Amazon negative keywords tell Amazon which shopper searches should never trigger your ad. Each one is set to negative exact or negative phrase, attached to a whole campaign or a single ad group, and stops you paying for clicks from people shopping for something else.
The hard part is judgment. Block too little and spend leaks; block too much, and you starve the queries feeding your best ASINs. In the accounts we manage, the negative list is often the quickest margin fix and the easiest one to overdo.
Here is how to get it right.
TL;DR
- Negative keywords come in two match types only: negative exact and negative phrase, and both also block close variations such as plurals and misspellings.
- Judge a zero-order search term against your own conversion rate, because 20 clicks with no sale proves far less on a product that converts at 5 percent than on one that converts at 20 percent.
- Leave the most recent seven days out of any zero-order review, since seller reports credit orders for up to seven days after the click.
- Use negatives to route traffic as well as block it: when a term graduates to an exact match campaign, negate it in the campaign it came from.
- Audit the negative list itself every quarter, because a stale block on a word that now converts costs you sales you never see.
Not sure which of your search terms deserve a block? Walk us through your search term report, and we will tell you which terms we would cut this week and which we would leave running. Book a Call
See What a Negative Keyword Actually Blocks
A negative keyword is an exclusion rule attached to a campaign or ad group. When a shopper's query matches it, that ad group sits out the auction for that search.
Say you sell a women's insulated winter coat and target the phrase match keyword "winter coat." Without exclusions, your ad is eligible for "boys winter coat," "winter coat rental," and "used winter coat." Each click on those costs the same as a click from your real buyer, and almost none of them will convert.
Three things negatives are good for:
- Removing wrong-intent traffic. Searches for a product you do not sell, a size or gender you do not carry, or a condition (used, refurbished, rental) that does not apply.
- Cleaning your data. Once the irrelevant clicks stop, your ACOS (Advertising Cost of Sales, the share of ad-attributed revenue spent on ads) and conversion rate describe your real buyers.
- Directing traffic inside your own account. A negative in one campaign can push a search toward the campaign where you want to pay for it (covered further down).
Amazon's own guide to Sponsored Products targeting frames the benefit the same way: fewer irrelevant placements can lift click-through rate (CTR) and return on ad spend (ROAS) while bringing down cost per click (CPC).
Choose Between Negative Exact and Negative Phrase
Amazon offers two negative match types. There is no negative broad match on Amazon; that option belongs to Google Ads, and looking for it in campaign manager is a common source of confusion.
| Negative exact | Negative phrase | |
|---|---|---|
| What it blocks | The exact query, plus close variations | Any query containing the complete phrase, plus close variations |
| Word limit | 10 words | 4 words |
| Character limit | 80 characters | 80 characters |
| Best use | One specific losing search you want gone | A word or short phrase that is wrong in every context |
| Main risk | Too narrow; the loser comes back in a slightly different form | Too wide; one word can wipe out good traffic |
Those limits come from Amazon's keyword match types reference, and they shape strategy more than most guides admit.
Two practical consequences:
- Long-tail blocks have to be exact or trimmed. A losing six-word query cannot go in as negative phrase. Either add it as negative exact, or find the two- or three-word core that makes it wrong ("rental," "for dogs," "replacement lid") and negate that as phrase.
- You cannot reliably split singular from plural. Close variations cover plurals and misspellings, so negating "coat hanger" can also block "coat hangers." If one form converts and the other does not, a negative keyword will not separate them; a bid change will.
Our default: start with negative exact on any query that is relevant in theme but losing money, and save negative phrase for words that are wrong everywhere you advertise, such as "free," "used," or a product type you do not make.
Pick the Right Level and Ad Type
Where a negative sits decides how much traffic it removes. A campaign-level negative applies to every ad group in that campaign; an ad group negative applies only to that ad group.
Campaign-level fits when:
- The term is wrong for every product in the campaign.
- You run a single ASIN per campaign, so the two levels behave the same.
- You are separating branded from non-branded traffic (more on that below).
Ad group level fits when:
- One ad group sells the product the term describes and another does not.
- You are routing a harvested term out of one ad group and into another inside the same campaign.
The tools also differ by ad type, and the search term report only covers some of them:
- Sponsored Products (SP), automatic targeting: negative keywords and negative products. Auto campaigns also show ASINs in the search term column, which you block with negative product targeting.
- Sponsored Products, manual keyword targeting: negative keywords.
- Sponsored Products, manual product targeting: negative products and brands.
- Sponsored Brands (SB): negative targeting matched to the campaign's targeting method.
- Sponsored Display (SD): negative products or brands. Display is audience- and product-driven, so there are no search queries to exclude.
Capacity is rarely the constraint. Amazon's sponsored ads campaign limits allow 10,000 negative keywords per campaign and 10,000 per ad group.
Build a Starter Negative List Before Launch
A new campaign has no search term data, but you already know some of what it should never pay for. Seed a short list on day one from your own listing, then let the report do the rest.
| Source | What to pull from it | Example negatives |
|---|---|---|
| Your product attributes | Sizes, materials, genders, or counts you do not carry | "toddler," "plastic," "single pack" |
| Condition and ownership words | Searches for a different buying situation | "used," "rental," "refurbished," "free" |
| Compatibility | Devices, models, or parts your product does not fit | A competing model number your accessory does not support |
| Your own catalog | Sibling products you advertise in a separate campaign | The other product type's name, negated in this campaign |
| Reviews and customer questions | Repeated confusion about what the product is | The product people expected instead of yours |
Keep the starter list short and mostly negative exact. Every word you add before launch is a guess, and a guess on a broad phrase can mute a campaign before it has shown you anything.
Find Negative Keyword Candidates in Five Steps
The search term report is where every good negative starts. Amazon's search term report documentation covers both Sponsored Products and Sponsored Brands, lists only searches that drew at least one click, and looks back 65 days.
Step 1: Pull the Full Window and Set the Last Week Aside
Download the summary report for the full 65-day lookback. For sellers, purchases in this report are counted within seven days of the click (vendor columns use 14 days), so the newest week of clicks has not finished collecting its orders.
Review zero-order terms on the older data only. A term that looks dead on Tuesday can show two orders by the following Monday.
Step 2: Sort Zero-Order Terms by Spend
Filter to search terms with no orders, then sort by spend from highest to lowest. The top of that list is where the money is; a term with one click and no sale is noise for now.
Export and keep each report. With only 65 days available in the console, a low-volume term that never builds enough clicks in one window can still show a clear pattern across three or four saved reports.
Step 3: Split Wrong Intent From Weak Performance
Read each high-spend, zero-order term and put it in one of two piles:
- Wrong intent. The shopper wanted something you do not sell. Block it now, whatever the click count.
- Right intent, no sale yet. The shopper wanted your product and did not buy. Hold it for the threshold test in the next section and the listing checks further down.
Not every row in the report is a typed search. Amazon notes that some terms are inferred as the best match for your product in non-search contexts, such as off-Amazon placements, so judge those rows by their numbers rather than their wording.
Step 4: Run a Single-Word Pass
Split every search term into individual words and total the clicks, spend, and orders for each word across the whole report. One word ("refill," "kids," "replacement") can be cheap in each query and expensive in aggregate.
A word that appears in dozens of losing queries and none of your converting ones is a strong negative phrase candidate. Check the converting list before you add it; one word inside your best-selling query is enough to rule it out.
Step 5: Add the Negatives at the Right Level
In Campaign Manager, open the campaign, choose the Negative keywords tab for a campaign-level block, or open the ad group and use its Negative targeting tab for an ad group block. Enter the terms, pick negative exact or negative phrase, and save.
Write down what you added, where, and why. That log is what makes the quarterly audit later in this guide possible.
Running more campaigns than your team has hours to review? We run search term reviews, routing, and negative list audits inside our managed Amazon advertising programs. Explore Our Services
Set a Click Threshold From Your Conversion Rate
Amazon suggests evaluating a keyword after at least 20 clicks before negating it, then letting negatives run for two weeks or more before changing strategy. Treat 20 clicks as a floor; the right number depends on how often your product converts.
The math is simple. If a search term were as good as your average, the chance it shows zero orders after a given number of clicks is (1 minus your conversion rate) raised to the number of clicks. A zero-order streak only becomes real evidence once it would be unlikely for a healthy term.
| Your conversion rate | Clicks per order, on average | Chance a healthy term shows zero orders after 20 clicks | Clicks before a zero-order streak is under a 10% fluke | Clicks before it is under a 5% fluke |
|---|---|---|---|---|
| 5% | 20 | 36% | 45 | 59 |
| 8% | 12.5 | 19% | 28 | 36 |
| 10% | 10 | 12% | 22 | 29 |
| 15% | 6.7 | 4% | 15 | 19 |
| 20% | 5 | 1% | 11 | 14 |
Read the first row closely. At a 5 percent conversion rate, more than a third of perfectly good search terms will sit at zero orders after 20 clicks, so a flat 20-click rule would cut real buyers.
How to use the table:
- Find your row. Use the ad group's own conversion rate from the last 60 days, since account averages hide big differences between products.
- Pick your tolerance. The 10 percent column suits cheap clicks and a deep keyword pool; the 5 percent column suits expensive clicks on terms you would hate to lose.
- Multiply by your CPC. The click count times your average cost per click is what it costs to prove a term is a loser. If that figure is larger than your profit on one order, lower the bid on the term while you collect data instead of paying full price for the test.
- Skip the math for wrong intent. A clearly irrelevant term does not need 59 clicks to prove itself.
A worked example, with illustrative numbers: an ad group converts at 10 percent and pays a $1.20 average CPC. At the 5 percent tolerance, a right-intent term needs 29 clicks with no order before you block it, which is about $34.80 of spend. If one order earns you $25 after fees and product cost, the test costs more than a sale is worth, so drop the term's bid while it collects clicks.
Use Negatives to Route Traffic Between Campaigns
Most accounts use negatives only to block. The more valuable use is steering each search to the campaign where you have set the right bid for it.
The common setup is a discovery layer (automatic and broad match campaigns) feeding a control layer (exact match campaigns). When a search term proves itself in discovery, you add it as an exact match keyword with its own bid. Then you add the same term as negative exact in the discovery campaign it came from.
Skip that second step and the two campaigns can both stay eligible for the same search. Amazon's match type guidance notes that when one keyword runs on several match types with different bids, the highest bid wins, so an unchecked broad bid can quietly override the exact bid you chose.
Branded traffic needs the same treatment. Negate your own brand name in your non-branded campaigns so branded searches run only through your brand campaign. Your generic campaigns then show their true cost of winning new customers, and you can see where defensive branded spend has reached diminishing returns and the budget would work harder on growth terms.
Competitor brand names call for a decision rather than a reflex. If you run a deliberate conquesting campaign, keep those names out of its negatives and negate them in every other campaign, so the spend sits in one place you can judge on its own numbers.
Rule Out a Listing Problem Before You Negate
A relevant search term with plenty of clicks and no orders is often a product page problem wearing a keyword costume. Negating it hides the symptom and leaves the cause in place.
Before you block a right-intent term, check:
- Price against the page-one results for that exact search, on the day you review it.
- Featured Offer status on the advertised ASIN, including how much of the review window you actually held it.
- Stock and delivery promise, because a slow ship date loses buyers who have faster options a scroll away.
- Main image and title against what the query asks for (size, count, material, compatibility).
- Rating and recent reviews, particularly any new negative review sitting at the top of the page.
- Variation landing, whether the click lands on the child ASIN the shopper searched for.
If one of these is the problem, fix it and give the term another full threshold window. If all six look healthy and the term still fails the threshold test, block it with a clear conscience.
Audit the Negative List on a Set Schedule
Negatives are easy to add and easy to forget. Amazon notes they are not permanent and can be removed at any time, which only helps if someone actually looks.
| Cadence | What to review | What you are looking for |
|---|---|---|
| Weekly | Search term report, older than seven days | New wrong-intent terms and high-spend zero-order terms past your threshold |
| Monthly | Single-word totals across all campaigns | Words that keep showing up in losing queries |
| Monthly | Harvested exact terms | Matching negative exact present in every discovery campaign |
| Quarterly | The full negative list, exported | Negative phrase entries that now block converting queries, and seasonal blocks that should come off |
| Quarterly | Organic and Brand Analytics search data | Terms you blocked months ago that now carry real demand for your product |
The quarterly check is the one most teams skip. Search a list of your top converting queries against your negative phrase list; any converting query containing a negated phrase means a block is cutting into sales you can no longer see in your ad reports.
Measure Whether Your Negatives Are Working
Good negatives show up in a few numbers within one or two report cycles. Track them on the ad group, since account totals blur the effect.
- Wasted spend share. Spend on zero-order search terms divided by total spend. Negatives should move this number first.
- CTR and CPC. Fewer irrelevant impressions should lift click-through rate; CPC tends to settle as the traffic mix tightens.
- Conversion rate and ACOS. Both should improve as wrong-intent clicks leave the mix.
- Impressions on core terms. A sharp drop on terms you want to win is the classic sign of an over-broad negative phrase.
- TACOS (Total Advertising Cost of Sales, which sets ad spend against all sales, organic included). If ACOS improves but total sales fall, you may be blocking demand along with waste.
Read them together. Lower spend with flat sales is a clean win; lower spend with falling sales means some of what you blocked was working.
Frequently Asked Questions
Quick answers to the follow-ups operators raise after their first negative list cleanup.
How Many Negative Keywords Can I Add to an Amazon Campaign?
Amazon allows up to 10,000 negative keywords at the campaign level and another 10,000 per ad group. Few accounts get close; if yours does, the campaign is probably targeting too broadly in the first place.
Do Negative Keywords Work in Automatic Campaigns?
Yes. Automatic Sponsored Products campaigns accept negative keywords and negative products, and they are where most wrong-intent traffic shows up. Review auto campaign search terms every week, including the ASIN rows.
Should I Add My Competitors' Brand Names as Negatives?
Negate them in every campaign that is not built to conquest them. If you do run a competitor campaign, keep the spend there so you can judge it separately against its own goal.
Can I Undo a Negative Keyword If I Block the Wrong Term?
Yes. Remove (archive) the negative in Campaign Manager, and the ad group becomes eligible for that search again. Give it a fresh threshold window before you judge it a second time.
Should I Negate a Search Term That Has Sales but a High ACOS?
Usually not as a first move. Lower its bid, or move it into an exact match campaign at a bid your margin can carry, and negate it in the source campaign. Block it outright only if it still misses your target at the lowest bid that keeps it serving.
Keep Negatives Working for Your Margin
A negative keyword list is a set of standing decisions about which shoppers you are willing to pay for. Build it from your own conversion rate, keep it out of the newest week of data, use it to route traffic as well as block it, and reopen it every quarter. Those habits keep each block earning its place.
Want a second opinion on your negative list? Book time with us, bring a recent search term export, and we will show you the blocks we would add, the ones we would lift, and why. Book a Call