Backend and generic keywords
The five generic_keywords fields — how Amazon indexes them, byte budgets, and what NOT to put there.
Backend keywords are the hidden indexing layer of your listing — words shoppers never see but Amazon uses to match your product to search queries. They live in five flat file columns (generic_keywords_1 through generic_keywords_5), shown in Seller Central UI as "Search Terms."
They exist for one purpose: capturing searches that don't appear in your title or bullets because the vocabulary doesn't fit, the format doesn't scan well in copy, or the term is a synonym/misspelling your front-end copy can't accommodate.
The five fields
Amazon provides five generic_keywords columns in the Category Listings flat file. While they appear as five separate fields, Amazon concatenates them into a single searchable index with one shared budget: 249 bytes total across all five fields.
Many sellers fill each field independently to ~200 characters and wonder why they're not seeing results — they've exceeded the budget by 900+ bytes and Amazon silently dropped everything past byte 249.
A practical organization system is to assign each field a distinct content purpose:
| Field | Purpose | Example content |
|---|---|---|
| generic_keywords_1 | High-volume synonyms | Product-type alternates, form-factor variants |
| generic_keywords_2 | Problem-solution terms | Skin concerns, use-case intent, outcome terms |
| generic_keywords_3 | Ingredient and scientific names | Active compounds, consumer-friendly ingredient names |
| generic_keywords_4 | Misspellings and spelling variants | Common typos, k-beauty vs kbeauty, British spellings |
| generic_keywords_5 | Spanish translations and demographic modifiers | Audience terms, life-stage qualifiers, Spanish keywords |
Total: target 249 bytes across all five.
Indexing rules
Amazon tokenizes on whitespace. Every space-separated word is indexed as an independent token. This means:
- No commas, semicolons, or other punctuation. They're not stripped — they become part of the token, breaking indexing.
- Plurals and singulars share a token. "pad" covers "pads" and vice versa. Include one form, not both.
- Stop words don't index. "a," "an," "the," "for," "with," "of" — Amazon's algorithm handles connectors. Don't waste bytes on them.
- Hyphens are inconsistently handled. The unhyphenated version covers more variants: "nonstick" covers "non-stick," "non stick," and "nonstick." When in doubt, include both the hyphenated and unhyphenated forms if they're meaningfully different in search behavior (e.g., "kbeauty" and "k-beauty").
- Duplicates waste bytes. Each unique word should appear exactly once across all five fields.
- Words already in your title or bullets are already indexed. Do not repeat them — those bytes are wasted.
What to include
Think in rationale categories, ordered by priority when you're tight on budget:
Synonyms — Alternate words shoppers use for the same product or action. If your title says "exfoliating toner pads," your backend should have "resurfacing," "peeling," "scrub" — the vocabulary of shoppers who don't use your exact terminology.
Product-form alternates — Different names for the physical format. "Wipes," "disc," "round," "towelette" — the same pre-soaked pad product is searched dozens of ways depending on the shopper's mental model.
Ingredient alternates and scientific names — Active compounds, consumer-friendly names, and derivative terms for ingredients that aren't in your front-end copy. If your title says "BHA," include "salicylic acid" in backend. If your bullets mention "Panthenol," include "vitamin b5" — the name ingredient-aware shoppers actually search.
Spanish translations — A significant share of US Amazon shoppers search in Spanish. Almost no competing sellers optimize for this. "Exfoliante," "tonico facial," "almohadillas," "cuidado piel" — these terms face almost no competition for the same shopper intent.
Use-case and problem-solution terms — The problem the shopper is trying to solve rather than the product attribute. "Blackhead," "clogged pores," "dull skin," "dark spot" — these connect your product to shopper goals rather than product descriptions, which is how COSMO and Rufus match intent.
Misspellings and spelling variants — Common typos and regional spellings Amazon's stemmer doesn't always catch. "Exfoliater," "moisturising" (British spelling), "kbeauty" vs "k-beauty." Low byte cost, meaningful incremental coverage.
Occasion and context terms — When or why someone buys: "gift," "self care," "morning routine," "after cleansing." These capture purchase-intent modifiers that don't fit naturally in copy.
Demographic modifiers — Audience qualifiers not in your title: life-stage terms, skin-type variants, gender qualifiers. "Perimenopause," "pregnancy-safe," "teen" — high-conversion, low-competition.
What to exclude
Any word already in your title or bullets. Amazon indexes front-end copy independently. Duplicating in backend wastes bytes without adding indexing.
Competitor brand names. Not even "like [Brand]." Using competitor brand names in backend keywords is a policy violation and a suppression risk.
Subjective claims. "Best," "top-rated," "premium," "amazing" — these don't index. Amazon's algorithm treats them as noise.
Promotional language. "Sale," "new," "on sale," "limited offer," "free shipping." Prohibited and not indexed.
ASINs and UPCs. Product identifiers are not search terms.
Stop words. "A," "an," "the," "for," "with," "of." Save the bytes.
Anything you can't substantiate. If your product isn't actually vegan, "vegan" in backend is a relevance violation that erodes conversion when those shoppers land on your page.
How RankASIN surfaces backend gaps
RankASIN's classification engine analyzes your harvested keywords and marks terms as opportunity — keywords appearing in your search report data but not yet converting for this ASIN. Many opportunity keywords aren't in your title or bullets. When they're not, backend is the logical home.
Review the Opportunity panel in the listing editor sidebar. Any tier_2 or tier_3 opportunity keyword that doesn't appear in your front-end copy is a backend keyword candidate.
After uploading, run an index check: search Amazon for <your ASIN> <keyword>. If your product appears, it's indexed. Wait 24-48 hours after any backend update before testing — indexing isn't instant.
Byte vs character budget
For standard ASCII (English letters, numbers, spaces), 1 character = 1 byte. The 249-byte limit and 249-character limit are the same.
The distinction matters for:
- Accented characters (á, é, ñ, ö) — 2 bytes each in UTF-8. Spanish keywords with accents cost double.
- Symbols (€, £, ©) — 2-3 bytes each.
- Emoji — 4 bytes each. Never use emoji in backend keywords.
Measure in bytes, not characters, before submitting. Exceeding 249 bytes by even 1 byte silently de-indexes the entire field — Seller Central shows no error and the keywords appear to be saved.
The Seller Central UI sometimes displays a 500-character limit for the Search Terms fields. Do not trust it. The indexing engine enforces 249 bytes. Filling to 500 characters means roughly half your keywords are silently ignored.