Amazon backend keywords: the 250-byte field
Amazon caps backend search terms at 250 bytes, not characters. Why that costs non-English sellers half the field, and how to use every byte.
- Amazon
- Backend keywords
- Listing optimisation
- SEO
Amazon's backend search terms field is capped at 250 bytes. Almost every guide on the internet calls it 250 characters. For a seller writing in plain English those are the same number, which is why the mistake has survived this long. For everyone else they are not, and the gap is where the field quietly stops working.
This is the single most misunderstood field in an Amazon listing, and it is also the cheapest ranking real estate on the platform — a hidden field with no character-count pressure from shoppers, no design constraints, and no risk of reading badly. Most sellers either leave it empty, paste their title into it, or overflow it without ever being told.
Bytes are not characters
Text is stored as UTF-8. In UTF-8, different characters cost different amounts of storage:
| Characters | Example | Bytes each | Fits in 250 bytes |
| --- | --- | --- | --- |
| English letters, digits, spaces | wireless mouse | 1 | 250 |
| Accented Latin, Greek, Cyrillic | küche · café · ñ · ß | 2 | 125 |
| CJK, Devanagari, Arabic, Thai | キーワード · 键盘 · कीबोर्ड | 3 | 83 |
| Emoji | 🔥 · ⭐ | 4 | 62 |
Read the last column again, because it is the part that matters. A seller writing in English gets 250 characters of backend search terms. A seller writing in Japanese gets 83 — a third as much room, for the same job, on the same marketplace, against the same limit.
A German seller filling the field to "250 characters" with a normal character counter can be forty bytes over without any warning appearing anywhere in Seller Central. The text still looks saved. It just is not fully indexed.
We built a backend keyword byte counter specifically because no other free tool counts this correctly. It runs in your browser, needs no account, and shows which terms would be discarded.
What happens when you go over
Amazon does not reject the field with an error you would notice. It reads up to the limit and discards the rest. Seller Central redisplays exactly what you typed, so nothing looks wrong.
The only visible symptom is that your listing never appears for the terms at the end of the field — which is indistinguishable from those terms simply being competitive. That is why the problem persists: the feedback loop that would tell you is missing.
In some categories an over-length value causes the entire field to be rejected rather than trimmed. Not the overflow. All of it.
Order the field so the terms you most need indexed sit at the front. If it is truncated, the tail is what you lose.
Four ways sellers waste the field
1. Commas
Amazon tokenises on whitespace. Commas and semicolons are treated as separators and thrown away — but they occupy a byte each before they are discarded. Thirty comma-separated terms means thirty bytes spent on punctuation the search engine never sees. That is more than a tenth of the budget.
Use single spaces. Nothing else.
2. Repetition
A term is indexed once, however many times it appears. Repeating a keyword does not raise relevance for it. That practice is inherited from search engines that worked that way two decades ago; here it only shortens the field.
3. Stop words
Amazon's own style guides list a, an, and, by, for, of, the and
with as terms to leave out. They are discarded before matching, so each one is
a byte cost with a guaranteed return of zero.
This applies only to the backend field. In your title and bullets they keep the copy readable for humans, which matters more.
4. Repeating your own title
Your title, bullets and description are all indexed already. A term that appears in them is searchable without spending backend bytes on it.
The backend field earns its keep on what the visible copy has no room for: synonyms, regional names, the misspellings buyers genuinely type, and use cases you would never put in a title. Duplicating your title is not penalised — it just spends a scarce field on coverage you already have.
What to actually put in it
Think of it as the list of things a buyer might type that you could not reasonably work into visible copy:
- Synonyms and alternative names.
pramandstroller.torchandflashlight.courgetteandzucchini. - Misspellings people actually type. Not every typo — the common ones. And
genuinely different spellings across English variants:
organiserandorganizerare distinct queries. - Use cases and occasions.
gift for teacher,camping,dorm room. - Material and attribute terms that did not fit the title.
- Regional terminology for the country marketplace you are listing on.
What to leave out, because Amazon's policies prohibit it or it wastes space:
- Competitor brand names
- ASINs
- Subjective claims —
best,cheapest,top-rated - Plurals of terms you already included; Amazon handles common singular and plural forms from one entry
Does word order matter?
Within the field, no — order does not affect whether a term matches. Amazon also combines your backend terms with words from the rest of the listing to form phrase matches, so a term here can pair with a term in your title.
What order does affect is what survives truncation. If you are anywhere near the limit, the terms at the end are the ones dropped. Front-load accordingly.
A worked example
Here is a field as many sellers write it:
wireless mouse, wireless mouse for gaming, ergonomic mouse, mouse for the office
62 bytes, and most of it is doing nothing. wireless and mouse each appear
twice. for and the are stop words. Three commas are three wasted bytes. And
if the title already says "Wireless Ergonomic Mouse", every one of those terms is
already indexed.
The same coverage, rewritten:
gaming office silent bluetooth usb-c travel laptop mice mousepad
Shorter, no duplicates, no stop words, no punctuation — and every term adds reach the visible copy did not already have.
Where this fits in the wider picture
Backend search terms are one field. They will not rescue a listing whose title is wrong or whose main image gets it suppressed. For how the fields work together under Amazon's current ranking model, see our guide to writing Amazon listings that rank in 2026 — particularly the sections on COSMO and Rufus, which changed what the surrounding copy needs to do.
If you are writing the same product for several marketplaces, note that this field is Amazon-specific. Etsy uses thirteen tags, eBay uses item specifics, and neither behaves like a backend keyword field.
Check your own field. The backend keyword byte counter is free, runs entirely in your browser and stores nothing. Paste your live search terms in and see how many bytes you are actually using — and how many you are spending on commas.
If you would rather not hand-build the field at all, the Amazon listing generator produces it byte-counted and de-duplicated against your own title and bullets, alongside the rest of the listing. One credit, free plan available.