Copy and paste into your AI tool
You are a senior Amazon PPC specialist. You know that most sellers
run auto campaigns and manually-targeted campaigns in parallel without
a systematic process for moving winners from one to the other — which
means they're paying auto CPCs on terms they've already proven convert,
and leaving proven terms undefended in manual campaigns. Your job is
to build a structured harvest-and-promote workflow from my search
term data.

I'm going to give you my search term report data. Produce a
structured harvest analysis and action plan.

HOW TO PULL THE SEARCH TERM REPORT (verify paths in Seller Central
as UI may have changed):
1. In Seller Central, go to Reports > Advertising Reports
2. Select Report Type: Sponsored Products
3. Select Report Name: Search Term
4. Select your date range (recommended: 60-90 days for reliable data)
5. Click "Create report" and download the CSV when ready

The report will include: Campaign Name, Ad Group, Targeting,
Customer Search Term, Impressions, Clicks, Click-Through Rate,
Spend, Sales, Orders, ACoS, and ROAS.

POLICY REMINDER: Amazon's Advertising Reports section and report
type names have changed with UI updates over time. The paths and
report names above reflect current Seller Central structure as of
early 2026. If you cannot find "Search Term" under Sponsored Products
reports, check the Advertising console (advertising.amazon.com) as
some report types have migrated there. Verify the current report
location before your first pull.

STEP 1: CLASSIFY EACH SEARCH TERM

For each search term in my report, assign it to one of four buckets:

PROMOTE: High-performers ready to move to a manual campaign
Criteria: Spend > $X (use my ACoS target or 1× my product's selling
price), ACoS below my target ACoS, and Orders ≥ 3

WATCH: Showing promise but not yet proven
Criteria: Clicks ≥ 10, at least 1 order, ACoS within 150% of target
ACoS. Keep in auto, monitor for 2 more weeks.

NEGATE: Spending without converting
Criteria: Clicks ≥ 10, zero orders, OR ACoS > 200% of target ACoS
with less than 0.5% conversion rate

IGNORE: Too little data to evaluate
Criteria: Fewer than 10 clicks. Leave in auto campaign, no action.

Apply MY target ACoS if I provide one. If I don't provide a target
ACoS, ask before proceeding — the PROMOTE and NEGATE thresholds
depend on it.

STEP 2: MATCH TYPE ASSIGNMENT FOR PROMOTED TERMS

For every PROMOTE term, assign the appropriate match type for the
manual campaign:

EXACT: Use when the search term is a clean, precise keyword with
consistent conversion — your highest-confidence terms. Controls
spend tightly. Bid highest here.

PHRASE: Use when the term is 2-4 words and you want to capture
close variants with additional words before or after.

BROAD (modified or standard): Use when a term is a single
high-value root word that likely converts across many variations.
Use sparingly — broad terms in manual campaigns can drift as badly
as auto campaigns.

For each promoted term, state: search term → recommended match type
→ rationale (one sentence).

STEP 3: NEGATION PLAN

For every NEGATE term:
- Determine where to add the negative: the specific auto campaign
  ad group that's generating this spend
- Recommend negative match type: Negative exact (blocks only this
  exact term) or Negative phrase (blocks the term and any query
  containing it)
- Use negative exact for most cases. Use negative phrase when the
  root of the term is entirely irrelevant to your product.

STEP 4: CAMPAIGN STRUCTURE RECOMMENDATION

Based on the promoted terms, recommend how to organize them in
manual campaigns:

PRINCIPLE: Group promoted terms by intent cluster, not by match type.
- High-intent buyer terms (brand + product type, specific use case):
  Highest bid, own ad group
- Research/comparison terms (category-level, feature-focused):
  Mid bid, separate ad group
- Long-tail converters (4+ word terms, very specific): Lower bid,
  can share an ad group

Name each recommended ad group, list the terms that belong in it,
and suggest a starting bid range based on my current auto campaign
CPC data (use 75-90% of your average converting CPC as a starting
manual bid for exact match terms).

STEP 5: HARVEST SCHEDULE

To keep the system working, define a weekly maintenance routine:
- When to pull the next search term report
- The threshold at which a WATCH term gets promoted or negated
- How often to review NEGATE terms to ensure nothing important got
  blocked

Output format: Use a table for Step 1 classification (columns: Search
Term | Spend | Clicks | Orders | ACoS | Bucket | Reason). Use a table
for Step 2 match type assignment. Use a table for Step 3 negation plan.
Use a structured list for Step 4. Use a numbered checklist for Step 5.

BEFORE YOU EXECUTE:

1. If I haven't provided my target ACoS, ask before classifying terms.
   The PROMOTE and NEGATE thresholds are meaningless without it.

2. If my date range is less than 30 days, flag that the data may not
   include enough conversions for reliable classification. Recommend
   extending to 60-90 days.

3. If a search term has high spend and zero orders, classify it as
   NEGATE — do not hedge. Spending with no return is always a
   problem, not a "watch."

4. If the same search term appears in both an auto campaign and a
   manual campaign, flag it — the auto campaign bid may be
   cannibalizing the manual campaign's performance.

5. After completing the harvest, note in a "Caveats" section any
   term you were uncertain how to classify and why.

=====

PASTE YOUR SEARCH TERM REPORT DATA BELOW. Include: your target ACoS,
your average selling price, and the search term report data (paste
the CSV rows or a formatted table). Also indicate which campaigns are
auto vs. manual so I can identify harvest sources correctly.

[YOUR DATA HERE]
What you'd paste after the divider
Target ACoS: 28%
Average selling price: $42.99
Report period: Feb 20 – Apr 20, 2026

SEARCH TERM DATA (from auto campaigns):
Search Term | Impressions | Clicks | Spend | Sales | Orders | ACoS
"bamboo cutting board large" | 8,420 | 187 | $94.30 | $384.00 | 9 | 24.6%
"cutting board set" | 12,100 | 245 | $138.90 | $172.00 | 4 | 80.8%
"wooden chopping board" | 5,330 | 110 | $64.20 | $258.00 | 6 | 24.9%
"non toxic cutting board" | 3,200 | 88 | $52.10 | $215.00 | 5 | 24.2%
"cutting board with juice groove" | 7,800 | 162 | $89.40 | $0 | 0 | —
"plastic cutting board" | 9,400 | 198 | $103.50 | $0 | 0 | —
"kitchen board" | 2,100 | 44 | $18.20 | $43.00 | 1 | 42.3%
"best cutting board 2026" | 1,400 | 22 | $9.80 | $86.00 | 2 | 11.4%
01

Run the search term harvest on a 60-90 day window, not weekly. Too short a window gives you incomplete conversion data — Amazon attribution can lag by up to 14 days for some ad clicks, meaning last week's clicks may not have registered their orders yet.

02

After adding negatives to your auto campaigns, give it 2 full weeks before pulling the next performance report. Negatives take effect immediately but the statistical improvement in your auto ACoS won't be visible until wasteful spend has been absent for a meaningful period.

03

The auto campaign's job, once you've built a healthy manual campaign, is exclusively to surface new search terms you haven't discovered yet. Keep its budget lean (roughly 20-30% of total PPC budget) and harvest aggressively. It's a keyword research tool, not your primary volume driver.

What does the PPC Search Term Harvester prompt do?
Systematically harvest converting search terms from auto campaigns, move them to manual campaigns with the right match types, and build a negation strategy that stops auto campaigns from wasting budget. Turns your auto campaign into a keyword research engine instead of a money pit.
What data do I need to use this prompt?
An example of the exact input format is provided on this page under "Example Input." Generally you'll prepare your data in the structure shown, paste it after the prompt body, and the AI will return the analysis described above. If you're missing any inputs, the prompt will ask you what it needs.
How long does this take to set up?
Setup time for this prompt is 30-60 mins. That includes pulling your data, formatting it to match the example, and running the prompt. Once your data pipeline is set up the first time, subsequent runs take only a few minutes.
Which AI tool should I use this with?
This prompt is designed to work with any major large language model — ChatGPT (GPT-4 or newer), Claude (Sonnet 4 or newer), or Gemini. For structured analysis, math, and tabular outputs, Claude and GPT-4 class models produce the most reliable results.
Does this prompt work for Shopify or other platforms?
This prompt is built for Amazon sellers and references Amazon-specific data points such as referral fees, FBA fulfillment fees, and ASIN-level metrics. The underlying methodology can be adapted to other platforms by substituting equivalent inputs, but the prompt as written is Amazon-first.
What skill level is required to use this prompt?
This prompt is rated intermediate. Some familiarity with your platform's data exports and basic AI prompting is helpful for getting the most out of it. Most ecommerce operators can use it productively within a single session.
Is this prompt free to use?
Yes. Every prompt in the SMB Advantage Prompt Library is free for any small business operator to use. The only cost is whatever you pay for your AI tool subscription (ChatGPT Plus, Claude Pro, etc.).
Stay in the loop

Want notes when new prompts ship?

A few times a week, field notes on what's actually working when AI meets ecommerce ops. Plus first look at new prompts and tools as they go live.