term in the only campaign that gets orders on it.
• Skipping a review. Negatives are hypotheses. Check that the campaign still sells.
• Treating the negative list as the fix. If an auto campaign needs hundreds of negatives, the listing may be attracting the wrong shoppers; fix the title and images as well.

A weekly rhythm that works

Weekly for active launches, every two weeks for steady campaigns: export the report, apply the spend threshold, split into the four buckets, add negatives, log them, and check the previous round’s effect. If you manage dozens of campaigns, a script or tool is worth it; browse Amazon PPC software options with the caveat that some of them are made by the author, and see the metrics that mislead before you act on any single column.