How to Use the Amazon Search Terms Report
Every week shoppers type things into Amazon search that nobody on your team would have thought to bid on, and some share of your budget goes to queries that were never going to convert. The search terms report is where Amazon tells you which is which, from its own data rather than a third-party estimate.
Two things make it worth a standing slot on the calendar rather than an occasional look. The raw query data expires. Amazon's advertising API team has said that while a report can be requested up to 95 days in the past, only the last 65 days reliably return data, so a quarter you never pulled is a quarter you cannot reconstruct. And two different Amazon reports carry nearly the same name, so plenty of teams follow a guide written about one and end up opening the other.
Quick answer: what the report shows and what to do with it
The Amazon search terms report is the advertising report that lists the exact shopper queries that triggered your Sponsored Products and Sponsored Brands ads, with impressions, clicks, spend, orders and sales for each one, alongside the keyword or product target you bid on and its match type. You pull it from the ad console under Measurement and Reporting, and it costs nothing. It feeds two jobs: harvesting converting queries into targets of their own, and negating queries that spend without converting. A second Amazon report, the Brand Analytics Top Search Terms dashboard, is marketplace-wide rather than account-specific and shows search frequency rank plus the top three clicked products for any term, whether or not you advertise on it.
Two Amazon reports share the "search terms" name and answer different questions
This is the single most common wrong turn on this topic, and it costs a real afternoon.
The advertising search term report shows the queries that triggered your ads
This one lives in the Amazon Ads console, inside Campaign Manager. It is account-specific and spend-specific. Every row is a query a shopper typed that resulted in your ad serving, paired with the keyword or product target that matched it, the match type that caught it, and the money that followed. It is available for Sponsored Products and Sponsored Brands. Sponsored Display is not keyword-driven, so there is no equivalent query column there.
In automatic campaigns and product-attribute targeting, the customer search term column will sometimes hold an ASIN rather than words. That means your ad served on that product's detail page rather than against a typed query, and it is a targeting signal, not a keyword.
Brand Analytics Top Search Terms shows what the whole marketplace is searching
The Top Search Terms dashboard sits in Seller Central under Brand Analytics and is open to professional sellers enrolled in Brand Registry. It reports popular queries across the Amazon store with each term's search frequency rank, which is a relative ranking against every other term in that marketplace rather than an absolute volume, plus the three products that captured the most clicks and their click and conversion share.
It will tell you what shoppers in your category are asking for. It will not tell you where your budget went. Use it for research and category reads, and use the ad console report for spend decisions.
| Dimension | Search term report (Ads console) | Top Search Terms (Brand Analytics) |
|---|---|---|
| Scope | Only queries that triggered your ads | Marketplace-wide, whether or not you advertise |
| Core metrics | Impressions, clicks, spend, orders, sales, ACoS | Search frequency rank, top three clicked ASINs, click and conversion share |
| Who can open it | Any advertiser running Sponsored Products or Sponsored Brands | Professional sellers enrolled in Brand Registry |
| Where it lives | Campaign Manager, Measurement and Reporting | Seller Central, Brand Analytics |
| Best used for | Harvesting, negation, bid decisions | Category research, launch validation, listing vocabulary |
| What it cannot answer | What shoppers search when your ads do not serve | Where your spend went, or which term wasted it |
Everything below is about the ad console report, which is what most people mean when they say search terms report.
What is inside the advertising search term report
Each row pairs a customer search term with the target that caught it. The columns you will work from are impressions, clicks, click-through rate, cost per click, spend, 7-day orders, 7-day sales, ACoS and conversion rate, with campaign, ad group, targeting and match type as the grouping fields. Broad and phrase match produce the widest spread of queries, which is exactly why those campaigns generate the most useful rows and the most waste. Exact match produces fewer surprises. If you need a refresher on how the match types behave, that shapes what shows up here more than anything else you control.
Your numbers keep moving after you pull the file
Attributed orders and sales fill in across the attribution window, which for Sponsored Products in a Seller Central account is seven days from the click and for Sponsored Brands is fourteen. A report pulled this morning covering yesterday will understate conversions, and the terms at the bottom of your zero-order list will not all still be there in a week.
The practical consequence: do not make negation decisions on the last seven days of data. Pull a window that has had time to settle, or accept that you will negate a few terms that were about to convert.
Some queries never appear at all
Amazon does not publish every query. Terms with very low impression counts get filtered or aggregated out, so the long tail in your report is thinner than the long tail in reality. That matters when you are trying to reconcile spend line by line, and it means an absent term is not evidence that nobody searched it.
The 65-day clock is the constraint most teams discover too late
Search term data is a decaying asset, and this is the part that turns a nice-to-have review into a standing one.
Amazon's advertising API team has confirmed that a single report request covers a maximum of 31 days, and that while a request can reach up to 95 days into the past, only the last 65 days reliably return data. The console has similar practical limits. Nothing warns you when a week ages out.
Two ways to handle it. Pull and archive the report on a fixed cadence so you build your own history, which is what most tools in this category are quietly doing on your behalf. Or accept a rolling two-month memory and structure your reviews so no window goes unread. What does not work is planning to look at Q4 in February.
How to download the Amazon search terms report
- Log in to your Amazon Seller Central account.
- Open Campaign Manager.
- Click Measurement and Reporting, then select Sponsored ads reports.
- Choose the report category, set report type to Search term, and pick your timeframe.
- Set the delivery schedule and run the report.
Amazon delivers a raw CSV. Schedule it to recur rather than requesting it by hand each time, which also solves half the retention problem for free.
Where to find the Brand Analytics Top Search Terms dashboard
Seller Central, Brand Analytics, Top Search Terms. Brand Registry enrollment is the gate. If you want the query-level view of how your own ASINs perform against specific searches, the Search Query Performance dashboard sits alongside it and is the better companion to the ad console report.
Harvest, negate, optimize: the loop the report exists to feed
A report you read is worth very little. A report that changes three things in the account each week compounds. The full mechanics, including how to structure campaigns so harvested terms have somewhere to go, are in our Amazon search term report workflow guide. Each step in brief:
Harvest the queries that convert
Sort by orders, then by ACoS. Any query converting at or below your target ACoS inside a broad, phrase or automatic campaign is a candidate to promote into an exact-match target where you can bid it deliberately. Add it as a negative exact in the campaign that discovered it, so the two do not compete for the same impression.
Negate against your break-even click count, not a round number
The common rule is to negate anything with 30 or more clicks and zero orders. That is a convention, not a law, and it is wrong for most catalogs in one direction or the other. Do the arithmetic instead: divide your allowed cost per order by your average cost per click. At a $40 average order value and a 30% target ACoS, you can spend $12 to win an order; at a $1.20 CPC that is ten clicks. A term at 25 clicks and no orders has burned two orders' worth of budget, and it does not need a thirtieth click to prove the point. A term at 25 clicks in a category with a $0.20 CPC has barely been tested.
Where the query is simply irrelevant, negate on the first sign rather than waiting for a threshold. Our guide to negative keywords on Amazon covers the phrase-versus-exact choice, which is where most negation lists go wrong.
Feed the vocabulary back into the listing
Converting queries are not only bid targets. They are the words shoppers use for your product, which makes them the best available input for titles, bullets and backend keywords. Run the promising ones through the Listing Analyzer to see whether your detail page currently supports the term at all before you bid harder on it.
Where a careful weekly pass runs out
A skilled operator with a pivot table does this well. On one account, with a few thousand rows, a careful pass beats most automation, because the judgment calls are semantic: is "stainless steel water bottle 32 oz" a competitor's model number, a use case you do not serve, or your next best keyword? A person answers that in a second.
The break point is not the reading. It is the repetition. The same judgment call, applied the same way, across tens of thousands of rows, every Monday, on nine accounts, by whoever is on shift that week. That is where a standard that lives in one person's head starts producing different answers depending on who ran it, and where the review quietly slips to fortnightly and then to whenever there is time.
A general-purpose AI tool helps with the reading and struggles with the standard. It will classify a batch of queries quickly and accurately from a clean export. Ask it the same question next Monday and you may get a slightly different cut line, because nothing is holding the definition steady between runs. That is a fine tradeoff for exploratory work and a poor one for a decision you have to explain to a client.
| Task | Careful manual pass | General-purpose AI tool | Platform that executes |
|---|---|---|---|
| Judge an ambiguous query | Best in class. Context and category knowledge win here | Good, and fast, on a clean export | Good, and only as good as the standard you wrote |
| Cover 40,000 rows weekly | Not realistically, so the tail goes unread | Possible in batches, with export and prompt overhead | Runs on connected data, no export step |
| Apply the same cut line every week | Varies by who is on shift | Drifts between runs | Locked logic, same output from the same inputs |
| Act on the finding | Manual, in the console or a bulk sheet | Not without an integration | Approval-gated actions, if it manages the bids |
| Catch the thing your rules did not anticipate | Strong. A person notices the weird row | Sometimes, if you ask an open question | Weak by design. Codified logic does what it was told |
| Explain what changed and why | Depends on the notes someone kept | A narrated rationale, not a reproducible one | Change history against the logic that ran |
Where Qore fits
Qore is where you write down the search-term review you already do by hand, as logic you can read back, and have it run on a schedule against every account on your roster. Internally we call it the codified-workflow layer, which is a duller way of saying the same thing.
For this report specifically, that means describing the loop once: which window to pull, what counts as a converting query in your catalog, what your break-even click count is per campaign type, which competitor and model-number patterns to negate on sight, and what the output should look like. Qore assembles that into visible logic, and the same inputs produce the same output next Monday, because the definition is written down rather than recalled. Negations and bid changes execute through Qinetix ads automation, with approval as the default gate, so you see the batch before it ships. Once a workflow has earned your trust, you can let it run without the review step.
What it does not do. The automated action loop presumes Trellis is the bid manager of record; if your bids live somewhere else, Qore can still produce the harvest and negation lists, and someone on your side applies them. And Qore will not decide your standard for you. Q, the assistant that helps you shape a workflow, will push on the vague parts and suggest structure, but if your team has never agreed what counts as a wasteful term, writing it into a workflow surfaces that argument rather than settling it. Most teams find that useful and slightly uncomfortable. Qore is on a public waitlist as of July 2026.
If you are weighing this against wiring up an LLM yourself, our piece on AI for Amazon ads walks through what each approach gets you.
What to do with your next Amazon search terms report
Schedule it rather than requesting it by hand, so the 65-day window stops being a risk. Confirm you are in the ad console report and not the Brand Analytics dashboard. Work a settled date window rather than the last seven days. Calculate your break-even click count once per campaign type and write it down where the next person can find it. Then harvest, negate, and push the converting vocabulary back into the listing, and do it again next week.
If the weekly pass is the part that keeps slipping, that is the problem worth solving before you optimize anything else. Book a walkthrough and we will show you what the loop looks like running on a schedule.
Frequently Asked Questions
They are usually the same thing, and the confusion is real. Most people mean the Sponsored Products or Sponsored Brands search term report in the Amazon Ads console, which lists the queries that triggered your ads. A separate Brand Analytics dashboard called Top Search Terms shows marketplace-wide popularity data instead. Check which one a guide is describing before you follow its steps.
Weekly is the common cadence and it works for most catalogs, because the underlying data expires in roughly two months. Fast-moving or high-spend accounts benefit from every three or four days. Schedule the report to deliver automatically rather than pulling it by hand, since a skipped week cannot be recovered later.
Amazon's advertising API team has said requests can reach up to 95 days into the past but only the last 65 days reliably return data, and a single request covers a maximum of 31 days. In practice, treat two months as your usable memory. Anything older needs to have been archived by you or by a tool pulling on your behalf.
Amazon filters and aggregates very low-impression terms, so the long tail in the file is shorter than the long tail in reality. Data also lands with a lag and attributed orders continue to fill in across the seven-day Sponsored Products attribution window. A term that is absent today is not proof that nobody searched it.
Rather than a fixed number, divide your allowed cost per order by your average cost per click. If your target ACoS and average order value allow $12 per order and your CPC is $1.20, ten clicks with no order is already a full order's worth of spend. Terms that are plainly irrelevant to your product should be negated immediately, without waiting for a click threshold.
No. Sponsored Display targets audiences and products rather than typed queries, so there is no customer search term column to report. Search term reports are available for Sponsored Products and Sponsored Brands. For Sponsored Display you work from the targeting and placement reports instead.
Yes. Both the ad console search term report and the Brand Analytics Top Search Terms dashboard are free, though Brand Analytics requires a professional selling account enrolled in Brand Registry. Amazon delivers the ad report as a raw CSV, so the cost is in the processing rather than the access.
eCommerce News You'll Actually Use