> For the complete documentation index, see [llms.txt](https://docs.sellermagnet.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.sellermagnet.com/selling-tools/listing-optimizer-overview/listing-agent.md).
# Listing Agent
{% hint style="info" %}
**Difficulty:** 🟡 Intermediate · **Reading time:** \~8 min
{% endhint %}
{% hint style="warning" %}
**Availability:** Listing Agent is part of the **Listing Manager AI** suite, which rolls out **per account**. While yours is pending you'll see an amber hourglass next to it in the **Listing-Manager** menu and a "you're on the list" note when you open it - at no charge until you decide to keep it. It needs a connected Amazon account. See [Listing Optimizer](/selling-tools/listing-optimizer-overview.md).
{% endhint %}
{% hint style="success" icon="rocket" %}
**Open this page in your dashboard:** [**Go to Listing Agent →**](https://dashboard.sellermagnet.com/listing-agent)
{% endhint %}
## 📋 Overview
The **Listing Agent** is the fastest way to find and fix content gaps across your entire Amazon catalog. Instead of opening one listing at a time, it scans every product you have synced, ranks them by how badly they need attention, and lets you draft AI copy and push it to Amazon in batches.
Think of it as bulk triage: the Agent shows you where the damage is across many listings at once, drafts a fix for each weak field, and submits the fields you approve, all without leaving the page.
A single scan looks at up to **six content signals**, but only scores the ones Amazon's API can actually see:
| Signal | What it checks | Severity |
| ----------------- | ------------------------------------------------- | ------------------------------------------------------------------------------------------------ |
| **Title** | Present and at least 50 characters | Missing = Critical · Too short = Warning |
| **Bullet Points** | All 5 bullet points present | 0 bullets = Critical · Fewer than 5 = Warning |
| **Images** | At least 5 images (7+ recommended) | 0 images = Critical · Fewer than 5 = Warning |
| **Description** | Only flagged when the body is actually visible | Not visible via the API = excluded (neutral chip) · Visible but under \~100 characters = Warning |
| **A+ Content** | Only flagged when an absence is actually observed | Status unknown (the usual case) = excluded (neutral chip) · Observed absent = Warning |
| **Video** | Only flagged when an absence is actually observed | Status unknown (the usual case) = excluded (neutral chip) · Observed absent = Info |
The reason Description, A+ Content, and Video behave this way is that Amazon's API (SP-API) usually cannot see them - it rarely returns the description body, and A+/video status often comes back unknown. When a signal is unknown, the Agent **excludes it from the score and shows it as a neutral chip** rather than penalizing you for something it cannot verify. An empty or unreadable description is therefore **no longer Critical**; only a description that is genuinely visible and under \~100 characters raises a Warning.
Each product gets a **quality score** that is the **share of the known signals it passes** - in practice that is usually Title + Bullets + Images (out of 3), not always 6 - shown as a percentage with a colored ring: green at 80%+, amber at 50-79%, red below 50%.
{% hint style="info" %}
**"Not visible via Amazon API" chips.** Fields Amazon's API can't see - description, A+ Content, and video - render as a neutral gray chip with a question-mark icon and the tooltip **"Not visible via Amazon API"**. A chip like this is **not** a failure and does not count against your score; it simply means the Agent has no reliable data for that field. Only Title, Bullets, and Images are consistently visible, so those are what the score is normally built from.
{% endhint %}
{% hint style="info" %}
**Where the scan data comes from.** The Listing Agent reads the **latest Listing Quality snapshot** for each product, it does not fetch fresh data from Amazon itself. If the Agent shows an empty state ("No Listing Data Found"), open [Listing Quality](/selling-tools/listing-optimizer-overview/listing-quality.md) and scan your products there first to populate the catalog data, then return to the Agent.
{% endhint %}
***
## The stat cards
Once a scan finishes, four cards summarize your catalog at a glance:
| Card | What it counts |
| -------------------- | ---------------------------------------------------------------------------------------- |
| **Total Products** | Every product returned by the scan |
| **Critical Issues** | Products that have at least one Critical issue (a missing title, description, or images) |
| **Warnings** | Products that have warnings but **no** Critical issue |
| **Perfect Listings** | Products with no issues at all across the known signals |
The same buckets appear as filter tabs above the product list: **All**, **Critical**, **Warnings**, and **Perfect**. A product with any Critical issue lands in the Critical bucket; a product with only warnings lands in Warnings; a clean product lands in Perfect. The list is sorted by severity first, so the most damaging listings (for example, a missing title) always rise to the top.
When a scan covers fewer products than your whole catalog, a line appears above the grid - **"Scanned {n} of {m} products - data as of {date}"** - so you always know how much of the catalog the current view reflects and how fresh it is.
***
## 🔧 Step by step: scan, fix, and push
{% stepper %}
{% step %}
### Scan the catalog
Click **Scan Catalog** in the top toolbar. The Agent walks your synced products and buckets them into All / Critical / Warnings / Perfect. Use the **search box** to filter by ASIN or title at any time.
{% endstep %}
{% step %}
### Pick a bucket
Click the **Critical** tab to work on the products that are hurting you most, or use **Warnings** for lighter clean-up. Each row shows the product image, title, ASIN and SKU, a row of per-field health chips (Title, Bullets, Description, Images, A+, Video), and the quality score ring.
{% endstep %}
{% step %}
### Open the Fix drawer
Click **Fix** on any product with issues (or click the row) to open the **Fix drawer** on the right. Issues are listed most-damaging first, each as a recommendation card explaining what is wrong.
{% endstep %}
{% step %}
### Generate with AI per field
For each fixable field, click **Generate with AI**. The Agent drafts replacement copy using your product context and shows it inline for review. AI drafting in the Agent covers **Title**, **Bullet Points**, and **Description**. Image and A+ gaps are surfaced in the drawer as findings, but you create those in [Listing Quality](/selling-tools/listing-optimizer-overview/listing-quality.md) or the AI Image Studio, not from the Agent's Generate button.
{% endstep %}
{% step %}
### Update on Amazon
When you are happy with the drafts, click **Update on Amazon** at the top of the drawer. A modal opens listing the elements you can push: **Title**, **Bullet Points**, **Description**, **Keywords**, and **Images**.
{% endstep %}
{% step %}
### Choose fields and Save
Tick the elements you want to send, then click **Save**. Only fields you actually generated content for get pushed, if you tick an element with no draft behind it, the Agent tells you to generate content first rather than sending an empty change. SellerMagnet then submits the selected fields to Amazon as a batch.
{% endstep %}
{% endstepper %}
{% hint style="info" %}
**Fields the Agent can draft vs. push.** The Generate button drafts Title, Bullet Points, and Description. The update modal lists all five elements for completeness, but a field is only submitted when it has AI-generated (or edited) content behind it. For image and A+ work, and for the full deep editor, use [Listing Quality](/selling-tools/listing-optimizer-overview/listing-quality.md).
{% endhint %}
***
## 🤖 AI modes
Behind the scenes the AI runs on pre-built prompt templates with four **modes**. The Agent's one-click **Generate with AI** button always uses **Optimize** (it improves the field you already have). The other modes are available in the deeper [Listing Quality](/selling-tools/listing-optimizer-overview/listing-quality.md) AI Studio.
| Mode | Use it to |
| -------------- | ------------------------------------------------------------------------------------------------------------- |
| **Create** | Write a field from scratch when there is nothing to start from |
| **Optimize** | Clean up and SEO-improve your existing title, bullets, or description (this is what the Agent uses) |
| **Competitor** | Generate copy designed to outrank a rival, paste a competitor title as the target to beat |
| **Translate** | Localize the field into the marketplace's language with native-sounding copy, not literal machine translation |
### Amazon style-guide compliance is built in
Every prompt has Amazon's title and content rules baked in, so the output is compliant by construction rather than needing a manual clean-up:
* **Title Case** (capitalize all words except short prepositions)
* **Character limits** per field (titles around 80-150 characters, bullets 150-250 each and under 1,000 total)
* **No promotional language** (no "best seller", "#1", "guarantee", and similar claims)
* **No emojis** and no forbidden special characters (`~ ! * $ ? _ { } # < > ^`)
* **Brand-first structure** (Brand + Product Name + Key Feature + Size/Color/Variant)
* **Front-loaded keywords** so the most important search terms come first
***
## 🔒 Push safety
The Listing Agent is deliberately draft-first. Nothing you generate touches your live listing until you take an explicit action.
{% hint style="warning" %}
**Nothing is sent to Amazon until you push it.** Every AI draft lands in the Fix drawer as a **suggestion**, never an applied change. Your live listing only changes after you open **Update on Amazon**, tick specific elements, and click **Save**.
{% endhint %}
{% hint style="info" %}
**Amazon queues submissions, it does not apply them instantly.** A successful push means Amazon **accepted your submission into its queue**. The listing updates once Amazon finishes validating it, and Amazon can still **reject** a change during that step. You will see the confirmation: "Changes submitted to Amazon and queued for processing. Amazon can still reject them during validation."
{% endhint %}
Additional safeguards on every push:
| Safeguard | What it does |
| ------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
| **Text-policy pre-flight** | Before anything is submitted, each field is screened for known-bad wording. Policy-blocked content is stopped locally and reported to you, not sent to Amazon. |
| **Marketplace ownership check** | The Agent confirms the target marketplace belongs to your connected Amazon account before submitting. |
| **Per-field results** | Amazon's own validation wording is shown per field in the modal, so a partial failure tells you exactly which element was refused and why, rather than a single generic error. |
| **Change log** | Every applied change is recorded (old value and new value), so there is an audit trail of what was sent and when, and changes can be reverted. |
***
## Listing Agent vs. Listing Quality
Both tools live in the Listing Manager AI suite and both can push to Amazon, but they are built for different jobs.
| | **Listing Agent** | **Listing Quality** |
| ------------- | ------------------------------------------------------------ | ------------------------------------------------------------- |
| **Built for** | Fast bulk triage across many listings | Deep work on a single listing |
| **Scope** | Whole catalog at once, bucketed by severity | One product at a time |
| **Checks** | Six content signals per product | 37 detailed checks with sub-scores |
| **Editing** | One-click AI draft per field, batch push | Full editor, AI Studio, keyword and history tabs, image tools |
| **Best when** | You want to sweep the catalog and fix the worst gaps quickly | You want to perfect a specific high-value listing |
Use the Agent to find where the damage is and clear the obvious gaps in bulk; drop into [Listing Quality](/selling-tools/listing-optimizer-overview/listing-quality.md) when a listing needs careful, detailed work.
***
## ⚠️ Caveats
* **Per-account rollout.** The Listing Manager AI suite rolls out per account. While yours is pending you'll see an amber hourglass next to it in the Listing-Manager menu and a "you're on the list" note when you open it - at no charge until you decide to keep it.
* **Requires a connected Amazon account.** The Agent reads your synced catalog and pushes through the Amazon API, so an active Amazon connection is required.
* **Needs Listing Quality data first.** The scan reads the latest Listing Quality snapshots. If nothing has been scanned yet, populate the data from [Listing Quality](/selling-tools/listing-optimizer-overview/listing-quality.md) before the Agent can show your products.
* **AI runs on platform credits, no API key needed.** The AI uses SellerMagnet's own models; you do not bring or pay for your own OpenAI or Gemini key. Drafting draws from a monthly **credit allowance** included with your plan, **shared with the** [**Bulk Sourcer**](/selling-tools/product-sourcer-overview/bulk-sourcer.md). If the balance runs out, the Generate button stops until the allowance refills.
***
## ✅ Best Practices
1. **Work the Critical tab first.** It surfaces the listings actively losing you conversion and discoverability, a missing title, description, or images. Fixing these delivers the biggest real-world gains.
2. **Sort by revenue in your head.** The Agent ranks by severity, but you decide the order. Clear the worst gaps on your highest-selling ASINs before the long tail.
3. **Always read the AI draft before you push.** Treat every draft as a strong first version, not a finished one. Read it, tweak if needed, then submit.
4. **Push only the fields you generated.** Ticking Keywords or Images without a draft behind them does nothing useful, generate the content first, or handle those fields in Listing Quality.
5. **Re-scan after Amazon processes the change.** A push is queued, not instant. Re-scan later to confirm the field cleared and the score climbed.
6. **Keep an eye on your credit balance.** Drafting spends from the monthly allowance shared with the Bulk Sourcer; the remaining balance is shown in the Listing menu. See [what AI actions cost](/selling-tools/listing-optimizer-overview.md) for the per-action credit costs.
***
## ❓ FAQ
Why is my Listing Agent empty even though I have products?
The Agent reads the latest Listing Quality snapshot for each product, it does not fetch fresh data from Amazon on its own. Open [Listing Quality](/selling-tools/listing-optimizer-overview/listing-quality.md), scan your products there, then come back to the Agent and it will show your catalog.
What is the difference between a Critical issue and a Warning?
Critical issues are gaps that actively suppress conversion or discoverability, a missing title, a missing description, or no images. Warnings are content that exists but is weak (too short, or fewer than five bullets) and worth improving when you have time. A product with any Critical issue lands in the Critical bucket; one with only warnings lands in Warnings.
Which fields can the Agent draft with AI?
Title, Bullet Points, and Description. The update modal also lists Keywords and Images for completeness, but the Agent's Generate button does not create those, use [Listing Quality](/selling-tools/listing-optimizer-overview/listing-quality.md) and the AI Image Studio for image and keyword work.
Do my changes go live on Amazon immediately?
No. When you push, Amazon accepts the change into its queue and applies it after validating, and it can still reject a change during that step. SellerMagnet logs every submission and records the old and new values so changes can be reverted. Treat it like submitting an edit in Seller Central, not an instant edit.
Is the AI output Amazon-compliant?
The prompts have Amazon's style guide built in, Title Case, character limits, no promotional language, no emojis or forbidden symbols, brand-first structure, and front-loaded keywords, so output is compliant by construction. On top of that, a text-policy pre-flight screens every field for known-bad wording before it is submitted.
Does the AI need my own OpenAI or Gemini key, and does it cost extra?
No key needed. The AI runs on SellerMagnet's own models and is included in your plan. Usage draws from a monthly credit allowance that is shared with the Bulk Sourcer. If the balance reaches zero, the Generate button pauses until the allowance refills.
When should I use Listing Agent instead of Listing Quality?
Use the Agent for fast bulk triage, sweeping the whole catalog, ranking by severity, and pushing several listings quickly. Use [Listing Quality](/selling-tools/listing-optimizer-overview/listing-quality.md) when a single listing needs deep, detailed work with the full 37-check breakdown, editor, and image tools.
***
## ➡️ What's Next?
{% content-ref url="/pages/FiimK849Zo1bXq80LTlA" %}
[Listing Quality](/selling-tools/listing-optimizer-overview/listing-quality.md)
{% endcontent-ref %}
{% content-ref url="/pages/taNm8GCbHtpSjd6CGv43" %}
[Keyword Research](/selling-tools/listing-optimizer-overview/keyword-research.md)
{% endcontent-ref %}
{% content-ref url="/pages/VXlGkA4PfkAz77id3Txp" %}
[Listing Optimizer](/selling-tools/listing-optimizer-overview.md)
{% endcontent-ref %}