What backend search terms actually are
The backend search terms field is the one part of your listing shoppers never see. It sits in Seller Central under the Keywords tab, away from the title, bullets and description that face the customer. Its job is to catch search terms that are true and relevant to your product but would look clumsy or repetitive if you forced them into visible copy.
That means synonyms, alternate spellings and the phrasing shoppers actually type when your front-end copy doesn't use it naturally. If you sell a "stainless steel water bottle", a shopper searching "steel flask" or "metal drink bottle" wants the same product in different words. Your title and bullets should read naturally for a person, so those alternate phrasings go in the backend field.
Most sellers never got a clear explanation of how the field works. They either leave it half empty or pack it with everything from competitor brand names to unrelated attributes, hoping volume beats precision. Neither earns anything. Amazon's indexing is far more selective and far more literal than most sellers assume.
Where to find and edit the field
In Seller Central, go to Inventory, then Manage All Inventory, find the product and choose Edit from the menu on the right. Open the Keywords tab and look for the field labelled Search Terms. If you manage listings through inventory files, the same field appears as generic keywords in the category template. Each marketplace has its own copy of the field, so a listing on amazon.com and the same product on amazon.co.uk or amazon.de need their terms written separately, in the local language and spelling.
The real limit is bytes, and the 250-character figure is a myth
Ask ten sellers what the backend search terms limit is and at least seven will say 250 characters. The number has been repeated so often that it has become folklore, and it is wrong in a way that costs sellers indexing.
The limit is 249 bytes per field. For plain English text using standard letters, one character is usually one byte, so the confusion rarely hurts US sellers writing in English. It becomes a real problem once you add accented characters, non-Latin scripts, or text copied with invisible formatting characters from a word processor, any of which can take two or more bytes per character.
Keyword lists copied out of a spreadsheet or a Google Doc often bring along smart quotes, long dashes or non-breaking spaces. They look identical to normal characters but take extra bytes and can silently truncate the field. If your backend terms have ever looked cut off in a way that doesn't match your character count, this is almost always the reason.
When you hit the byte limit, Amazon doesn't stop indexing early and leave the rest visible for your reference. The field truncates where the byte count runs out, and anything past that point is gone. The last keyword you typed, the one you thought was safely tucked in at the end, may not even be saved, never mind indexed.
The fix is simple. Write backend terms in plain text, paste them into a plain text editor first if you're moving them from somewhere else, and keep the field comfortably under 249 bytes. Treat the limit as a ceiling, not a target to fill to the last byte.
How to count bytes
Character counters in word processors count characters, which is the wrong measure. Use a byte counter instead. Plenty of free online tools report UTF-8 byte length, and if anyone on your team writes code, a one-line script that measures the encoded length of the text does the same job. Some examples of the difference: the plain word "bottle" is 6 characters and 6 bytes, while "café crème" is 10 characters but 12 bytes, because each accented letter takes two. A phrase wrapped in curly quotes, like “steel” bottle, is 14 characters and 18 bytes, because each curly quote takes three. Sellers writing for the German, French, Spanish or Italian stores run into this most, since accented letters are common in normal search terms there.
Words Amazon silently ignores or penalizes
Amazon's indexing filters some kinds of words out of the backend field entirely. Putting them there does nothing except use byte budget you could spend elsewhere.
- Your own brand name and ASIN. These are already indexed automatically from the brand field and product identifiers, so repeating them adds nothing.
- Competitor brand names. As well as being useless for indexing, they cross into trademark territory and are the most common trigger for a listing suppression tied to an intellectual property complaint. Don't use them, even if you've seen competitors do it to you.
- Subjective claims such as "best", "amazing", "top quality" or "guaranteed". This is marketing language that nobody types as a search, and Amazon's system treats it as noise.
- Temporary claims. Anything tied to a sale, a discount percentage or a promotional window doesn't belong in a field meant to describe the product itself.
- Words already in your title, bullets or description. Amazon's search algorithm already indexes your front-end copy, so repeating a title word in backend terms doesn't boost its weight. It just takes up space you could use for a new phrase.
All of these fail for the same reason. The backend field exists to add coverage, and Amazon already knows everything in this list. Every byte spent restating something indexed elsewhere is a byte that could have caught a search you're currently invisible for.
Punctuation falls into the same category. Words in the field are separated by spaces, and since Amazon indexes individual words, commas and semicolons between them add bytes without adding anything a shopper can search for.
Duplicate keywords are wasting space you don't have
This is the most common mistake we find when auditing new client accounts, and it's almost never deliberate. Sellers build their keyword list over months, adding a few terms at a time, and never review the whole field as one piece of text.
The result looks something like "steel bottle metal flask insulated bottle steel water bottle metal water flask", where "steel", "bottle", "metal", "flask" and "water" each appear two or three times in different combinations. Every repeat is wasted. Amazon's search algorithm indexes individual words within the field, not only exact phrases, so listing every permutation of the same five words gives you no more coverage than listing each word once.
Write out the full backend field, then go through it and make sure no word appears more than once. The only exception is a word that's part of a genuinely distinct multi-word phrase, like "gift for her" and "gift for him" both containing "gift". If a word repeats without a new phrase around it, delete the second instance and use the space for a term you aren't covering at all.
A quick way to catch this is to paste your backend terms into any word frequency counter and look for words that appear more than once. In most audits we run, this one fix frees up 15 to 30 percent of the field, which is often enough room for three or four new search terms that were missing entirely.
Before and after: an example
Take the bottle example above. As written, it uses 78 bytes and covers only six distinct words. Deduplicated to "steel metal water bottle flask insulated", it uses 40 bytes and covers exactly the same words. That saves 38 bytes. Add "vacuum thermos canteen gym hiking office leakproof" and the whole field comes to 91 bytes, only 13 more than the cluttered original, while covering 13 distinct words instead of six, including synonyms and use cases the original never reached. It still leaves well over half of the limit free for more terms. These words are illustrative; your own replacements should come from search term reports and keyword research for your product.
A practical framework for what belongs there
Once you've stripped out what doesn't work, decide what to keep using a simple order of priority:
- True synonyms and alternate names: the different words shoppers use for exactly the same product. Think "sneakers" and "trainers", "flask" and "bottle", or regional spellings like "colour" if you sell internationally.
- Use cases and contexts your visible copy doesn't mention. If your title focuses on what the product does but shoppers often search by occasion or setting, like "gym", "office" or "travel", that context goes here.
- Common misspellings with real search volume. Include one only if you can confirm actual search traffic behind it, not just any typo you can imagine.
- Abbreviations and acronyms. If your category has a common shorthand you haven't spelled out in visible copy, include it here.
Generic adjectives are missing from this list, along with category words already covered by your browse node and anything unrelated to what the product is. Relevance matters in a very literal sense. Amazon's policy explicitly prohibits keywords unrelated to the product, and irrelevant terms that do get indexed bring traffic that leaves straight away. That drags down your conversion rate and can hurt your organic rank on the terms that matter.
Your best source for new terms is your own advertising data. In the Amazon Ads console, the search term report for your automatic and broad match Sponsored Products campaigns shows the exact queries shoppers typed before clicking and buying. Any converting query that uses words absent from your title, bullets and backend field is a strong candidate.
How to verify what's actually indexed
Writing a good backend field is half the job. The other half is confirming that Amazon indexed what you wrote, and no dashboard simply lists your indexed keywords for you to check.
The most reliable manual method is a live search test. Search your exact keyword phrase on Amazon, then use the storefront's own filters to narrow the results to your brand or seller. If your listing appears in that filtered set, you're indexed for the term. If it doesn't appear even with the brand filter applied, either the term isn't indexed or your organic rank for it is so weak that Amazon isn't showing you in the results it displays.
The manual method doesn't scale past a handful of terms. For a full field, the Search Query Performance report in Brand Analytics is a better diagnostic, because it shows real impression and click data for actual search terms. That tells you whether you're indexed and also whether the indexing is producing any visibility.
Allow a short delay after editing before you test. Changes to listing attributes can take a while to process, so a term that doesn't show up an hour after you saved it may simply not have updated yet. If it still doesn't appear a few days later, check that the edit saved by reopening the Keywords tab and confirming the full text is there.
A backend terms audit you can run today
To check your own listings now, use the same sequence we run during a client account audit:
- Pull the current backend search terms for your top five listings by revenue.
- Paste each field into a plain text editor and check for duplicate words using the one-word, one-instance rule above.
- Remove anything that repeats a word already present in your title or bullets.
- Remove any competitor brand names immediately, however long they've been there without causing trouble.
- Confirm you're under 249 bytes, measured in bytes and not characters, especially if any of the text was ever copied from a spreadsheet or document.
- Use the freed-up space for two or three genuine synonym or use-case terms you weren't covering before.
Most listings we audit have never had this cleanup since the product launched, so the backend field has been collecting years of copy-paste debris while doing a fraction of the indexing work it could do. Thirty minutes per listing is usually enough to fix it.
Repeat the audit whenever you rewrite a title or bullets. Words you've just added to visible copy can come out of the backend field, and words you removed from the title may need to move into it.
Frequently Asked Questions
It's 249 bytes per field. The 250-character figure is a myth. For plain English text a character and a byte are usually the same thing, so the myth rarely hurts US sellers writing in English. It becomes a real problem once you add accented characters, non-Latin scripts or invisible formatting characters, which can take two or more bytes each.
Your field almost certainly ran past 249 bytes. Keyword lists copied from a spreadsheet or Google Doc bring along smart quotes, long dashes and non-breaking spaces that look like normal characters but take extra bytes. Once the count runs out the field truncates, and anything past that point is gone completely, which is worse than merely going unindexed.
Never. Competitor brand names do nothing for indexing and they cross into trademark territory, which makes them the most common trigger for a listing suppression tied to an intellectual property complaint. Remove them from your listings today, however long they've been there without causing trouble.
No. Amazon already indexes your title, bullets and description, so repeating those words in backend terms adds no weight and only burns byte budget. The same goes for your own brand name and ASIN, which are indexed automatically from your brand field and product identifiers. Spend that space on synonyms and use cases you don't cover anywhere else.