| name | keyword-research |
| description | Amazon keyword research and keyword mapping for product listings. Use when the user wants to find keywords, build a keyword strategy, analyze search terms, identify keyword gaps, map keywords to listing elements, or optimize keyword coverage. Also use when the user provides a Helium 10, Jungle Scout, Brand Analytics, or Search Term Report export and wants analysis.
|
Amazon Keyword Research
Builds structured keyword maps for Amazon product listings, organizing terms
by priority tier and mapping them to specific listing elements (title, bullets,
description, backend).
Prerequisites
Load references/backend-keywords-guide.md — used for backend string
construction rules, byte counting, and what to include or exclude.
Load references/amazon-style-guide.md — used when checking keyword
compliance (prohibited terms, stop words, formatting rules).
If brand-context.md exists in the project root or .agents/ directory,
load it for shared target audience guidance and past keyword learnings.
If products/[ASIN]/product-context.md exists, load it for product-specific
details. Prefer User Confirmed Data, then Amazon Observed Data, then AI Inferences.
Resolve product references by SKU or ASIN using the product context file's
SKU: and ASIN: fields.
If multiple product folders exist and the user has not specified a product, ask
which SKU or ASIN to use before building the keyword map.
If the user is working on a variation family, load the parent product context
plus the relevant child product contexts before finalizing keyword placement or
backend terms. Map shape, count, capacity, and configuration terms to the child
they actually describe rather than inheriting them across the family by default.
How This Skill Works
This skill does NOT have access to live Amazon search volume data. It works by:
- Leveraging product knowledge and category expertise to generate
comprehensive keyword candidates based on what buyers actually search
- Parsing data the user provides — CSV exports from Helium 10, Jungle Scout,
Amazon Brand Analytics, or Search Term Reports
- Structuring keywords into actionable tiers that map directly to listing elements
- Identifying gaps between current listing content and potential keyword coverage
If the user has access to keyword tools, recommend they pull data and bring it
back for analysis. If not, generate research-informed suggestions based on the
product description and category knowledge.
Keyword Tier Framework
Tier 1: Primary Keywords (1-3 terms)
- Highest relevance, highest volume
- MUST appear in the product title (within first 80 characters)
- These define what the product IS
- Example: "glass vacuum storage container"
Tier 2: Secondary Keywords (5-10 terms)
- High relevance, moderate-to-high volume
- Should appear in bullet points and/or description
- These add specificity and capture related searches
- Example: "airtight food container", "vacuum seal container", "glass meal prep container"
Tier 3: Long-Tail Keywords (10-25 terms)
- Moderate relevance, lower volume, higher conversion intent
- Distribute across bullets, description, and backend
- These capture specific buyer scenarios and questions
- Example: "glass container for meal prep", "vacuum container for berries",
"airtight container for pantry storage"
Tier 4: Backend-Only Keywords (10-30 terms)
- Synonyms, translations, misspellings, use-case terms
- Do NOT appear in visible listing copy — backend search terms only
- These expand reach without cluttering customer-facing content
- Example: "tupperware recipiente vidrio contenedor hermético cocina leftovers marinate"
Tier 5: Rufus/Intent Keywords (5-15 terms)
- Conversational phrases and question-based queries
- Integrate naturally into bullets and description
- These target how Rufus-era shoppers discover products
- Example: "keeps food fresh longer", "safe for microwave and freezer",
"best container for meal prep", "won't stain or absorb odors"
Keyword Research Process
Step 1: Seed Keyword Generation
Ask the user for:
- Product name and type
- Seller SKU or ASIN (if known)
- 3-5 words they'd use to describe the product
- What problem it solves
- Who the target buyer is
- 2-3 competitor products (names, SKUs, or ASINs)
From these inputs, generate an initial seed list of 10-15 terms.
Step 2: Keyword Expansion
For each seed keyword, expand using these methods:
Synonym expansion:
- "container" → "canister, jar, keeper, holder, organizer, vessel"
- "vacuum" → "airtight, sealed, hermetic, airless"
Modifier expansion:
- Material: "glass", "borosilicate", "BPA-free"
- Size: "small", "large", "32 oz", "liter"
- Feature: "stackable", "leakproof", "microwave safe"
- Use case: "meal prep", "lunch", "leftover", "pantry"
- Audience: "for kids", "for office", "for camping"
Question-based expansion (Rufus-ready):
- "how to keep food fresh longer"
- "best container for meal prep"
- "glass vs plastic food storage"
- "does vacuum seal container work"
Spanish translations (US marketplace):
- Product terms: "contenedor de vidrio", "recipiente al vacío"
- Use-case terms: "almacenamiento de alimentos", "preparación de comidas"
Step 3: Competitive Gap Analysis
If the user provides competitor ASINs, competitor SKUs, or listing content:
- Extract all keywords from competitor titles, bullets, descriptions
- Identify terms they target that the user does NOT
- Flag high-value gaps (relevant terms competitors rank for that user misses)
- Note terms competitors use that may not be compliant (and should be avoided)
Step 4: Keyword Mapping
Map every keyword to a specific listing element. Example (illustrative — the
final output should follow the per-tier table structure in templates/keyword-map.md):
| Keyword | Tier | Placement | Status |
|----------------------------------|------|------------------|-----------|
| glass vacuum storage container | 1 | Title (pos 1-5) | Primary |
| airtight food container | 2 | Bullet 1 | Include |
| meal prep container glass | 2 | Bullet 5 | Include |
| vacuum seal food keeper | 3 | Description | Include |
| contenedor vidrio hermético | 4 | Backend | Include |
| keeps food fresh longer | 5 | Bullet 1 natural | Rufus |
Step 5: Backend String Construction
After mapping, take all Tier 4 keywords and build the backend search terms string:
- Remove any words already present in Tier 1-3 placements
- Remove stop words
- Order by estimated value (highest first)
- Calculate byte count
- Trim to under 235 bytes (safety buffer for 249 limit)
For variation families, build backend strings per child ASIN. Add only the
shape/configuration terms that apply to that child and remove family terms that
would misdescribe the variant.
See references/backend-keywords-guide.md for detailed formatting rules.
Working with User-Provided Data
If the user provides a Search Term Report (from Amazon Advertising):
- Sort by conversions (descending)
- Identify converting search terms NOT in current listing
- Flag high-spend, zero-conversion terms to consider excluding from ads
- Recommend adding converting terms to organic listing content
If the user provides Helium 10 / Jungle Scout export:
- Sort by search volume and relevance
- Cross-reference against current listing content
- Identify high-volume terms missing from listing
- Flag terms with high volume but low relevance (avoid)
If the user provides Brand Analytics data:
- Analyze click share and conversion share
- Identify terms where the product has impressions but low click share (title needs work)
- Identify terms where clicks are high but conversion is low (bullets/description need work)
Output Format
Load templates/keyword-map.md and populate every tier table and the backend
string section with research findings. Do not reproduce the template inline —
fill it out and present the completed keyword map to the user.