Negative Keywords on Amazon: How to Find and Use Them

Negative keywords on Amazon are search terms you tell Amazon not to show your ads against. They are the cheapest optimization in Amazon PPC, and in most accounts we audit they are either missing or set by gut feel.

This guide covers how they work, where to add them, how to find them, and the math behind the hardest question: how many clicks a term gets before you give up on it.

What Are Negative Keywords on Amazon?

A negative keyword is a word or phrase you add to a campaign or ad group so your ads don't appear when a shopper's search matches it. Amazon calls this negative keyword targeting, and it works in Sponsored Products and Sponsored Brands.

Regular keywords say where you want to show up. Negative keywords say where you don't, and they override the positive side: if a search matches both, the negative wins and no ad serves.

Negatives exist because Amazon's matching is deliberately loose. Amazon's match type documentation says a broad match keyword can match synonyms and related terms, and the keyword itself "may not be contained" in the shopper's query. Auto campaigns go further, matching by close match, loose match, substitutes and complements.

Our match types guide covers each one.

That reach is useful for discovery. It also means a campaign for "white shirts" can end up paying for "black shirt", "shirt hangers" and a competitor's brand name.

Negatives do three jobs in a working account:

• Cut waste. They stop spend on searches that will never convert for this product.

• Route traffic. They push a search to the campaign and bid you chose for it, instead of whichever campaign happens to win.

• Separate products. They keep a "dark chocolate" ad off "milk chocolate" searches when you sell both.

Negative Exact vs Negative Phrase

Amazon offers two negative match types. There is no negative broad in Sponsored Products.

According to Amazon's help page on negative targeting:

• Negative exact blocks searches that match the exact phrase or a close variation (plurals, misspellings). Up to 10 words and 80 characters.

• Negative phrase blocks any search that contains the complete phrase, or a close variation, anywhere in it. Up to 4 words and 80 characters.

You can't change the match type of an existing negative. To switch, archive it and add it again.

Our default is negative exact. It removes exactly the search that failed and nothing else.

We use negative phrase only for concepts that are wrong in every combination: "free", "used", "refurbished" for a new product, "kids" for an adult-only item, a size or material you don't sell.

Be careful with short phrases. A negative phrase of "case" on a phone accessories account can quietly block "phone case with card holder", which might be your best search.

Campaign Level vs Ad Group Level

You can add negatives at the campaign level or at the ad group level. The Amazon Ads API documentation describes the difference plainly: a campaign-level negative applies to every ad group in the campaign, an ad group-level negative applies only to that ad group.

How we choose:

• Campaign level for terms no product in the campaign should ever show on: irrelevant concepts, wrong categories, competitor brands you've decided not to buy.

• Ad group level for terms that are wrong for one product but right for another, and for search terms that failed in one ad group on performance.

One rule from our click-threshold work still holds. When you negate a term because it didn't convert, negate it as exact at the ad group level. Performance negatives are evidence about one product, not the whole campaign.

Negative Product Targeting

Negative product targeting does for detail pages what negative keywords do for searches. You exclude an ASIN or a whole brand, and your ads stop appearing on those product pages.

Amazon's negative targeting help page says it affects automatic targets and product attribute targets. An ad group-level negative product applies to targets in that ad group; a campaign-level one applies across the campaign.

ASINs show up as search terms too. In an auto or product targeting campaign, the search term report lists entries like "b00ipgvvz4", which are product pages your ad appeared on. Amazon notes you can block these through negative targeting.

We use negative product targeting for four things:

1. ASINs that spend without converting, judged with the same click threshold as keywords.

2. Competitors you can't beat on price or rating. Your ad on their page loses the comparison.

3. Your own ASINs in conquest campaigns, so spend meant for competitor pages doesn't land on yours.

4. Accessories or cheap items where a shopper on that page is never your buyer.

How to Find Negative Keywords for Amazon

The search term report is the source. Amazon's search term report lists every search that got at least one click, with spend and orders, and its lookback window is 65 days.

Pull it monthly at minimum, and save it. Anything older than 65 days is gone from the console. Our search term report guide covers the columns in detail.

1. Pull the Search Term Report and Drop the Last 7 Days

Sellers' Sponsored Products sales are attributed for 7 days after a click, and vendors' for 14, per the report's column definitions. A term clicked yesterday may still convert.

So exclude the most recent 7 days (14 for vendors) from any decision. Most premature negatives we find in audits were judged on clicks that hadn't finished attributing.

2. Negate Irrelevant Terms on Sight

Sort by spend and read the terms. Anything that is the wrong product, wrong category, wrong size or wrong intent goes now, even after one click.

Relevance needs no threshold. A dog bed ad showing on "cat tree" doesn't need 30 more clicks to prove anything.

3. Apply a Click Threshold to Relevant Terms

Relevant terms that spend without converting need evidence before you negate them. The next section explains how to set that number.

4. Run N-Gram Analysis

Single search terms hide patterns. Split terms into their words and pairs, and add up spend and orders per word.

If "toddler" appears in 40 different search terms that together spent $300 with no orders, that is a negative phrase candidate even though no single term crossed the threshold. Our n-gram analysis guide shows how to build it.

5. Add Negatives at the Right Level

Performance failures go in as negative exact at the ad group level. Irrelevant concepts go in as negative phrase at the campaign level.

Then note what you negated and when. Negatives take time to apply: Amazon says up to 72 hours for keywords and 96 hours for products.

For brand-new campaigns, add the obvious negatives at launch: "free", "used", competitor products you won't sell against, sizes and variants you don't carry. Our Amazon PPC keyword research guide covers building that list before the first click.

How Many Clicks Before You Negate a Search Term?

This is the question our old click-threshold post answered, and the one most negatives guides skip. Too early and you cut terms that would have converted. Too late and you pay for the same failure for weeks.

Amazon's own targeting guide recommends evaluating a keyword after at least 20 clicks before negating it. That is a sensible floor, but the right number depends on your conversion rate.

The Math: Zero Orders Is Normal for a While

If a term converts at your product's normal rate, some runs of clicks will still produce no orders by chance. The question is how unlikely a long run of zero orders would be.

The chance of zero orders in n clicks, when the true conversion rate is p, is (1 − p) to the power of n. At a 10% conversion rate:

• After 10 clicks with no order, there is still a 35% chance the term is a normal converter.

• After 20 clicks, 12%.

• After 30 clicks, 4%.

This gives a rule you can use on any product. Work out your clicks per order (1 ÷ conversion rate), then wait for a multiple of it.

At two times clicks per order, you are about 86% confident the term converts below your normal rate. At three times, about 95%. Statisticians know the three-times version as the rule of three).

We use 2x for cheap, high-volume terms and 3x for anything relevant that we'd hate to lose. A home goods product converting at 15% gets 13 to 20 clicks. An apparel product at 5% gets 40 to 60.

Use the conversion rate of the product in that ad group, not the account average. Our Amazon conversion rate guide explains where to find it and what's typical.

Tie the Threshold to Cost, Not Just Clicks

Converting below average isn't the same as losing money. The better test is whether a term converts below your break-even rate.

Your break-even conversion rate is your average CPC divided by the most you can pay for an order. The most you can pay for an order is price times margin before ad spend, which is your break-even ACOS in dollars.

Here is the worked example, with illustrative numbers:

• A $25 product with a 40% margin before ads can pay up to $10 for an order.

• At a $1.00 average CPC, it breaks even at a 10% conversion rate (1.00 ÷ 10).

• Three times 10 clicks per order is 30 clicks, or $30 of spend with no order.

Our old post used the simpler version: maximum cost per order ÷ CPC. With a $10 ceiling and a $0.25 CPC, that is 40 clicks. It's the one-times point, about 63% confidence, so treat it as the earliest you'd look, not the moment you negate.

When Not to Negate at All

Three kinds of terms shouldn't be negated on a click count:

• Your brand terms. They are cheap, defend your listing and usually convert. If one doesn't, fix the listing or the ad, not the targeting.

• Visibility terms. Head terms you need to rank for, like "wireless gaming mouse" for a gaming mouse brand. Move them to exact match and manage the bid.

• Terms with orders but high ACOS. They work, just at the wrong price. Lower the bid in an exact campaign instead of cutting them off.

Negation is for terms that don't convert. Bids are for terms that convert at the wrong cost. Mixing the two is how accounts lose good traffic.

How Negatives Work With Keyword Harvesting

Harvesting is moving converting search terms from discovery campaigns into exact-match campaigns where you control the bid. Negatives are what make it work.

1. Discover in Auto and Broad

Auto and broad campaigns find terms you'd never think to target. Amazon recommends running an automatic campaign for about two weeks before building manual campaigns from it.

2. Harvest Winners Into Exact

When a term passes your harvest rule, say two orders in 30 days at or under target ACOS, add it as an exact keyword in a performance campaign with its own bid.

3. Negate It as Exact at the Source

Then add the same term as a negative exact in the auto or broad campaign that found it. Without this, the term keeps serving from both campaigns, and you can't tell which bid is setting your price.

Use negative exact here, never negative phrase. Negative exact on "dog bed large" blocks that search in the auto campaign but still lets it find "orthopedic dog bed large washable", which is the next term you might harvest.

4. Keep Discovery Lean

Over time the auto campaign stops paying for your proven terms and spends only on new ones. That is the point: discovery budget buys discovery.

Some advertisers add a catch-all campaign with very low bids for the long tail. It needs the same funneling negatives, or it ends up competing with your exact campaigns.

Our campaign structure guide shows how auto, broad, phrase and exact campaigns fit together. The keyword harvesting funnel guide covers the harvest rules.

What Changed in 2025 and 2026

Three changes affect how we manage negatives today.

Search terms are no longer only searches. Amazon's search term report page, updated May 2026, says the report now includes terms Amazon infers for non-search placements, such as off-Amazon placements on social sites. A strange-looking "term" may be an inferred one, so check relevance before you negate it.

Prompts run on your existing targeting. Sponsored Products and Sponsored Brands prompts launched in open beta in November 2025 and became generally available in the US on March 25, 2026. Campaigns are enrolled automatically, use your campaign targeting, and are billed per click. Review the new Prompts report and pause prompts that spend without orders.

More targeting is automated. At unBoxed 2025, Amazon introduced Ads Agent, which recommends keywords and audiences, and Sponsored Products keyword groups are in beta in the US. The more Amazon chooses your targets, the more your negatives are where you keep control.

What hasn't changed: still two negative match types, the same word limits and the same 65-day report lookback.

Mistakes We See in Real Accounts

• Negating on too few clicks. Five clicks with no order on a product converting at 8% means almost nothing. There's a 66% chance of that from a perfectly good term.

• Negative phrase on short, common words. One two-word negative phrase can switch off a quarter of a campaign's traffic.

• Campaign-level negatives for product-level failures. A term that failed for one variant may be the best term for another.

• Harvesting without negating at the source. The term then serves from two campaigns and your exact bid stops controlling the price.

• Judging the last week of data. The 7-day attribution window means recent clicks still owe you orders.

• Never reviewing old negatives. Products, prices and listings change. A term negated two years ago may convert now.

A full Amazon PPC audit checks for every one of these, alongside bids, budgets and structure.

How We Run Negatives in Adbrew

Everything above can be done by hand in the Amazon Ads console. It takes a weekly afternoon per account, and it's the first task to slip when you manage twenty of them.

Adbrew is an agentic retail media management platform for Amazon and Walmart, and negation runs inside it on rules you write and can read:

• Search term negation rules. Set the lookback, the attribution buffer, the click threshold and the match type and level. Adbrew can calculate the threshold from each campaign's own conversion history, the "statistically significant clicks" approach from our old post, or you can set an absolute number.

• Harvesting and negation together. One rule moves a converting term into exact and adds the negative exact at the source in the same step, which is the funneling described above.

• Guardrails. Exclude brand terms and your visibility keywords from negation, and every change is logged with the rule that made it.

• Search terms and n-grams in one view. Review terms across every campaign, filter and negate in bulk from the Sponsored Ads management workspace.

• Adbrew Intelligence. Ask an agent for "search terms that spent over $50 with no orders in the last 60 days, excluding brand terms", review the list and approve. Adbrew Intelligence and our MCP server draft the changes; nothing applies until you approve it.

Brands on Adbrew cut ACOS by 46% on average, and Seller Presto saves more than 250 hours of manual optimization a month. Search term cleanup is a large share of those hours.

Spend Less on Searches That Never Convert

Negatives are a weekly habit with a little math behind it: relevance first, a click threshold tied to your conversion rate, and a negative at the source for every term you harvest.

Want to see search term negation and harvesting run on rules you control? Book a demo with our team and we'll walk through it on your account.