Amazon Negative Keywords: Choose Match Type, Scope and When to Exclude

Quick answer: exclude a defined mismatch, not every term without sales

Amazon negative keywords restrict ad eligibility for specified shopping queries within the scope where you apply them. Start with the advertised product and the reason for excluding traffic. Then select the negative entity, matching option and campaign or ad-group scope. Keep a record of the previous configuration, the supporting evidence and the relevant queries you intend to preserve.

Use negative exact when the evidence supports a narrower query-level exclusion; consider negative phrase when the unwanted phrase represents a broader intent you do not want anywhere in the selected scope. Neither is a literal-string-only guarantee: Amazon describes close variations for both negative matching options. Product exclusions address product targets rather than substituting an ASIN into a keyword decision. Amazon match-type documentation

The practical sequence is evidence → match and scope → authorized change → configuration check → time-aligned review. If the evidence is incomplete, retaining the candidate for review is a valid outcome. Adding more negatives is not itself a performance objective.

Negative exact, negative phrase and negative products solve different problems

Compare the entity before comparing the match type

Keep keyword matching and product targeting separate. A query about a material is not the same object as the product detail page where an ad appeared. Use this comparison to identify the decision, then confirm the supported controls in your campaign.

Swipe horizontally to view all table columns.

The keyword definitions follow Amazon's negative-match documentation. Product negatives can prevent product-targeted or automatically targeted ads from displaying against the matching product-page ASIN. Amazon match types, Amazon negative-target setup

Negative exact is narrower, but not an exact-text firewall

Do not build the approval around “only these characters will be affected.” Instead, state which intent should be excluded and which adjacent intents matter. A reviewer can then judge whether the proposed matching option is acceptable even with close variations.

Equally, do not copy an exhaustive list of positive-keyword matching behaviors into the negative-keyword section of your operating procedure. Check the documentation for the negative option itself. Where a particular plural, translation or spelling difference is business-critical, verify that edge case rather than asserting an unsupported universal rule.

An ASIN-like value needs its report context

Amazon's search term report can include product identifiers as well as words. The report also includes inferred terms in non-search contexts and only terms associated with at least one ad click. It is not a complete list of everything shoppers typed or every impression the campaign received. Amazon Sponsored Products search term report

Retain the source campaign and targeting context when reviewing such a row. Do not paste an identifier into a keyword exclusion merely because it appeared in a column labeled customer search term. First establish which entity the row represents.

Use six checks to decide whether a term belongs on the negative list

1. Product relevance: what requirement cannot this offer meet?

A useful relevance explanation names the mismatch: material, compatibility, size, intended use or another verified attribute. “Bad keyword” is not enough. Check the actual advertised offer, not just the parent category or a remembered version of the listing.

Separate clear incompatibility from weak sales performance. An incompatible accessory query may have a straightforward relevance problem. A relevant query with poor conversion could instead point to price, availability, content, competition or traffic economics. Exclusion is only one possible response; the Amazon conversion-rate improvement guide covers the offer-side investigation.

2. Evidence quality: are you looking at a comparable, mature period?

Keep account, marketplace, campaign, ad group, currency and report dates attached to the candidate. A blank sales field is not automatically a recorded zero. Do not combine different products and declare the resulting average to be evidence against the phrase everywhere.

Recent results can also change. Amazon explains that conversions are credited to the interaction date and that attribution metrics remain incomplete until the applicable lookback window ends. A reporting update delay and a completed attribution window are different things. Amazon ad campaign attribution

For a repeatable way to prepare candidates, use the search term report analysis workflow. Its output should remain a review queue, not an automatic upload list.

3. Entity: keyword or product exclusion?

Document whether the candidate refers to a shopping-query intent or a product target. If the source is ambiguous, resolve it before selecting an action. The same visible text can lose important meaning when separated from its original report fields.

4. Matching coverage: which neighboring intents could be affected?

Write down examples you want to exclude and examples you want to preserve. These protected examples are especially important for short phrases, brand terms, materials and words with multiple meanings. They are review aids, not a complete simulator of Amazon's matching system.

5. Scope: do all affected products justify the same decision?

A campaign-level change deserves a campaign-level review. If the argument concerns one ad group, do not silently turn it into a restriction on other groups. Conversely, repeatedly adding the same narrow restriction everywhere without checking the common rationale can create a maintenance problem. Choose the narrowest scope that actually satisfies the documented requirement.

6. Verification: who will check the accepted setting and its consequences?

Assign an owner and a review point before execution. That owner needs the intended target, the previous state and the relevant report period. Without those, a later reviewer cannot distinguish a correct exclusion from a configuration mistake or unrelated traffic change.

Amazon's targeting guide suggests assessing performance after at least 20 clicks. That is evaluation guidance, not a rule that the twentieth click without a purchase must trigger exclusion, and not a universal statistical threshold. Product relevance and the evidence checks still matter. Amazon Sponsored Products targeting guide

Example: why an ad-group problem can become a campaign-wide mistake

The two-ad-group setup

Consider a fictional homeware campaign with two groups: Group A advertises stainless-steel mixing bowls; Group B advertises plastic storage bowls. A reviewer sees “plastic mixing bowl” associated with Group A and proposes the negative phrase “plastic” at campaign level.

That proposal changes the question from “Does this intent fit Group A?” to “Should every group in the campaign reject plastic-related intent?” The second question has a different answer because Group B intentionally sells plastic products.

Swipe horizontally to view all table columns.

These rows illustrate decision boundaries, not measured traffic or an exhaustive prediction of matching. Actual eligibility, variations and delivery require account-level verification.

What a better proposal says

The revised record might say: “Review a negative exact candidate for the unsuitable mixing-bowl intent in Group A. Do not apply a campaign-wide negative phrase for plastic. Check related intended queries before approval.” This does not automatically approve the exact candidate; it makes the proposed scope and the unresolved matching risk visible.

Amazon supports negative targeting in both automatic and manual campaigns, with campaign or ad-group application as documented for the relevant controls. That flexibility is why identifying the selected level matters. Amazon Sponsored Products targeting guide

A reusable list should contain reasons, not universal forbidden words

Words such as “cheap,” “free” or a material name should not become account-wide negatives just because they appear in a generic template. The right decision depends on the offer and query context. A term can be unsuitable for one product and useful for another.

Build a reusable review format rather than a supposedly universal blocklist. Preserve the rationale when copying an approved decision to another scope, and require a fresh relevance check there.

Check negative keyword word limits and preserve the intended meaning

Validate both words and characters

Amazon lists a maximum of four words and 80 characters for negative phrase, and ten words and 80 characters for negative exact. Check both constraints and the actual acceptance response; a long entry can fail even when its word count seems acceptable. Amazon negative keyword match types

Do not solve an invalid long phrase by cutting it down to a broad noun without review. Removing context can change which intent the exclusion expresses. Return the candidate to the reviewer if the supported format cannot represent the intended decision adequately.

Keep exact text alongside any normalized comparison key

A spreadsheet may normalize capitalization or spacing to find duplicate candidates. Retain the original value as well as the campaign, ad group, entity and matching option. Two rows with the same visible phrase are not duplicate decisions if they address different scopes.

The review key should therefore include the target context, not just the keyword. This prevents an automated deduplication step from dropping an important exception or treating an already-approved campaign decision as a new ad-group change.

How to add Amazon negative keywords through a reviewed workflow

Prepare the change outside the live account first

1. Preserve the source report and note the extraction date, reporting period and marketplace.
2. Record the candidate, relevance explanation and any unresolved evidence issue.
3. Select the proposed entity, matching option and exact scope; list protected intents.
4. Check existing negatives and related positive targeting for contradictions or redundancy.
5. Obtain the account owner's approval for the specific change, not a blanket instruction to clean up traffic.

The output is a small, reviewable change set. If the reviewer cannot identify which products it affects, stop before entry. A spreadsheet sorted by spend alone is not an adequate specification.

Enter and verify in the advertising console

In the appropriate account, open the intended campaign and, where relevant, its ad group. Use the available negative-targeting controls to add the keyword, select its negative matching option and save. Product exclusions use their relevant product-target controls instead. Confirm the labels and options in the current account. Amazon negative-target setup

After saving, inspect the resulting configuration. Record the entity, text or product target, matching option, scope, status and time. Do not treat the presence of a save confirmation as proof that every intended row was accepted correctly.

For bulk operations, inspect row results rather than only upload status

Amazon offers bulk-file campaign updates. Use a fresh account template and validate the supported fields and operations for the entity you intend to change. Amazon campaign measurement and improvement guide

A file-level completion message is not a substitute for reviewing accepted and rejected rows. Preserve the submitted version, the processing results and the configuration check. Never repeatedly resubmit an entire file just because one row failed: reconcile the accepted rows before deciding what to retry.

Why negative keywords may appear not to work

Separate acceptance, processing and attribution

Amazon's negative-target help lists processing buffers of 72 hours for negative keywords and 96 hours for negative products or ASINs. Use the current account guidance when planning a check; do not promise an immediate effect. Those buffers are not conversion-attribution windows or evidence that performance will improve after a fixed delay. Amazon negative-target setup

A later report can still contain earlier activity. Compare the interaction period with the change time before calling a row evidence of failed exclusion. Similarly, a lower spend total after the change does not prove a causal improvement: other targeting, bids, availability, demand and competition may have changed.

Run a six-point diagnosis

1. Identity: Is this the intended account, marketplace, campaign, ad group and negative entity?
2. Acceptance: Was the particular entry accepted, with the intended matching option and valid length?
3. Timing: Has the documented processing allowance passed, and what does the current configuration show?
4. Report period: Does the apparent contradictory activity actually belong to the period after the change?
5. Other coverage: Could another campaign or an organic listing explain what you observed?
6. Matching context: Is the observed query or product context within the intended exclusion, including the documented matching behavior?

A manual search is only one observation. It does not identify every serving campaign, reproduce every shopper context or explain historical reporting. Do not click your own ad to manufacture a test result. Use identifiable configuration and report evidence, and send unresolved discrepancies to support with the relevant IDs, timestamps and screenshots.

Monitor the traffic you intended to keep

Check the protected intents as well as the excluded candidate. If useful traffic declines, investigate whether the exclusion's scope or meaning was broader than intended. A disappearance of the unwanted row alone does not establish that the whole change was successful.

Keep the conclusion proportionate: “configuration accepted and reviewed” is different from “exclusion confirmed for the observed period,” and both differ from “profit improved because of this change.”

Correct an overly broad negative without promising instant recovery

Changing the matching option requires a new decision

Amazon states that the matching type of an existing negative keyword cannot be changed. Use the currently supported removal or archival and replacement path for your account rather than assuming an editable match-type switch. Verify each resulting state. Amazon negative-target setup

Before correcting the entry, preserve the erroneous setting and its scope. Identify the intended replacement, if one is necessary, and obtain approval for that correction. Do not add a narrower negative while leaving an unintended broader restriction in place and assume the problem has been solved.

Recheck the cause, not just the symptom

If a shared list or scheduled process created the problem, correct the source decision too. Otherwise, the next run may recreate the same exclusion. Record why the original rationale did not apply, including any affected product group the reviewer missed.

Removing a setting cannot recover auctions already missed or refund prior spend. It also does not guarantee that traffic returns to a previous level. Review the corrected configuration and subsequent evidence without presenting reversal as a performance guarantee.

Routing a winning term is not the same as excluding irrelevant traffic

Some negative decisions are intended to channel a useful term into a different targeting structure. Label that purpose explicitly. It needs evidence that the intended destination is configured and viable; a newly added positive target is not, by itself, proof of replacement delivery.

Do not classify a profitable intent as unwanted merely because it is being reorganized. Keep routing decisions separate from relevance exclusions so the approval and recovery checks address the actual goal.

Choose manual review, native controls or automation according to the risk

Manual and native workflows suit small, explainable changes

For a few candidates, a report, a decision record and the advertising console may be enough. The advantage is direct inspection; the weakness is inconsistent documentation when several people work in the account. Require a shared scope and status convention before introducing more tooling.

Native targeting controls apply the approved setting. They do not replace the team's judgment about product relevance or acceptable collateral effects. Keep the explanation outside a person's memory so another operator can review it later.

Scripts and no-code workflows can prepare candidates, not manufacture certainty

Automation can help structure files, detect missing scope fields and assemble a queue. Use explicit input validation and a read-only or dry-run stage before any separately authorized write integration. Reject ambiguous entities, incomplete IDs and unreviewed scope expansion.

At larger scale, preserve per-row decisions, acceptance results, exceptions and review ownership. Stop the process when its source format changes or when the proposed scope no longer matches the approval. A repeated workflow needs monitoring, not just a recurring schedule.

Agent assistance needs the same evidence and permission boundaries

An agent can help draft the explanation of a candidate, but a plausible explanation is not verified product knowledge. Require a reviewer to check the advertised offer, original report and affected groups. Do not infer permission to modify campaigns from permission to analyze a file.

Any connected execution path needs its own current checks for account authorization, supported operations, platform requirements and recovery behavior. This guide does not establish that a particular third-party integration is approved or available. For the broader operating model, see Amazon PPC automation.

Use OpenMax to evaluate the review handoff, not assume an Amazon executor

Where collaboration fits

The coordination problem starts when the analyst identifies a candidate, the product owner knows whether it is relevant, and a separate operator can change the campaign. OpenMax presents itself as a human-and-agent collaboration platform. That positioning makes a shared review handoff a relevant workflow to evaluate, but does not establish an Amazon negative-keyword integration. OpenMax

This article has not verified OpenMax reading an Amazon advertising account, applying negative targets or recovering a live change. Keep execution in the currently authorized tool unless the required integration and controls have been separately demonstrated. A single operator with a small queue may not need another collaboration layer.

A seven-field decision record for a narrow pilot

Use the following fictional record to test whether the handoff preserves the information each participant needs. It is a review specification, not an Amazon bulk-upload format or a claim about an OpenMax screen.

1. Scope: homeware campaign / Group A / stainless-steel mixing bowls
2. Candidate: plastic mixing bowl; keyword intent, not a product ID
3. Evidence: source report, period and verified material mismatch
4. Proposal: review negative exact in Group A only; not yet authorized
5. Protect: Group B plastic storage-bowl intent; inspect close variants
6. Ownership: named reviewer and authorized operator; decision pending
7. Verification: before-state, accepted result, change time, follow-up owner

Start with a read-only review of one campaign. Check whether the receiving person can explain the intended exclusion and the protected traffic without asking the analyst to reconstruct the work. Only then evaluate additional delegation and any separately verified integration. Do not put account credentials or unnecessary customer information into the review packet.

FAQ: Amazon negative keyword decisions

What is the difference between negative exact and negative phrase?

Negative exact addresses a specified query and close variations; negative phrase can exclude queries containing the phrase or close variations. Choose according to the intent you mean to remove, not just the shortest available entry. Check nearby relevant intents before approving either option.

Do Amazon negative keywords include plurals and close variants?

Amazon explicitly describes close variations for negative matching. Do not promise literal-text-only blocking or assume every rule documented for positive keywords transfers unchanged. Verify any specific plural or variation that materially affects your decision.

Can I add negative keywords to an automatic campaign?

Yes, Amazon documents negative targeting for automatic as well as manual campaigns. Confirm the available keyword or product control, the campaign type and the level where you are applying it. Automatic targeting does not remove the need for scope review.

Should I add negatives at campaign or ad-group level?

Use the level justified by the evidence. A mismatch affecting one ad group should not automatically restrict other groups with different products. A campaign-level proposal needs a review of all affected groups and the relevant intent each should retain.

Should every search term with no sales become a negative keyword?

No. Check relevance, report completeness, attribution maturity and offer-side explanations first. Missing sales data is not the same as zero sales. A relevant term with weak performance may need investigation or a different advertising decision rather than exclusion.

How long do negative keywords take to take effect?

Amazon's negative-target help lists a 72-hour keyword buffer and a 96-hour product-target buffer. Confirm current account guidance and the accepted setting before investigating an apparent failure. These processing allowances do not determine when conversion reporting becomes complete.

Can I change a negative keyword from phrase to exact?

The existing matching type is not editable according to Amazon's help. Review the supported correction and replacement process in the account, preserve the prior state and verify that an unintended broader restriction is no longer left active.

Is there a universal Amazon negative keyword list I can upload?

No list can establish relevance across every catalog and campaign. Reuse a decision-record template, then evaluate the specific offer, candidate, matching option and scope. Generic words can describe valuable intent for some products and unsuitable intent for others.

Next step: review one exclusion and the traffic it must preserve

Choose one candidate from a validated search term report. Write the mismatch, proposed match and scope, protected intents and verification owner in the seven-field record. If those details are not clear, resolve them before changing the account.

For a team handoff, evaluate that read-only case with OpenMax before considering connected execution. The useful outcome is a decision another person can inspect and correct—not a longer negative list or an unsupported promise of lower advertising costs.