Alexa for Shopping (formerly Rufus; Amazon renamed it on May 13, 2026) answers shoppers from your catalog data, reviews, community Q&A and the web, so the detail page is the dataset it talks about you from.
- Fix the catalog first: the product type sets which attributes exist, the browse node sets which shelf you compete on, and a wrong value becomes a wrong answer.
- A+ text, comparison charts and substantive reviews are the evidence that backs your claims. Text baked into images is a weak bet.
- Measure demand with Search Query Performance and total ad dependency with TACOS; ACOS alone hides ads buying sales you'd have had anyway. AMC answers the path-to-purchase questions.
- Structure ads so every dollar maps to one product (single-product ad groups), and pick Seller Central, Vendor Central or both knowing who controls price and the page.
Short answer: on Amazon in 2026, listing work and ads work are one operation. Alexa for Shopping (formerly Rufus), Amazon's shopping assistant, reads your catalog data, reviews and Q&A and answers shoppers in its own words. So the order of work is catalog first (product type, attributes, browse node), then evidence (A+ content and reviews), then measurement (Search Query Performance, TACOS, Amazon Marketing Cloud), then an ad structure that keeps spend attributable to one product. The account model, Seller Central or Vendor Central, decides how much of that you control.
- Alexa for Shopping
- Amazon's AI shopping assistant, called Rufus until Amazon renamed it on May 13, 2026, when it combined Rufus and Alexa+. All US customers can use it free in the Amazon app and website (no Prime membership needed), including straight from the search bar. Amazon describes it as Rufus's product expertise brought together with Alexa+'s personal knowledge, and it still answers from the catalog, reviews and the web, so listing work done for Rufus still applies.
This guide walks the stack from the bottom up, one merchant question per section. Each section ends with the article in this cluster that goes deeper, so you can treat this page as the index for the whole Amazon program.
What does Alexa for Shopping read on an Amazon listing?
Amazon has been specific about the inputs. Its original Rufus announcement says the assistant was trained on Amazon's product catalog, customer reviews, community Q&As and information from across the web, and that it generates answers from relevant information across Amazon and the web (About Amazon). Amazon's May 13, 2026 Alexa for Shopping launch post adds what shoppers now do with it: ask questions in the search bar, compare selected products side by side, read AI-generated overviews in search results and on product detail pages, check up to a year of price history, and on eligible items let the Buy for Me feature complete the purchase. It also says the assistant compares features, prices and reviews across Amazon and the web.
Amazon doesn't publish how it ranks or weighs those inputs, so treat anything beyond the input list as observation. What follows from the list is enough to plan around: every field on the detail page is now something the assistant may repeat to a shopper, and every empty field is a question it can't answer about you.
| Element | Classic role | Role with Alexa for Shopping | Failure mode |
|---|---|---|---|
| Title | Keyword rank | Identity and the primary facts it states | Keyword-stuffed title the assistant can't parse |
| Bullets | Skimmable benefits | Claims it repeats when asked why | Superlatives with no checkable fact |
| Structured attributes | Hidden search terms | Answers to constraint questions (size, material, fit) | Empty or wrong values |
| A+ content | Brand visuals and conversion | Corroborating detail, if it's text | Key facts locked inside images |
| Reviews and Q&A | Social proof | Evidence of real-world use | Thin, star-only reviews; unanswered questions |
The practical move is to read your own page the way the assistant would: list the questions shoppers ask in your category and check whether each one has a plain answer somewhere on the page. That self-audit, as 20 yes-or-no questions, is the Alexa for Shopping listing test (the Rufus Test). Finish it by reading the AI overview Amazon now shows on your own detail page: it is the assistant's summary of your product.
Which Amazon attributes should I fill first?
Start with the product type, because it decides which attributes exist at all. Amazon's developer terminology defines a product type as a hierarchical categorization of catalog items and says an item's data requirements are tied to it (SP-API terminology). The Product Type Definitions API returns that requirement set as a JSON Schema, which is the same rule set behind the category templates and flat files sellers fill out.
- Product type
- Amazon's classification of a catalog item (for example, LUGGAGE) that determines which attributes the listing can carry and which are required. A wrong product type means the wrong attribute set, so the right facts have nowhere to go.
Inside the right product type, prioritize by shopper question rather than by field order. Suitability (who or what it's for), compatibility (what it fits), materials and care, safety, and dimensions are the attributes behind the constraint-heavy questions shoppers type: will it fit, is it dishwasher safe, is it BPA-free. Fill those accurately before you chase completeness elsewhere.
- 01
Confirm the product type
Check that each ASIN sits in the product type that matches what it is. A misfiled type carries the wrong attribute schema.
- 02
List the category's real questions
Pull them from your reviews, competitors' reviews, customer messages and returns reasons. Map each question to the attribute that answers it.
- 03
Fill required, then question-mapped fields
Required fields keep the listing live; question-mapped fields get it recommended. Do both before optional extras.
- 04
Verify values against the product
Accuracy beats coverage. A wrong material or size becomes a wrong answer, then a return.
- 05
Re-check after template changes
Amazon revises product type definitions. Schedule a review when you bulk-upload or Amazon flags new required fields.
The field-by-field method, including flat-file discipline at catalog scale, is in the Amazon backend attributes guide. The same principle applies on Shopify and in feeds; see the catalog enrichment playbook.
Why is my product in the wrong Amazon category, and does it matter?
- Browse node
- A named location in Amazon's catalog hierarchy, used for navigation, product classification and site content. An ASIN's browse node decides which category pages, filters and best-seller lists it appears in.
Browse node and product type are related but not the same: the product type governs your data, the browse node governs your shelf. Products end up on the wrong shelf at listing creation, during bulk uploads, and when Amazon migrates product types, and they rarely fix themselves. Amazon's own tooling treats placement as something to check: the Product Type Definitions API can return Amazon-recommended browse nodes and item type keywords for a product type (SP-API use case guide), and the Browse Tree report exposes the hierarchy so you can compare where you are against where you should be.
Does it matter? Amazon doesn't document category placement as a ranking factor, so be careful with anyone who quantifies it. The mechanical effects are clear enough: the node sets which category filters apply, which competitive set and best-seller list you sit in, and which category context sits next to your product data. A kitchen scale filed under office supplies is competing for the wrong shopper.
Browse node errors and product type errors compound: a wrong type means the wrong attribute schema, and the wrong schema often brings the wrong node. Review both in the same pass, quarterly and after every bulk upload.
How to spot a misplaced ASIN and file a correction with evidence is in browse tree and category corrections.
Does A+ content help with Alexa for Shopping?
A+ content is the module-based section below the bullets. To publish it you need a Professional selling account and a Brand Representative or Reseller role on a brand enrolled in Brand Registry (Amazon A+ Content). Amazon says Basic A+ can lift sales by up to 8% and well-implemented Premium A+ by up to 20%, citing its own internal data. Those are Amazon's numbers, not independent results, and they measure conversion, not assistant visibility.
For Alexa for Shopping, the useful question is which parts of A+ carry checkable facts in a readable form. Two rules hold up. First, treat text inside images as invisible: if a spec, dimension or care instruction matters, put it in a text field or the alt text, not only in a designed graphic. Second, the comparison chart is the highest-value module, because it states attributes side by side in a structured grid, the same shape as the comparisons Alexa for Shopping builds for shoppers.
- Text modules: use them for claims you want repeated, with specifics (materials, measurements, test standards) rather than adjectives.
- Comparison chart: compare your own variants or range on the attributes shoppers ask about. Amazon notes Basic A+ charts are shoppable.
- Image modules: keep them, but caption the facts in text and write descriptive alt text.
- Brand Story: useful for trust and cross-selling the range; light on product facts.
Module-by-module priorities and how to rework existing A+ without starting over are in A+ content that Alexa for Shopping can read. Whether the Premium upgrade earns its cost is a separate decision, covered in Brand Story vs A+ Premium.
How do reviews affect AI shopping answers?
Reviews are on Amazon's own list of what the assistant was trained on, and Alexa for Shopping compares reviews across products when shoppers ask it to. That changes what a good review base looks like. A star average tells the assistant little; a review that says the jacket ran small in the shoulders, or the pan warped on an induction hob, is an answer to a future shopper's question. Substance beats volume.
Get more substantive reviews the ways Amazon allows. Amazon Vine lets eligible brands send units to a community of trusted reviewers, up to 30 reviews per product, if you're a Professional seller with eligible FBA offers and the right Brand Registry role (Amazon seller blog). The other safe lever is Amazon's own: the Request a Review button on the order details page sends a standard request to every buyer, so no one is filtered.
Everything else is a policy risk. Amazon's communication guidelines tell sellers not to attempt to influence ratings, feedback or reviews, and not to ask customers to post positive reviews or remove negative ones (Amazon Customer Reviews tool). Paying for reviews, offering rewards or asking only happy customers breaks those rules; the full customer review policy sits in Seller Central help (login required).
Review manipulation is the fastest way to lose everything else in this guide. Anything that offers a reward in exchange for a review, or asks only happy customers for one, is out of bounds. Read the current customer review policy in Seller Central before you run any review program.
Community Q&A sits next to reviews in Amazon's description of the assistant's inputs. Where questions appear on your listing, answer them with the same specifics you'd put in an attribute. Responding to negative reviews with a fix (not an argument) is also visible behavior. The full program is in reviews as machine-readable evidence.
How do I know if my listing is winning demand on Amazon?
- Search Query Performance (SQP)
- A Brand Analytics dashboard that lists the search queries shoppers used to find your brand, with query volume and the impressions, clicks, cart adds and purchases your brand or ASINs earned for each, alongside the totals for the whole query.
SQP is the closest thing Amazon gives brands to a share-of-demand report. Access needs a Professional selling account and a Brand Representative role on a brand enrolled in Brand Registry (Amazon Brand Analytics); the dashboard sits under Search Analytics, next to Top Search Terms. Because it compares your counts with the query's totals, it shows not just traffic but share at each stage of the funnel.
Read it as a chain. Low impression share on a query you should own is a visibility problem: relevance, attributes, browse node, or not enough ad coverage. Click share well below impression share means the search result loses: main image, title, price or rating. Cart-add and purchase share below click share points at the detail page and offer: content, reviews, delivery promise. Each break has a different owner, which is why SQP is the report to open first in a weekly review.
SQP blends organic and sponsored activity and has no spend fields, so pair it with the Ads search term report before you move bids. The weekly workflow and the three patterns to look for are in how to read the Search Query Performance report.
TACOS vs ACOS: which number should run the account?
Both, for different jobs. Amazon Ads defines ACOS as ad spend divided by ad revenue, times 100, and ROAS as its inverse (Amazon Ads ACOS guide). ACOS tells you whether a campaign is efficient. It can't tell you whether the ads are creating sales or just buying ones you'd have made organically, because organic sales aren't in it.
- TACOS
- Total advertising cost of sales: ad spend as a percentage of total sales, organic and paid combined. Amazon Ads describes it as showing whether total sales are growing in relation to ad spend.
That's the gap TACOS closes (Amazon Ads). If ad spend rises and TACOS rises with it while total sales stay flat, the ads are cannibalizing organic demand. If TACOS drifts down while revenue holds or grows, organic demand, including demand the assistant sends you, is doing more of the work. There's no universal good TACOS; the trend against your margin is what matters.
| Metric | Formula or source | Answers | Blind spot |
|---|---|---|---|
| ACOS | Ad spend ÷ ad-attributed sales | Is this campaign efficient? | Ignores organic sales; rewards bidding on your own brand |
| ROAS | Ad-attributed sales ÷ ad spend | Same as ACOS, inverted | Same as ACOS |
| TACOS | Ad spend ÷ total sales | How ad-dependent is the business? | Doesn't say which ads drive the change |
| SQP share | Brand Analytics, per query | Are we winning the search, and where do we lose it? | No spend; blends organic and sponsored |
| AMC queries | Clean-room event data | New-to-brand share, overlap, time to purchase | Needs a question first; outputs are aggregated |
Run the worked example and the cannibalization check in TACOS, not ACoS, or put your own numbers into the free TACOS calculator.
When is Amazon Marketing Cloud worth opening?
- Amazon Marketing Cloud (AMC)
- Amazon's clean-room analytics workspace, built on AWS Clean Rooms, where advertisers analyze pseudonymized ad and conversion events and get back only aggregated, anonymous results.
Amazon says AMC is available at no cost to eligible advertisers and that Sponsored Ads advertisers can use it self-service, with no-code insights alongside custom queries (Amazon Ads). The cost is attention, not money: AMC is only worth opening when you have a question standard reports can't answer.
Three questions usually justify it for a mid-market brand. What share of ad-driven purchases are new to the brand, so you know whether ads are acquiring or just re-buying existing customers. Where two ad products touch the same order, so you stop paying twice. And how many days the purchase really takes, which tells you whether a short attribution window is undercounting upper-funnel work. AMC also lets you define your own attribution approach in a query (different lookback, first-touch or position-based) instead of standard reporting's last-touch view.
Those three queries, how to read them together, and when another query is the wrong next step are in Amazon Marketing Cloud: the first three queries.
How should Amazon PPC campaigns be structured?
Amazon describes Sponsored Products as cost-per-click ads that promote individual product listings, with keyword and product targeting you set or Amazon sets automatically (Amazon Ads). The structure question is what goes in each ad group. Put several products in one group and Amazon's systems choose which one to show per search, so spend drifts to whichever ASIN wins early, the others starve, and your reports can't say which product a dollar bought.
- SPAG (single-product ad group)
- A campaign structure with one advertised product per ad group, so every bid, search term and dollar of spend maps to exactly one ASIN.
SPAG fixes attribution and bid precision at the cost of more ad groups. Manage the count with naming conventions, portfolios and bulk-sheet templates rather than by bundling products back together. It also makes per-SKU TACOS possible: you can only divide one product's ad spend by its total sales if the spend is isolated to that product. Reserve the full treatment for hero and mid-tier ASINs, and run a simpler catch-all structure for the long tail.
We split a product into its own ad group once it carries enough spend that you'd change its bid on its own numbers. Below that, a shared catch-all group is fine. The test is practical: if you can't name the decision a separate ad group would let you make, don't create it yet.
Build steps and the common mistakes are in SPAG PPC structure, explained.
Seller Central or Vendor Central: which account model fits?
Amazon's developer terminology states the difference in one line each: a seller lists and sells their own goods on Amazon, and a vendor supplies inventory that Amazon sells. Everything else follows from who the retailer is. In Seller Central (3P) you set the price, own inventory until it sells, pay Amazon's selling fees and control the detail page through your Brand Registry role. In Vendor Central (1P) you sell wholesale on purchase orders, Amazon sets the retail price, and terms, allowances and chargebacks replace referral fees.
| Question | Seller Central (3P) | Vendor Central (1P) |
|---|---|---|
| Who is the retailer? | You | Amazon |
| Who sets the retail price? | You | Amazon |
| Who owns inventory risk? | You, until the sale | Amazon, after it buys on a PO |
| How you get paid | Recurring disbursements of sales minus fees | Invoice payment on agreed terms, minus deductions |
| How you get in | Open registration | Typically by invitation from Amazon's retail team |
| Who controls content? | Brand owner via Brand Registry | Brand owner, with Amazon as retailer |
For most mid-market brands, Seller Central is the default because price control and listing ownership are worth more than wholesale convenience. Vendor can earn its seat for high-volume replenishment products where Amazon's retail reach outweighs the margin you give up, and many brands run a hybrid. Run net contribution per ASIN under both before you chase an invitation; the full decision checklist is in Seller Central vs Vendor Central.
Inside Seller Central the next call is fulfillment. Fees, Prime and the Featured Offer rules are compared line by line in FBA vs FBM in 2026.
On Seller Central your offer also has to win the page itself. What Amazon ranks for the Featured Offer, and why it disappears, is in the Amazon Buy Box (Featured Offer) guide.
In what order should I fix an Amazon account?
Bottom of the stack first. Ads and measurement amplify whatever the catalog already is, so a thin catalog with a clever ad structure just buys traffic for a page that can't convert or be recommended. The sequence below is the one this guide follows, and it's the order the free audit scores in.
- 01
Decide the account model
Seller, Vendor or hybrid, per product line. It sets who can edit what.
- 02
Fix the catalog on hero ASINs
Product type, browse node, question-mapped attributes, then title and bullets in plain, specific language.
- 03
Add the evidence
A+ text and a comparison chart; a review program that stays inside Amazon's rules; answered questions.
- 04
Set the baseline
SQP share by top query and TACOS by product, recorded before you change ads, so you can see what moved.
- 05
Restructure ads
SPAG on heroes, catch-all for the tail, negative keywords from the search term report.
- 06
Add AMC when you have the question
New-to-brand share, overlap and time to purchase, once standard reports run out.
Work on three hero ASINs before the whole catalog. Three is enough to see whether the changes move SQP share and TACOS, and few enough to finish properly in a month. Once the pattern works, roll it across the line.
If you want an outside read first, the free Amazon listing audit scores your top three ASINs on listing quality, catalog readiness, PPC efficiency and Alexa for Shopping surfacing, and returns a sequenced fix list.
Is keyword optimization still worth doing on Amazon?
Yes. Search placement, organic and sponsored, still drives the traffic the assistant then mediates, and SQP is a keyword report at heart. What changed is that keywords alone don't win any more. A title packed with search terms and an empty attribute set ranks for the query and then loses the shopper's follow-up question to a competitor whose listing answers it.
Write titles that state what the product is in the words shoppers use, bullets that answer the category's real questions with specifics, and backend search terms for the synonyms and misspellings that don't belong in visible copy. Keep the keyword discipline and add intent and attribute completeness on top. For how AI assistants beyond Amazon choose products, see how products get recommended by AI.
FAQ
Is Alexa for Shopping the same as Rufus?
- It replaced Rufus. Amazon renamed Rufus to Alexa for Shopping on May 13, 2026, bringing Rufus's product expertise together with Alexa+ in the Amazon app and website. Amazon doesn't publish how either version weighs listings, but the inputs it describes (catalog, reviews, web) are the same, so listing work done for Rufus still applies. Sellers still search for Rufus, which is why the name stays in titles like this one.
What data does Alexa for Shopping use to answer questions?
- Amazon says Rufus was trained on its product catalog, customer reviews, community Q&As and information from across the web, and that answers draw on information across Amazon and the web. Alexa for Shopping adds the shopper's own history and Alexa conversations to personalize answers. Amazon doesn't publish how those inputs are weighted, so the safe plan is a complete, accurate detail page backed by real reviews.
Do backend attributes matter if shoppers can't see them?
- Yes. Attributes are the structured facts behind constraint questions such as size, material, compatibility and safety. Which attributes exist depends on the product type, so check that first. An empty attribute leaves a question unanswered; a wrong one creates a wrong answer and, often, a return. Prioritize accuracy over coverage.
Can Alexa for Shopping read text inside A+ images?
- Amazon doesn't say, so plan as if it can't. Put every fact that matters, such as dimensions, materials, care and test standards, in A+ text modules, a comparison chart or descriptive alt text. Keep images for persuasion, and caption the facts they show in text that any system can read.
What is a good TACOS on Amazon?
- There's no universal number. TACOS is ad spend divided by total sales, so it depends on your margin, category and growth stage. Watch the trend: falling TACOS at steady or rising revenue means organic demand is doing more of the work. Rising TACOS with flat sales means ads are buying sales you'd have had anyway.
Do I need Brand Registry for this?
- For most of it, yes. A+ content, Brand Analytics (including Search Query Performance) and Amazon Vine all require a role on a brand enrolled in Brand Registry, which needs a registered or pending trademark. Attributes, browse nodes, Sponsored Products and TACOS tracking work without it, but you'll be measuring with less data.
Is Amazon Marketing Cloud free?
- Amazon says AMC is available at no cost to eligible advertisers, and Sponsored Ads advertisers can use it self-service. The real cost is analyst time. Open it when you have a question standard reports can't answer, such as new-to-brand share, overlap between ad products, or how many days a purchase takes.
Should a mid-market brand use Seller Central or Vendor Central?
- Usually Seller Central, because you control price and the detail page and get paid on a recurring disbursement cycle. Vendor Central is typically by invitation and makes Amazon the retailer, which can suit high-volume replenishment products. Many brands run both. Compare net contribution per ASIN under each before deciding.
Sources
- 1.Amazon announces Rufus (editor's note) · About AmazonOn May 13, 2026 Rufus was renamed Alexa for Shopping
- 2.Meet Alexa for Shopping · About AmazonMay 13, 2026: Rufus and Alexa+ combined; AI overviews on product pages; free to all US customers
- 3.SP-API terminology · Amazon Selling Partner APIDefinitions of product type, browse node, seller and vendor
- 4.Product Type Definitions API use case guide · Amazon Selling Partner APIAttribute requirements per product type and recommended browse nodes
- 5.A+ Content · Amazon (Sell on Amazon)A+ eligibility and Amazon's own figures on sales lift
- 6.What is Amazon Brand Registry? · Amazon (Sell on Amazon)Trademark requirement; unlocks A+, Brand Analytics and protection
- 7.Amazon Brand Analytics · Amazon (Sell on Amazon)Search Query Performance: impressions, clicks, cart adds, purchases
- 8.ACOS: advertising cost of sales · Amazon AdsAmazon's own definition of ACOS, the ratio TACOS widens to total sales
- 9.Common holiday advertising myths · Amazon AdsTACOS counts organic and paid sales against ad spend
- 10.Amazon Marketing Cloud · Amazon AdsThe clean room, who can use it, aggregated outputs
- 11.Sponsored Products · Amazon AdsCost-per-click ads for individual listings; keyword and product targeting
- 12.Tips for new product success · Amazon (Sell on Amazon)Amazon Vine eligibility and review limits
- 13.Customer Reviews tool · Amazon (Sell on Amazon)Request a Review, and the rules against influencing reviews
Founder of GigaCommerce, part of Gigaverse. Works with mid-market Shopify and Amazon merchants on agentic commerce installs, AI-ready catalogs, and Commerce GEO.
LinkedIn