Every auto, broad or phrase campaign pays for searches that were never going to buy. The search term report shows which ones, but it runs to thousands of rows, so most sellers skim it.
The 13 prompts below turn that report into a reviewed list of negative keywords and negative ASINs: clean the export, set thresholds from your own margin, separate irrelevant traffic from terms that need more data, pick the match type, and check for conflicts before anything goes live.
Run them in order, paste your data where a prompt says [paste ...] and replace the bracketed numbers. Approve each negative yourself: a wrong one quietly switches off traffic that was selling.
What the search term report shows and what it leaves out
In the Amazon Ads console, create a Sponsored Products report with Search term as the report type. Download it weekly and keep your own archive, because the console only offers a limited look-back window (check the date range limit when you create it). For a broader campaign health check first, see our ad performance prompts.
Column
What it means
Use for negatives
Customer Search Term
What the shopper typed, or an ASIN when your ad showed on a product page
The candidate negative
Targeting, Match Type
The keyword or auto target that matched
Where the negative belongs
Clicks, Spend
Traffic and cost
What a negative would save
7 Day Total Orders (#), 7 Day Total Sales
Orders and sales attributed within 7 days of a click
Zero orders after enough clicks marks a candidate
7 Day Advertised SKU Sales, 7 Day Other SKU Sales
Sales of the advertised item versus your other items
A term may sell a different product of yours
Total Advertising Cost of Sales (ACOS)
Spend divided by attributed sales
Compare with your break-even ACoS
Two limits matter. Attribution lags: for sellers, Sponsored Products credits orders up to 7 days after the click (Sponsored Brands and vendor accounts use 14 days, and the column names change to match), so the latest week always looks worse than it is. And the report is built around clicks, so impression-only searches mostly do not appear. Column names shift between report versions, so check your header row.
Negative exact, negative phrase and negative ASINs
Sponsored Products has two negative keyword match types, no negative broad, plus negative product targeting. Amazon's help pages say both keyword types also block close variations such as plurals and misspellings, and currently limit negative phrase to 4 words and negative exact to 10, with 80 characters for either. You cannot change the match type of an existing negative; archive it and add a new one. Limits get revised, so check the Amazon Ads help center if an upload rejects a row.
Type
Blocks
Use it when
Risk
Negative exact
That exact search and close variants
One query wastes money but its words work elsewhere
Low
Negative phrase
Any search containing the phrase in order, plus close variants
A word is wrong for your product in any context
High: one phrase can block hundreds of searches
Negative product target
Your ad on one ASIN's detail page
ASIN rows with clicks and no orders
Medium
Negatives sit at campaign or ad group level. Amazon notes they can take a few days to apply fully.
Clean the report and set your thresholds
Prompt 1 · Clean the report and set your thresholdsChatGPT Claude
Act as an Amazon PPC analyst. Below is my Sponsored Products search term report for [date range] with Campaign Name, Ad Group Name, Targeting, Match Type, Customer Search Term, Impressions, Clicks, Spend, 7 Day Total Orders (#), 7 Day Total Sales, 7 Day Advertised SKU Sales and 7 Day Other SKU Sales. If there is a date column, exclude the last 7 days because attribution is incomplete; if not, tell me. Lowercase and trim every search term and flag terms that look like ASINs (10 characters starting with b0). Aggregate by search term across all campaigns, and separately by search term plus campaign plus ad group. Return both tables sorted by Spend with: search term, campaigns, clicks, spend, orders, sales, ACoS, conversion rate (orders divided by clicks). Report: [paste search term report]
Tip: If exporting and pasting this report every week is the step you skip, DataDoe connects your Amazon Ads search term data to Claude or ChatGPT through MCP, refreshed daily, with a ready-made negative keywords Skill that runs a version of this workflow.
Prompt 2 · Clean the report and set your thresholdsChatGPT Claude
Act as an Amazon advertising consultant setting negative keyword rules for one product. Selling price [price], landed cost per unit [COGS], referral and FBA fees per unit [fees], ad conversion rate [conversion rate], average CPC [CPC]. Calculate my margin before ads and break-even ACoS; the click count at which a zero-order term is unlikely to be a normal converter at my conversion rate (show the method and give a conservative and an aggressive threshold); and the spend at which a zero-order term has cost more than the profit on one unit. Finish with a short reusable rule set, such as zero orders and at least X clicks, or ACoS above Y with at least Z orders. List every assumption.
Why: At a 10% conversion rate, a perfectly normal term still has about a 1-in-8 chance of zero orders after 20 clicks. Thresholds from your own numbers stop you negating on gut feel.
Prompt 3 · Clean the report and set your thresholdsChatGPT Claude
Act as an Amazon PPC analyst running an n-gram analysis. Split every search term in the table below into single words, two-word and three-word phrases. For each n-gram, sum clicks, spend, orders and sales across the terms that contain it, and calculate ACoS and conversion rate. Ignore stop words such as for, with, and, the as single words. Return the 30 n-grams with the most spend and near-zero orders, and the 15 with the most sales, with three example terms each. My product: [one-line description with key attributes]. Data: [paste aggregated table from Prompt 1]
Why: A wrong size or audience word can hide across 60 searches with two clicks each. No row crosses your threshold, but the word does.
Find the search terms that waste spend
Prompt 4 · Find the search terms that waste spendChatGPT Claude
Act as an Amazon PPC manager reviewing negative keyword candidates for [product, with material, size, use and audience]. Apply this rule set: [paste rule set from Prompt 2]. Classify each term that meets it as irrelevant (different product, use or audience), relevant but low intent (research searches), relevant (it describes my product, so the problem is likely price, images or the listing), or unclear. Return a table with search term, clicks, spend, orders, class and a one-line reason. Do not recommend negating anything relevant; list those separately as listing or bid issues. Data: [paste aggregated table]
Why: A relevant term that does not convert points at your listing or price. Negating it hides the symptom.
Prompt 5 · Find the search terms that waste spendChatGPT Claude
Act as an Amazon advertising analyst. List search terms with at least [N] orders and ACoS above my break-even of [break-even ACoS]. For each, recommend lowering the bid on the matching keyword or target (name it from the Targeting column), moving the term to its own exact match keyword with a lower bid, or negating it. Negate only if the term describes a different product, or if 7 Day Advertised SKU Sales are near zero and the sales come from SKUs I could advertise directly. Return search term, targeting, orders, ACoS, recommendation, reason. Data: [paste table]
Prompt 6 · Find the search terms that waste spendChatGPT Claude
Act as an Amazon PPC analyst working on product targeting. Below are search term rows where Customer Search Term is an ASIN, so my ad showed on that detail page, with each ASIN's title and price: [paste ASIN, title, price]. My product is [product and price]. Judge each placement as complementary, a competitor I can beat on price or features, or a poor fit, and flag those with at least [N] clicks and zero orders as negative product targets. Put my own ASINs [paste your ASINs] in a separate table so I decide on them myself. Data: [paste ASIN rows]
Choose the match type and the level
Prompt 7 · Choose the match type and the levelChatGPT Claude
Act as an Amazon PPC manager. Choose negative exact or negative phrase for each approved candidate below. Use negative phrase only when the phrase is wrong in every search it could appear in and has no more than 4 words; otherwise use negative exact (up to 10 words, 80 characters). For every proposed negative phrase, list any converting term below that contains it, and if there is one, switch to negative exact or drop it. Return: negative, match type, reason, converting terms blocked (this should end up empty). Candidates: [paste]. Terms with at least one order: [paste]
Why: Negative phrase does the real damage when it is wrong, because it blocks searches you have not seen yet.
Prompt 8 · Choose the match type and the levelChatGPT Claude
Act as an Amazon PPC account manager. My structure: [each campaign with its ad groups, targeting type and products]. For each negative below, recommend campaign or ad group level and name the campaign and ad group. Use campaign level when the term is wrong for every product in the campaign, and ad group level when another ad group in the same campaign could still convert it. Flag negatives that belong in more than one campaign. Negatives: [paste output of Prompt 7]
Prompt 9 · Choose the match type and the levelChatGPT Claude
Act as an Amazon PPC manager setting up search term isolation. Below are terms that converted in my auto, broad and phrase campaigns, with orders, ACoS, CPC and source campaign: [paste]. For each with at least [N] orders and ACoS below [target ACoS], draft a new exact keyword for [exact campaign name] with a starting bid based on its CPC, and a negative exact for the same term in each source campaign. Return two tables: keywords to add, and negatives to add with campaign, ad group, term and match type.
Why: These negatives are about control, not waste: one term, one campaign, one bid. To plan that structure, see our campaign strategy prompts.
Check the list before anything goes live
Prompt 10 · Check the list before anything goes liveChatGPT Claude
Act as a careful Amazon PPC auditor. My proposed negatives with match types and locations: [paste]. My active keywords by campaign and ad group: [paste]. Every search term with at least one order in the last [60] days: [paste]. Find every conflict: a negative that blocks an active keyword in the same campaign or ad group, a negative phrase contained in a converting term, and any negative containing my brand [brand]. Return only the conflicts, each with a fix, or say plainly that there are none.
Prompt 11 · Check the list before anything goes liveChatGPT Claude
Act as an Amazon Ads bulk operations specialist. Convert the approved negatives below into Sponsored Products bulk upload rows. Use exactly the headers, entity names and match type values from this sample from my own account: [paste the header row, one ad group negative keyword row and one campaign negative keyword row]. Campaign and ad group IDs: [paste names with IDs]. Output CSV only, one row per negative, operation set to create, then the row count per campaign. Negatives: [paste final list]
Tip: Upload a small file first and read the processing report before sending the full list. If your AI client is connected to your ad account through DataDoe, the approved list can be applied without a bulk file: Actions stay off until you enable them, most AI clients ask before a change, and every change is logged.
Measure the result and keep the list clean
Prompt 12 · Measure the result and keep the list cleanChatGPT Claude
Act as an Amazon PPC analyst. I added the negatives below on [date]. Compare the [14] days before with the [14] days after, starting the after window [3] days after the change. Inputs: Spend, Clicks, Orders, Sales and ACoS per campaign for both windows [paste], and both search term reports [paste]. Report spend saved on negated terms, whether orders moved to other terms, and the change in ACoS and total orders per campaign. End with keep, adjust or remove for each negative that touched a term with past orders. Negatives: [paste]
Why: A negative can improve ACoS while cutting sales you wanted, so watch total orders. For what ads cost the business overall, pair this with our TACoS vs ACoS prompts.
Prompt 13 · Measure the result and keep the list cleanChatGPT Claude
Act as my weekly Amazon PPC assistant. Each week I paste the last [21] days of my Sponsored Products search term report. In order: exclude the latest 7 days; apply my rule set [paste from Prompt 2]; classify relevance for [product description]; propose negatives with match type and level using [paste rules from Prompts 7 and 8]; check them against my converting terms and active keywords [paste]; and flag existing negatives that now block a product I have added. Output one table I can approve line by line: term, spend, clicks, orders, negative, match type, campaign, ad group, reason. Put anything uncertain under Needs a human look. Report: [paste report]
Tip: Save the rules in a Claude Project or custom GPT so you only paste the new report. To find new terms worth bidding on, use our keyword discovery prompts.
FAQ
What is the difference between negative exact and negative phrase on Amazon?
Negative exact blocks one specific search and its close variants, such as plurals. Negative phrase blocks every search that contains the phrase, so it reaches much further.
How many clicks should a search term get before I add it as a negative keyword?
Base it on your conversion rate, not a fixed number. A common rule is two to three times the clicks you would expect per order, unless the term is clearly irrelevant.
Should I add negative keywords at the campaign or ad group level?
Use the campaign level when the term is wrong for every product in the campaign. Use the ad group level when one ad group's product does not fit the term but another ad group in the same campaign could still convert it.
How long do Amazon negative keywords take to work?
They are not instant. Amazon notes that negative keywords and negative product targets can take a few days to apply fully, so wait at least three days before judging the effect.
Can I add negative keywords to Amazon auto campaigns?
Yes. Auto campaigns accept negative exact and negative phrase keywords and negative product targets, and usually need the most negatives.
DB
David Brooks
Amazon seller and automation expert with hands-on experience in FBA, PPC, and logistics. Shares proven AI prompts to streamline selling and boost profits.
Free download
27 prompts top sellers actually use.
Listings, ads, reviews, analytics and support: the Amazon Seller Prompt Pack as one PDF.
