How Amazon's AI decides what to recommend, what makes a listing machine-readable, how compliance pre-checks and POA appeals work, and how to connect ListingGood to your own AI assistant over MCP.
What is ListingGood?
ListingGood is an AI recommendation engine for Amazon sellers. It rewrites and structures your listing so Amazon's AI discovery systems (COSMO, Rufus) and AI shopping assistants can parse, trust, and recommend the product. It has three core capabilities: (1) AI listing writing and optimization, (2) compliance pre-check against four risk libraries, and (3) appeal and negative-review rescue including POA generation. It is agent-native: the same tools are exposed as an MCP server so Claude, ChatGPT, Cursor and other MCP clients can call them directly.
How is Amazon's AI recommendation (COSMO, Rufus, Alexa) different from the old A9 algorithm?
A9 was primarily keyword matching plus sales rank. Since 2023 Amazon has layered AI systems on top: COSMO, a large-model knowledge graph that infers shopper intent and use-case relationships; Rufus, a conversational shopping assistant that answers buyer questions; and Alexa shopping capabilities. These systems read your listing for meaning, entities, and use-case context — not just keyword density. A listing stuffed with keywords but semantically confusing is now penalized in discovery rather than rewarded. See A9 vs COSMO vs Alexa for Shopping.
Why is my Amazon listing not being recommended or surfaced by AI shopping agents?
The most common causes are keyword-stuffed copy with incoherent semantics, missing use-case and entity information (who it is for, what it is made of, where it is used), compliance red lines that trigger silent suppression, and thin structured signals on a new ASIN. ListingGood's free AI Recommendation Readiness check scores a listing across four dimensions — AI Readability, Entity Clarity, Structured Signals, and Compliance Health — so you can see which dimension is holding the listing back before rewriting.
What is an Amazon listing compliance check, and can it prevent takedowns?
A compliance check scans your title, bullets and description against Amazon policy and a compliance knowledge base covering prohibited claims, intellectual property, GPSR and category mismatch, then returns Critical, Warning and Info level findings with suggested fixes. No tool can guarantee a listing will never be taken down, but a pre-check removes the large majority of the triggers — prohibited words, unsubstantiated medical or promotional claims, and category mismatches — before they are published at scale. Run it free at /scan, or read Amazon prohibited & restricted words.
How do I write a POA (Plan of Action) for an Amazon suspension or listing takedown?
A POA has three parts: root cause, corrective actions taken, and preventive measures. The most common rejection reason is a vague root cause that Amazon reads as denial of responsibility. ListingGood's appeal tool matches the notice against the relevant policy clauses and drafts a POA with a specific root cause, dated corrective actions, and preventive controls, citing the policy language — which materially improves acceptance rates for IP complaints, authenticity complaints, GPSR notices and restricted-product takedowns.
Is ListingGood an MCP server? Which AI tools does it work with?
Yes. ListingGood exposes compliance check, review analysis, appeal/POA generation and listing generation as MCP (Model Context Protocol) tools. Any MCP client can call them: Claude, ChatGPT, Cursor, and other MCP-compatible assistants. You connect with an API key only — there is no local install. Setup instructions and the tool reference are at /developers, or read using ListingGood MCP with Claude.
Which Amazon marketplaces and languages are supported?
ListingGood covers the major Amazon marketplaces including the United States, the United Kingdom, Japan, and the European stores (Germany, France, Spain, Italy), with localization across English, German, French, Spanish, Italian and Japanese. The compliance knowledge base is organized across 19+ marketplaces so findings are matched to the rules of the store you actually sell in.
How does ListingGood pricing work? Is there a free tier?
ListingGood is pay-per-use with star credits rather than a seat-based subscription: compliance pre-checks, listing generation, deep compliance reports and POA appeals each consume a different number of stars. New accounts receive a free allowance, and the AI Recommendation Readiness check and the basic compliance scan are free — no credit card required to try it. Current plans are listed at /pricing.
Is my listing data safe, and what is the difference between listinggood.com and listinggood.cn?
They are separate deployments with separate databases. listinggood.com is the international site, served from Singapore and integrated with international models. listinggood.cn is the China site, served from Shanghai with data stored in mainland China and using domestic models, built for PIPL data compliance. Your API key works per site, and data is not shared between the two.
How is ListingGood different from Helium 10 or other Amazon listing tools?
Most listing tools optimize for keyword discovery and rank tracking. ListingGood optimizes for being understood and recommended by AI: it rewrites for semantic and entity clarity, pre-checks compliance before publishing, and covers the post-sale rescue layer (POA appeals and negative-review root-cause analysis) that discovery tools do not touch. It is also agent-native — the same capability is callable as MCP tools from your own AI assistant instead of being locked inside a dashboard.
What are the first steps to make my Amazon listing AI-recommendable?
Start with the free AI Recommendation Readiness check to get a four-dimension baseline. Then run the compliance pre-check to clear any Critical findings, rewrite title and bullets so each bullet states a concrete benefit plus a use case rather than repeating keywords, and add the structured attributes buyers and agents filter on (material, dimensions, compatibility, intended user). Re-score after each change so you can see which edit moved the number. The full playbook is in How to get your Amazon products recommended by AI.