A keyword can be added to a listing and still fail to index for four broad reasons: the edit never actually landed on the live page, the words landed somewhere Amazon's search index does not reliably draw on, Amazon refused the term, or the listing is not eligible to be returned in search at all. Work through them in that order, because the cheapest checks rule out the most common causes — confirm the public detail page really contains the words, confirm the listing is buyable and not suppressed, allow for recrawl time before concluding anything, and only then suspect the term itself. In practice most cases turn out to be a reverted save or an ineligible listing rather than anything to do with keywords.
First rule out a false negative in the test
Searching Amazon for the exact phrase followed by your ASIN is the right test, but it has failure modes that look identical to a missing index entry, and they are worth eliminating before you touch the listing again. Run it in the marketplace you are asking about rather than on amazon.com by habit, from an unfiltered search bar with no department selected, and against the child ASIN that actually holds the copy rather than the parent — a variation family's parent is not what shoppers buy and not what the words are attached to.
The bigger source of false negatives is eligibility. A listing with no purchasable offer can drop out of search results entirely, and a suppressed or search-suppressed listing behaves the same way while its detail page still loads perfectly for anyone holding the link. In both cases the test returns nothing although the words are indexed, and the seller spends a week rewriting a field that was never the problem.
The cheap way to separate the two is a control. Pick a distinctive phrase you already know is indexed — an unusual string from your own existing title — and run the same ASIN test on it. If the control fails too, the problem is the listing's eligibility or your test method, not the new keyword, and nothing you do to the copy will change the result until that is fixed.
Confirm the edit actually landed, and stayed
Seller Central confirming a submission is not evidence that the change is live. Read the public detail page as a shopper sees it and find the words with your own eyes, in the rendered title and bullets, rather than re-reading the form you typed them into. Amazon can also accept a submission and drop individual attributes within it, so a clean submission report is compatible with the one field you cared about never changing.
Then check it again a few days later, because the more interesting failure is an edit that lands and is then reverted. A single ASIN can be fed from several sources at once — the Manage Inventory interface, a category flat file, an integration or repricer pushing listing data, A+ and brand content submissions, and on a shared ASIN the contributions of other sellers. Where sources conflict, one wins, and an edit made from a lower-priority source can quietly disappear hours or days afterwards. If the old copy is back when you look again, the problem is not that Amazon ignored your keyword; it is that something else is overwriting it, and you need to change it at that source instead.
Backend search terms have no rendered page to read, which makes them the hardest field to verify and the easiest to be wrong about. There is no shopper-visible proof they saved, the field is capped by a size limit measured in bytes rather than characters, and words past that cap do nothing for you whether or not you were warned. Check the limit the field itself shows for your marketplace rather than a figure quoted in a blog post, keep comfortably inside it, and treat the index test as the only real confirmation that anything in there worked.
Where the words sit changes whether they index
The fields that reliably feed Amazon's search index are the title, the bullets, the backend search terms and the structured attributes you fill in per category. How much the description contributes, and whether A+ content text is indexed for search at all, are genuinely contested — accounts differ and the behaviour has changed over time. The practical consequence is not to argue about it but to stop depending on it: if a phrase you need exists only in the description or only inside A+ modules, move it somewhere you can verify and re-test, rather than assuming it is working.
It also matters that Amazon can match a query from words present across different fields, so a shopper searching a four-word phrase may be returned a listing that never contains those four words as a block. The corollary catches people out: all the words being indexed is not the same thing as the full phrase returning your product. A long phrase that fails the test while every uncommon word in it passes the test individually is not an indexing problem at all — the words are in the index and Amazon is making a relevance judgement, which is a ranking question answered by click-through and conversion rather than by editing another field.
That test sequence is worth running before anything else, because it is free and it splits the two problems cleanly. Test the full phrase with your ASIN, then test each distinctive single word in it with your ASIN. All words pass and the phrase fails: relevance. One word fails on its own: that word is the thing to investigate. Repeating a word you have already added cannot help either way — if once did not index it, twice will not, and the duplication costs you space and readability.
Terms Amazon will not index for you
Some terms are refused rather than missed, and the giveaway is that the submission looked completely successful. Competitor brand names and trademarks are the clearest case: using them is against Amazon's policy, it exposes the listing to complaints and takedowns, and it does not buy you the ranking the seller was hoping for. There is no technique that indexes you for another company's brand name, and the honest answer when a client asks is that the phrase is not available to them.
The looser category is wording Amazon treats as a prohibited or restricted claim — medical and health language, safety and compliance claims, and certain superlatives — which varies by category and by marketplace. The published guidance does not amount to a complete word list, the lists circulating online are unreliable and go stale, and the restriction can apply to one marketplace and not another. So do not try to clear a phrase against a list; test it. If a phrase keeps failing and one word in it is the plausible suspect, test that word alone with your ASIN and let the result decide.
When a term is genuinely unavailable, the useful move is to find the query shoppers actually type that you are allowed to own. Your own search term reports are the best source for that, because they show real queries that already produced clicks on your product rather than phrases you would like to rank for.
How long to wait, and when to escalate
A recrawl after a text edit is not instant and it is not reliably same-day. Treat a few days as normal before concluding that a term failed, and re-run the identical test rather than editing again in the meantime. The discipline that makes this work is changing one field at a time: three edits made while waiting leaves you unable to say which one indexed the phrase, which means you have paid for the experiment and learned nothing you can reuse on the next product. Record the previous value verbatim with the date before each edit, so that reverting is available to you and so that a shift in impressions weeks later can be attributed rather than guessed at.
There is a defensible point to open a case with Seller Support, and it is after the checks above rather than instead of them. By then you can state that the words are on the live detail page, that the page has not been overwritten, that the listing is buyable and not suppressed, that a control phrase indexes correctly, that each individual word passes the ASIN test, and that the term is not a brand name or a restricted claim. A case carrying the ASIN, the exact phrase, the date of the edit and the test you ran gets a different reply from one that says a keyword is not working.
One last thing worth deciding before you spend more time on it: indexing only buys eligibility. Once a term indexes, position is decided by click-through and conversion, so a phrase you force into the index and then convert badly on delivers impressions, a falling click-through rate and no sales. Some of the keywords sellers fight hardest to index are ones they would be better off not appearing for, and the question of whether you want the term is cheaper to answer than the question of why it will not index.
Or read how the ongoing work runs before you book: Amazon SEO and listing optimization.