Image Moderation API for Marketplaces
Automatically screen user-uploaded listing photos on classifieds, car-listing sites, and peer-to-peer marketplaces before they go live.
Published 2026-05-19
TL;DR. To screen images on a marketplace automatically, route every seller upload through an image moderation API before the listing publishes. Pixicular's image analysis API returns content-moderation flags (nudity, violence, weapons) and object labels (counterfeit and prohibited-goods indicators) in one call, so your pipeline approves safe listings instantly and escalates only the rest.
Why marketplaces need automated listing photo screening
On any classifieds site, car-listing platform, or peer-to-peer marketplace, a fraction of seller-uploaded photos will contain prohibited content: explicit imagery disguised as a product listing, weapons and regulated goods, counterfeit brand signals, or content that violates listing quality standards. Manual pre-screening is not viable at scale — a trust-and-safety team cannot review every photo on a platform that ingests tens of thousands of new listings a day, and buyers encounter the bad ones before a human ever sees them.
Automated screening inverts the default. Instead of publishing first and reacting to reports, every image is scored by an API at upload time. The safe majority of listings go live without any human reviewer in the loop. Reviewers focus their time on the ambiguous cases and the dangerous ones — the fraction that the API flags above your policy thresholds. This is how platforms with high listing volume keep moderation costs proportionate while maintaining buyer safety. See also product image moderation for e-commerce for a comparison with retail catalogue workflows.
How the marketplace moderation pipeline works
A typical listing photo moderation pipeline has four stages: a seller uploads a photo, it is sent to the moderation API in the same request that processes the listing, the API returns content-moderation flags and label scores with confidence values per category, and your backend applies per-category thresholds to decide approve, queue, or reject.
The decision step is deliberately yours, not the API's. The API returns scored signals; your trust-and-safety policy determines what confidence on which category triggers an auto-reject, what goes to the queue, and what passes directly to publication. Keeping that separation means moderation behaviour is under your control, auditable, and adjustable without API changes. This matters for the regulatory obligations covered below.
What the moderation and label services return
The detect-moderation service returns a confidence score per unsafe category. The detect-labels service returns the objects, brands, and scene concepts present in the image. Together they cover the categories a marketplace trust-and-safety team screens for.
| What you screen for | Pixicular service | Example on a listing |
|---|---|---|
| Explicit nudity and sexual content | detect-moderation | Adult imagery disguised as a clothing or furniture listing |
| Violence and weapons | detect-moderation | Knives, firearms, or graphic injury in a listing photo |
| Drugs and controlled substances | detect-moderation | Drug paraphernalia visible in a product shot |
| Suggestive or borderline content | detect-moderation | Borderline imagery that needs human review rather than auto-block |
| Counterfeit and brand indicators | detect-labels | Designer logos on suspiciously generic goods, replica-style framing |
| Prohibited objects | detect-labels | Currency, regulated items, or policy-restricted goods visible in the photo |
| Listing quality and relevance | detect-labels | Screenshots, stock-photo backgrounds, or images unrelated to the listed item |
Note: Pixicular's AI-generated image detection service is currently disabled while accuracy is being improved. The content-moderation and label-detection services above are live and are what marketplace platforms use for listing photo screening today.
Code: calling the moderation API from your listing pipeline
One request sends the image and asks for both services. The image is uploaded once; the response is a single JSON document with one key per requested service and confidence scores per category within each.
curl
curl -X POST https://api.pixicular.com/detect \
-H "Authorization: Bearer $PIXICULAR_API_KEY" \
-F "image=@./listing-photo.jpg" \
-F "services=detect-moderation,detect-labels"Example response (abbreviated)
{
"detect-moderation": {
"flags": [
{ "name": "Nudity", "confidence": 0.02 },
{ "name": "Violence", "confidence": 0.01 },
{ "name": "Weapons", "confidence": 0.00 }
]
},
"detect-labels": {
"labels": [
{ "name": "Bicycle", "confidence": 0.97 },
{ "name": "Sports Equipment", "confidence": 0.89 },
{ "name": "Outdoor", "confidence": 0.82 }
]
}
}Your backend compares the confidence scores against per-category thresholds you define. See the Pixicular API documentation for the full response schema, authentication, and exact field names per service.
API cost vs an in-house moderation team
The economics of marketplace image moderation favour an API model heavily, especially at the volumes classifieds platforms encounter.
| Approach | Monthly cost (100k images) | Notes |
|---|---|---|
| Pixicular API (Scale plan) | ~$99 | Scales linearly with volume; human reviewers only see flagged cases |
| 1 FTE moderator + overhead | $4,000–$6,000 | Fixed cost regardless of volume; limited throughput; psychological support required |
| 2 FTE moderators + overhead | $7,000–$12,000 | Throughput doubles but cost scales with headcount, not API calls |
The API does not eliminate the need for human moderators — it dramatically reduces the volume they see. The API handles the safe majority automatically; your moderators focus on the borderline cases the API queues and on appeals. For a 100k-image-per-month marketplace, even a conservative 95 percent automated clearance rate means reviewers see at most 5,000 items per month rather than 100,000. Check the pricing page for current plan details. Cost estimates above are illustrative; confirm with qualified HR and finance advisors.
Regulatory context: DSA Article 22 and Article 16
The EU Digital Services Act places image moderation squarely inside a compliance perimeter for online marketplaces. Article 16 requires all hosting providers and online platforms — including classifieds and peer-to-peer marketplaces — to operate a notice-and-action mechanism: any person or entity must be able to flag content they consider illegal, and the platform must process those notices in a timely, diligent, non-arbitrary and objective way, then notify the reporter of its decision and the available redress channels.
Article 22 goes further for very large online platforms (100 million or more EU users), requiring them to deploy specific automated tools to detect and remove illegal content as part of their systemic risk mitigation. Automated image moderation is the detection layer that feeds this obligation.
Automated screening strengthens your compliance posture in three concrete ways. First, proactive detection reduces the volume of illegal imagery that ever reaches buyers, shrinking your notice queue and the exposure window before removal. Second, every screened listing carries stored moderation and label scores, which is the evidence trail your team needs to action a notice quickly and produce the statement of reasons Article 17 requires when content is removed. Third, consistent automated thresholds make decisions non-arbitrary and reproducible — exactly the standard regulators expect when they audit moderation behaviour. Pair the API with a human review queue for borderline cases and you have a defensible, auditable pipeline rather than a reactive backlog. Learn more about content moderation principles on the what is image content moderation page.
This is general information, not legal advice. Confirm your obligations and the applicability of DSA provisions to your platform with qualified legal counsel.
When to auto-reject, queue for review, or approve
- Auto-reject on high-confidence unsafe categories: explicit nudity, graphic violence, and weapon imagery above a strict threshold should never be seen by buyers. These are the cases where you have high confidence the content violates policy and the risk of showing it outweighs the risk of a false positive.
- Queue for review on borderline confidence, on counterfeit-indicator labels, or on categories where your policy requires a human decision before removal. The reviewer sees the scored evidence alongside the image, which makes decisions faster and more consistent.
- Approve automatically when no category exceeds its threshold and the detected labels match the item category the seller declared. The listing publishes immediately, with no human in the loop for a photo that passes all checks.
- The same decision model extends beyond listing photo moderation — see how it applies in a broader context on the image content moderation overview page.
Frequently asked questions
How do you screen images on a marketplace automatically?
Route every seller-uploaded listing photo through an image moderation API before publication. Pixicular's detect-moderation service scores the image for nudity, violence, weapons, and drugs, while detect-labels identifies the objects and scene concepts present. Your backend compares those scores against per-category thresholds you own and routes the listing to approve, queue-for-review, or reject — without a human reviewing the safe majority.
Which Pixicular services should a marketplace use for listing photo moderation?
Two services cover the core screening needs: detect-moderation for unsafe imagery (nudity, explicit sexual content, violence, weapons, drugs, and suggestive material) and detect-labels for objects, brands, and scene concepts that surface prohibited-goods and counterfeit indicators. Both run in a single POST /detect request, so every listing photo costs one API call rather than several.
Does automated image moderation satisfy DSA Article 22 obligations?
DSA Article 22 requires very large online platforms (100 million or more EU users) to deploy specific automated tools to detect and remove illegal content as part of their systemic risk mitigation. Automated image moderation is the detection layer that feeds those obligations. For all platforms covered by Article 16, storing per-listing moderation scores gives your team the evidence trail needed to action notices quickly and produce the Article 17 statement of reasons. This is general information, not legal advice.
How much does an image moderation API cost compared to an in-house moderation team?
At 100,000 listing images per month, Pixicular's Scale plan costs around $99/month. Two full-time human moderators covering that volume typically cost $7,000–$12,000/month in salary and overhead, before factoring in training, management, and psychological-support costs. The API handles the safe majority automatically; human reviewers then focus only on borderline and high-risk cases the API flags for review.
Can image moderation detect counterfeit or prohibited goods on a marketplace?
Image moderation does not make a legal counterfeit determination, but detect-labels surfaces strong indicators — designer logos on generic products, replica-style framing, weapons, currency, or regulated items — that you can combine with listing text and seller history to flag for human review. The detect-moderation service handles unsafe imagery such as nudity and violence independently. The two services complement each other in a single API call.
Add image moderation to your marketplace listing pipeline
The fastest way to evaluate Pixicular for marketplace photo screening is to point a curl request at it with a real listing photo. Pick a plan on the pricing page and follow the API documentation for authentication and the full response schema for detect-moderation and detect-labels.