Features and Amazon integration

What Rufusly does

Rufusly helps Amazon selling partners write compliant, well-structured listings and understand how those listings perform. It is a subscription web application at rufusly.ai. Sellers sign in, add a brand profile, and use the tools described below on their own catalogue.

The platform has two halves. The first writes and checks content: listing copy, product imagery, and the brand rules that govern both. The second reads a seller's own Amazon data so that the content decisions are informed by real performance rather than guesswork.

What it does that a spreadsheet does not

A spreadsheet will hold anything you type. Amazon will not. The gap between those two facts is where most rejected uploads come from, and it is the gap Rufusly closes.

Every field Rufusly writes or reads is measured against Amazon's published limits before you see it, using the same rules on the server that generates the content and in the browser that audits your uploaded file. Those checks are deterministic. They involve no AI, they cost no credits, and they give the same answer every time for the same input.

The byte distinction matters more than it sounds. An accented character, a currency symbol or an emoji costs more than one byte, so a backend keyword string that looks comfortably short in a spreadsheet can still be rejected, so Rufusly counts the bytes rather than the characters.

Alongside the limits, titles are stripped of characters that cause problems in an Amazon title. That covers Amazon's banned set and, as our own policy, decorative symbols and dash variants: the exclamation mark, dollar sign, question mark, underscore, braces, caret, angle brackets, asterisk, tilde, the pipe, and both en and em dashes. Backend terms are checked for words repeated inside the field and for words already used in the title or item highlights, because repeating a word there spends bytes without adding anything Amazon will index.

Who it is for, and who it is not for

It is built for Amazon sellers and the agencies that manage listings for them, particularly anyone maintaining more than a handful of SKUs by hand, and anyone who has had an upload come back with errors on every row and no obvious reason why.

It is not a product research tool. There is no supplier sourcing, no competitor price tracking, and no sales estimate for products you do not own. When the advertising manager is released, bids will be changed only inside a plan the seller sets, if they choose to set one. The advertising manager is not yet generally available. Rufusly reads your own account and works on your own catalogue. If the question is what to sell next, this is the wrong tool.

Listing management

Listing Studio is the core of the product and is available today. You paste or upload your existing listing content and it rewrites it to be compliant and easier for shoppers and Amazon's own search tools to understand. Once you connect your Amazon account, it reads that content directly from your catalogue instead.

What it does

• Rewrites listing content. Title, bullet points, product description, and backend search terms, one listing at a time or in bulk. A+ content copy is generated alongside a rewrite, with a dedicated A+ Content Builder on the Growth and Agency plans.
• Enforces Amazon's limits. Every generated field is checked against Amazon's rules before it is shown, including the 75 character title limit, banned characters, byte limits on backend search terms, and duplicate terms across fields.
• Listing Quality Score. Scores a listing out of 100 and lists exactly which fields fail and why, so a seller can see what to fix before publishing anything.
• Bulk rewriting from a flat file. A seller uploads their Amazon Category Listings Report and Rufusly rewrites the non-compliant rows, then exports a file that can be uploaded back to Amazon. Every other column is returned byte for byte as it came in. Free accounts are limited to 10 rows.
• Brand profile. Sellers record their brand name, product category, key claims, target customer, and certifications once. Every rewrite for that brand is written against those details, so generated copy stays consistent with the brand.
• Brand Safety Vault. Sellers keep a confidential list of forbidden claims, restricted words, and required disclaimers. Rufusly checks every generated field against that list after generation and flags anything that breaches it. The list is held server side and is never sent to the AI model.
• Version history. Previous versions of a listing are kept with their scores, so a seller can compare a rewrite against what came before.

What a flat file is, and why Amazon rejects them

A flat file is the spreadsheet Amazon accepts for changing many listings at once. Most sellers get theirs by exporting a Category Listings Report from Seller Central. It looks like an ordinary workbook and behaves like nothing of the sort.

Cell A1 carries a settings string that tells you how the sheet is laid out. In a typical report, row 4 holds the human-readable column labels, row 5 holds the fully qualified machine attribute names, row 6 holds Amazon's own example row, and the real data starts at row 7. Duplicate labels on row 4 are normal, because "Bullet Point" legitimately appears five times.

Amazon also moves the columns between template exports. That single fact causes a large share of failed uploads: a seller reuses last quarter's template, the columns have shifted by one, and every value lands one column to the left of where it belongs. Rufusly reads the settings string in A1 and resolves every column by its attribute name or its label instead of by position, so a shifted template is handled rather than silently corrupted. Where a file uses Amazon's qualified attribute names, such as a bullet point tagged with the marketplace and the language, those resolve too.

What the pre-upload check inspects

The audit runs in your browser on the file you uploaded. No AI, no credits, no waiting. Every row and every field is measured against the limits in the table in section 1, and the result is grouped into sections you can act on: titles, item highlights, bullet points, backend search terms, variation structure, data integrity, size and variation, required attributes, and brand compliance where a compliance profile is set.

On top of that audit, a ten step pre-upload gate runs on the bulk updater before anything is exported. It works through the steps in order, and reports every one, including the steps it had to skip because you had not linked approved copy or a previous processing summary: row count sanity, whether the content matches copy that was approved, price anomalies, product identifiers such as EANs, variation structure, title compliance, any field that failed a previous upload, required universal attributes, value correctness against the lists of values Amazon accepts for that field, and finally a sweep for en and em dashes. It ends on one of two verdicts, ready or fix first, and names exactly what is blocking.

The point is that you find out before Amazon does. A processing summary arriving hours later with an error code on every row is the expensive way to learn that a column moved.

What you get back

An .xlsx file built from the workbook you uploaded, with only the fields you accepted overwritten. Fields that passed are never rewritten, are never sent to the AI, cost no credits, and come back exactly as they went in. That is the part sellers usually want confirmed: a bulk fix should touch the broken rows and leave a working catalogue alone. Where a file has no item highlights column, one is inserted immediately after the title column so the content has somewhere to go.

What the Listing Quality Score measures

It scores one listing out of 100 across five sections with fixed maximums: the title out of 20, bullet points out of 30, the description out of 20, backend search terms out of 15, and shopper question readiness out of 15. The total is graded A at 90 and above, B at 80, C at 65, D at 50, and F below that, and every failed check is listed with what to change and why.

It is worth being clear about what it does not measure. It is not an Amazon score, and Amazon neither produces it nor sees it. It says nothing about your sales, your conversion rate, or where you rank, because it only ever reads the text of the listing. The shopper question readiness section is tuned towards consumable products: it asks whether ingredient names are spelled out rather than abbreviated, whether the product format and the serving size are stated, and whether dietary suitability is addressed, on the basis that a shopping assistant can only answer "is this vegan" if the listing says so somewhere.

Amazon data used

Listing content and attributes for ASINs in the seller's own catalogue, plus the product type schema Amazon publishes for each category, which we use to detect missing required attributes.

Inventory and orders

The inventory view shows a seller which of their products need attention, and connects those problems to the tools that fix them.

What it does

• Listing and SKU view. Every SKU the seller has listed, with its ASIN, product type, listing status and the count of issues Amazon reports against it. Stock quantities and days of cover are not currently shown, and are on the roadmap rather than available today.
• Suppressed listing detection. Reads the issues Amazon reports against a listing and identifies which ones are search suppressions. Where the cause is a main image that does not meet Amazon's guidelines, for example a lifestyle photo used where a plain white background is required, Rufusly flags it and offers to generate a compliant replacement image.
• Guided fixes. Suppression reasons that need seller action in Seller Central, such as category approval, are shown as read only with an explanation, rather than being presented as something the tool can fix.

What suppressed means, and why it happens

A suppressed listing has been pulled out of Amazon's search results because something about it breaches a rule. The product page usually still exists and the stock is usually still there. Shoppers simply cannot find it, which is why a suppression can go unnoticed for weeks while sales quietly stop.

Rufusly reads the issues Amazon reports against each listing and treats only errors as a suppression, so warnings and informational notices do not send you chasing something that is not broken. It then splits the errors into ones it can help with and ones it cannot, and says which is which.

How an image suppression is classified

Main image problems are not all the same problem, and treating them as one is how a seller ends up paying for an edit that changes nothing. Rufusly reads the wording of Amazon's message and sorts it into one of five kinds.

• Background. The background is not pure white, or carries shadows or props. Removing the background addresses this.
• Overlay. Text, a logo, a watermark, a badge, a border or a promotional banner is burned into the photograph. This needs the offending element painted out, not the background swapped, and Rufusly tests for it first because Amazon's overlay message often mentions the white background requirement as well.
• Content. What is depicted is wrong: more than one product, accessories that are not included, a model, or packaging. No edit makes the wrong photograph the right one. Rufusly says so rather than generating something that cannot lift the suppression.
• Technical. The file is too small, the wrong shape, or an unsupported format, all of which are mechanical faults with nothing to do with the image content.
• Unrecognised. Amazon said something Rufusly does not have a rule for. It reports the message as received and does not guess at a repair.

Only the main image is offered as an automatic fix. A suppression on a secondary image is reported but not repaired, because patching a new main image would not resolve it and would needlessly change an image that was compliant.

Amazon data used

Listing status and the issues Amazon reports against each listing. Order data is read only to identify which recent orders are eligible for a review request. We do not read buyer names, addresses, or contact details.

Brand Analytics

Brand Analytics data tells a seller which search terms actually convert for their own brand. Rufusly uses it so that listing copy is written around proven terms rather than guesses.

What it does

• Search query performance. Top search terms for the seller's brand with impression share, click share, and conversion share, reported by ASIN. Available to AI agents through our MCP server; it does not yet have a screen in the web app.
• Keyword intelligence. Search query performance for a chosen ASIN, ranked by purchase share, click share, and cart add rate, so the seller can see which keywords earn sales rather than only clicks. These terms are then used by Listing Studio when the title and backend terms are written. Requested through our MCP server. We do not use the market basket or repeat purchase reports.

Amazon data used

Aggregated Brand Analytics reports for the seller's own brand and ASINs only, retrieved on the seller's behalf and shown only to that seller.

Selling Partner Insights

Sellers need to see whether a listing change made a difference. The insights view reports performance over time so the effect of a rewrite is visible.

What it does

• Sales performance. Revenue, units sold, and order counts by ASIN over a chosen date range. Available to AI agents through our MCP server; it does not yet have a screen in the web app.
• Version history. Listing versions are stored with the date they were published, so a seller can see exactly what the content said at any point. Automatic correlation of a content change against the sales that followed it is on the roadmap, not built.

Amazon data used

Sales and traffic figures for the seller's own account and ASINs. We also read review and return themes for the seller's own ASINs through Amazon's Customer Feedback API, which is a separate API from Brand Analytics, to inform bullet points and flag claims that reviews contradict.

Finance and accounting

Content decisions have a margin consequence, and money Amazon owes back is easy to miss. Both are surfaced here.

What it does

• Financial event summary. Referral and fulfilment fees and other financial events over a chosen window, so a seller can see what a product actually costs them before investing in new content or imagery for it. Available to AI agents through our MCP server. We do not currently read settlement reports or compute a net margin figure.
• FBA reimbursement finder. Reviews financial events over a chosen window to surface inventory that was lost, damaged, or incorrectly charged, and reports pending and approved reimbursement totals with an itemised list. Rufusly reports what it finds. The seller decides whether to raise a case.

Amazon data used

Financial event and settlement data for the seller's own account. We do not access payment instruments or bank details.

Review requests

Reviews carry weight with shoppers deciding whether to buy. Rufusly lets a seller ask for them using Amazon's own mechanism.

What it does

• Amazon templated review requests. Submits Amazon's standard review request for orders that are eligible under Amazon's own rules. Eligibility is determined by Amazon: we submit the request and Amazon accepts or rejects it against its own permitted window.
• No custom messaging. Rufusly uses Amazon's Solicitations API and Amazon's own template. We do not write, alter, or add to the message, and we do not offer incentives or attempt to influence review content.
• You confirm before anything sends. Rufusly shows the eligible orders and sends only once you confirm. Review requests are managed from your Amazon connection settings.

How eligibility is decided

Amazon accepts a review request only within a window after delivery, which its documentation puts at between five and thirty days. Rufusly filters the order list to that window before submitting anything, so it does not make calls Amazon is going to refuse.

That filter is a convenience, not a substitute. Amazon's order data does not always carry a delivery date, in which case Rufusly falls back to the purchase date, which is earlier and therefore slightly more generous at the start of the window. Amazon re-checks the window itself on every request and rejects anything outside it. The decision is Amazon's, and Rufusly does not present it as its own.

Amazon data used

Order identifiers and eligibility status only, for the purpose of issuing the request. Rufusly does not request, store, or process buyer names, email addresses, postal addresses, or any other buyer personal data, and no buyer data is sent to any AI model.

Product imagery

Image Studio generates Amazon-ready product images from a seller's own photography. It does not require an Amazon role of its own, but it is used alongside the inventory tools to resolve image-based suppressions.

• Generates main images on a pure white background, lifestyle scenes, and infographic-style images with editable text labels.
• Removes or replaces backgrounds on an existing product photo.
• Where a listing is suppressed for a main image that breaks Amazon's guidelines, the replacement is previewed and only published to Amazon once the seller approves it.

AI search visibility (in development)

Some shoppers now ask an AI assistant instead of browsing search results. This part of the platform, currently in development and not yet released, will track how a brand is described when they do.

• Will run a set of buying questions against public AI assistants and record whether the brand is mentioned, how it is described, and which sources are cited.
• Will report share of voice against named competitors over time, and the sources those answers draw on.
• Assistant Radar will simulate how an Amazon shopping assistant is likely to summarise a listing, based on the listing's own content. Its output will be labelled as simulated responses, not live output from Amazon Rufus.

This feature will use public AI assistants and the seller's own listing content. It will not use Amazon Brand Analytics data. AI search visibility and Assistant Radar are not yet available. They are in development. Which plans they will be available on, and how they will be metered, will be confirmed when the feature is released.

Advertising (in development)

A seller who advertises on Amazon usually reads their ad reports in one place and edits their listings in another, so it is hard to tell whether a content change made the advertising work harder. Rufusly has been granted access to the Amazon Ads API and is building an advertising manager that sits beside the listing and imagery work already described on this page. It is not yet generally available.

It is being built for a brand owner rather than an agency, so the seller will see products, money and a record of what happened, not campaign structure. Campaigns, ad groups and match types stay behind an advanced view.

• Products. Ad spend, ad sales and cost of sale for every product, with a status that reflects whether the product can convert at all: a suppressed listing, a lost Featured Offer or an empty stock position is flagged, because spend on a product that cannot sell returns little or nothing at any bid.
• Results. Whether the advertising is making money, with the attribution window stated on every figure and a note whenever the underlying data is too recent for Amazon to have finished reporting it.
• Activity. Everything Rufusly noticed, suggested or did, newest first, each entry with its reason and the evidence behind it. Planned: changes made outside Rufusly, in Seller Central or by another tool, will appear here too, so the record is complete rather than only Rufusly's side of it.
• Plan. One short form: which products to advertise, a starting daily budget for each, the cost of sale to aim for, a daily spending ceiling if the seller wants one, and how much to let Rufusly do.

How changes are made

When released, Rufusly will start by watching. It will read the reports and show what it would have changed, and why, without changing anything. Applying changes will begin only when the seller sets a plan and chooses a level above Suggest only, which is the default and under which every change waits for approval. Handle the routine will apply changes inside the seller's limits, which can be undone where Amazon allows, while pausing anything still waits for approval. Handle everything within my limits will also allow pausing. Every applied change will be recorded with what it was before, what it became and the reason, and can be reversed from within Rufusly wherever Amazon allows it. Archiving a keyword or negative keyword on Amazon is final: Rufusly can add the same keyword back as a new one, but the archived one itself cannot be restored. A reversal that would raise spend, such as putting a lowered bid back up or re-enabling a paused target, is subject to the daily spending ceiling, if one is set, like any other increase. Reversing restores the setting, not the spend in the meantime. If a ceiling is set, once reported spend for the day reaches it Rufusly will raise no bids; it cannot cap what Amazon delivers within the seller's own campaign budgets, and Amazon's reporting can run several hours behind. A single control will pause all automation immediately.

What it will not do

Once released, Rufusly will only set up and run Sponsored Products advertising, meaning campaigns, ad groups, product ads, keywords, product targets and negative keywords, for products the seller has chosen. It will never advertise a product the seller has not chosen. The starting daily budget for each campaign it creates is the amount the seller confirmed, and Rufusly will never change that budget afterwards, or the budget of any campaign it did not create. Through nightly review of search term reports it can add and archive keywords and negative keywords, but only inside campaigns it created itself. Amazon's own advertising system can deliver somewhat more than a campaign's set daily budget on a given day, under Amazon's own delivery and reporting mechanics; that is Amazon's behaviour, not a change Rufusly makes, and Rufusly does not control or guarantee what Amazon actually spends against a budget on any given day. Within the seller's limits, Rufusly may also raise or lower bids and pause or re-enable targets in the campaigns it created. The one exception to Rufusly never touching a campaign it did not create is the handover: if the seller chooses that option, Rufusly will pause, never delete, their existing ads for those same products, and only those, 48 hours after the new ones start, so the two do not compete for the same shoppers. The seller can resume a paused ad themselves at any time. Rufusly will not apply any change outside the limits of the seller's plan. No advertising spend is committed by Rufusly beyond what the seller has authorised by setting that plan, and Rufusly does not guarantee any advertising outcome.

Amazon data used

When released, advertising reports for the seller's own campaigns and ASINs will be retrieved on that seller's behalf and shown only to that seller, in line with Amazon's Acceptable Use Policy and the terms governing the Amazon Ads API. As with every other Amazon data source on this page, we do not combine one seller's data with another's, we do not build benchmarks from it, we do not use it to train AI models, and we do not sell, license, or otherwise share it. Advertising data is reached through a separate Amazon Advertising authorisation, not through the Selling Partner API roles listed at the top of this page, so a seller who does not authorise it is unaffected.

The advertising manager is not yet generally available. Which plans it will be available on, and how it will be metered, will be confirmed when it is released.

Agent access (MCP)

The tools described above are also available to AI assistants through a Model Context Protocol server at mcp.rufusly.ai, so a seller can work in an AI chat client instead of the web interface. Content tools such as listing rewriting work today. Tools that read Amazon account data become available once you connect your Amazon account. Access is authenticated with either a scoped API key or OAuth 2.1, and an agent can only ever reach the data of the account that authorised it. The same approval step applies: an agent can draft content and can list the orders eligible for a review request, but anything that writes to Amazon or reaches a buyer requires the seller to confirm it first. Setting or changing an advertising plan through an agent, when the advertising manager is released, will carry the same confirmation step as the web app. MCP access is included on every plan.

Why this matters

Rufusly does not assume you are sitting in front of a dashboard. Because the tools are exposed over MCP, a seller can ask an assistant to score a listing, pull last month's search query report, and draft the rewrite, all in one conversation, without opening the web app at all. The rules are identical either way. The same character limits, the same Brand Safety Vault check, the same approval step before anything is published.

What an agent can call

Tools fall into three groups. Content tools work for every account. Tools that read your Amazon data need the connection first, and until it exists they return a plain error telling the agent to connect, rather than returning something invented.

• Content, available now. Rewrite a listing, rewrite listings in bulk, score a listing, read its version history, generate images, save an image from a URL, produce an A+ content brief, and draft a support case.
• Your Amazon data, once connected. Listing data, inventory status, sales performance, Brand Analytics, the search query report, keyword intelligence, customer voice, finance summary and reimbursements.
• Actions that reach outside. Review requests, which still require your explicit confirmation before anything is submitted to Amazon.

How access is authenticated

The server speaks JSON-RPC 2.0. You can authenticate with a scoped API key, which is stored only as a SHA-256 hash so the key itself cannot be read back out of our database, or with OAuth 2.1 using PKCE, where the access token is valid for thirty days and the client then refreshes it. Either way the credential is bound to one account. There is no shared or organisation-wide token, and no way for an agent holding one seller's credential to see another seller's data.

Connecting your Amazon account

Rufusly reads Amazon data only after a seller explicitly connects their Selling Partner account, and only for the account that authorised it.

• Amazon's own consent flow. Connection uses Amazon's standard authorisation flow from within Seller Central or from the Rufusly app. The seller sees exactly which roles are being requested and must approve them.
• Least privilege. We request only the six Selling Partner API roles listed at the top of this page, each of which maps to a feature described here. The advertising manager, when released, will require its own separate Amazon Advertising authorisation.
• Read first, write only on approval. Reading data happens in the background so the tools have something to work with. Writing anything back to Amazon through the Selling Partner connection, whether listing content or a main image, always requires the seller to review the exact change and approve it. The advertising manager, when released, will be the one exception, on the terms in section 10: changes inside a plan the seller has set and confirmed.
• Revoke at any time. A seller can disconnect from within Rufusly, or revoke authorisation in Seller Central under Apps and Services. Access stops as soon as the authorisation is revoked.
• More than one account. Agencies can connect several selling accounts and keep each brand's data and rules separate.

How we handle data

Full detail is in our Privacy Policy. In summary:

• No buyer personal data goes to AI models. The content we generate is written from product information, not from buyer data. Buyer names, email addresses, and other contact details are never included in any prompt.
• Processing is server side. All AI processing happens on our servers, never in your browser. Amazon credentials and tokens are held server side and are never exposed to the browser.
• Not used for training. We do not use your content to train AI models. The AI providers we use are listed in our Privacy Policy, along with exactly what each one receives.
• Not shared or sold. One seller's Amazon data is never pooled into benchmarks and never sold or licensed. It is shown only to that seller and to anyone they have explicitly invited into their brand, such as an agency they have chosen to connect to.
• Encrypted in transit using TLS 1.2 or above, and encrypted at rest by our database and storage providers, with access restricted to the account that owns the data.
• International transfers. Some processors are based in or transfer data to the United States. Where that involves a transfer of personal data outside the UK, we rely on Standard Contractual Clauses, the UK International Data Transfer Addendum, or an equivalent recognised safeguard.

Plans and pricing

Rufusly is sold as a monthly or annual subscription. Current plans and the full feature comparison are listed on our pricing section.

What separates the plans

Almost entirely credits. Every plan carries every feature, including agent access through the MCP server, and the difference is how many listing rewrites and image generations you get each month. The free trial is a full 30 days of full access rather than a stripped-down tier, and when it ends the core tools stay available until you pick a plan.

Credits are spent where AI is spent. A listing rewrite costs a listing credit, an image generation costs an image credit, and the deterministic work costs nothing: the flat file audit, the pre-upload gate, and the Listing Quality Score do not touch a credit, because no model is called. In a bulk run, rows that already pass are exported unchanged and are never charged. Monthly and annual billing are both available, and annual is billed at a lower monthly rate.

Prices are in pounds sterling. No VAT is charged, because Rufusly is not VAT registered. You can cancel at any time from your account settings, which stops the next renewal. Because subscriptions are billed in advance, cancelling does not refund the current period, except where you have a statutory right to a refund. Your rights under the Consumer Contracts Regulations are set out in our Terms of Service.

Contact us

Questions about the platform, our Amazon integration, or this page:

• Email: hello@rufusly.ai
• Support for existing customers is available from within the app.

Rufusly is a trading name of LUMINOUS EMPORIUM LTD, registered in England and Wales, company number 16201976. Registered office: 124-128 City Road, London, EC1V 2NX, United Kingdom.

Questions sellers ask

Answered against what the product does today, not what is planned.