1
Send listing images
Send images as sellers upload them, index existing inventory in batch, or use both.
Send listing imageswith your listing metadata
Similarity AI by NewByte Solutions
Stop scam listings built with stolen photos.
Fraudsters reuse photos from legitimate car, rental, and classified listings to make fake offers look real. Similarity AI finds those images before they reach your users.
Detect reused and edited listing imagesRoute high-confidence matches to reviewIndex images individually or in bulk
Edited copyGeometric evidence
From listing image to review decision
Evidence before a suspicious listing reaches your users.
Give your Trust & Safety team the context they need to act in three steps.
2
Find reused photos
Compare each image against your marketplace inventory and return likely matches.
Edited copygeometric evidencegeometric_copy
3
Route matches to review
Use match tiers, geometric evidence, and your rules to send relevant matches to review.
Likely reusedReview evidence
What we detect
A copied photo is rarely an exact copy.
Scammers crop, resize, recolor, and edit images to avoid simple duplicate checks. Similarity AI is designed to find the underlying image relationship.
Exact copy
The same listing photo appears again.
Crop or resize
A seller uses part of an existing image or changes its dimensions.
Color or lighting edit
Brightness, contrast, saturation, and filters change the look but not the source.
Watermark removal
A listing image is reused after branding or attribution has been removed.
Mirrored image
A horizontal flip is used to create a superficially different photo.
Edited or reused image
A photo has been modified but still points back to an earlier listing.
Get in touch
Let's find the right fit for your marketplace.
Tell us about your listings and what you want to improve. Our team can help you choose a practical next step.
Tell us what you need
Share your marketplace, image volume, and the problem you want to solve.
Review the right approach
We can discuss detection, review workflows, pricing, and a focused pilot.
Plan the next step
Talk through API integration or a custom rollout with our team.
Why not build it yourself?
Your team can build image matching. That is not the whole job.
Open-source embeddings and a vector database can produce similar-looking-image results. Turning those results into a dependable fraud-control workflow takes more.
Let engineering focus on your marketplace.Avoid a new internal ML product.
What your team needsBuild internallySimilarity AI
Time to first usable workflowModel selection, image pipeline, indexing, search, and review toolingAPI, batch indexing, and review-ready match data
Crops and editsOngoing evaluation and tuningTuned for reused, cropped, mirrored, and edited images
Marketplace contextBuild tenant separation and metadata handlingTenant-aware workflows and flexible metadata
Review workflowDesign queues, confidence rules, and integrationsMatch tier, geometric evidence, and API results
MaintenanceMonitor models, infrastructure, re-indexing, and edge casesNewByte Solutions manages the detection layer
Built for the way marketplace teams ship
Integrate where listing images enter your product.
Index your catalog, then search by image URL or indexed image ID. Inspect match tiers and geometric evidence, and filter results using your catalog fields.
Metadata and attribute filters for your listingsCatalog updates and deletion without extra usage chargesAccount-scoped catalogs for separate marketplaces or brands
POST /v1/images/match
curl 'https://similarity-api.newbytesolutions.com/v1/images/match' \
-H 'Authorization: Bearer YOUR_TENANT_API_KEY' \
-H 'Content-Type: application/json' \
-H 'Idempotency-Key: REPLACE_WITH_A_FRESH_UUID' \
--data-raw '{
"images": [
{
"query_id": "listing-query",
"image_url": "https://example.com/listing-photo.jpg"
}
],
"top_k": 5,
"filters": []
}'200 response · illustrative example
{
"request_id": "EXAMPLE_REQUEST_ID",
"results": [
{
"item_index": 0,
"query_id": "listing-query",
"status": "succeeded",
"matches": [
{
"image_id": "CATALOG_IMAGE_ID",
"external_id": "listing-123",
"image_url": "https://example.com/catalog-photo.jpg",
"metadata": {
"title": "Catalog listing"
},
"attributes": {},
"embedding_cosine": 0.96,
"phash_distance": 8,
"phash_similarity": 0.875,
"ranking_score": 0.96,
"decision": "match",
"match_reason": "geometric_copy",
"tier": "edited_copy",
"normalized_cosine": null,
"shadow_evaluation": null,
"verification": {
"inliers": 42,
"inlier_ratio": 0.68,
"inlier_coverage": 0.47,
"transform_scale": 1.1,
"rotation_degrees": 2.4,
"transform_plausible": true,
"flipped": false
},
"combined_score": 0.93875
}
],
"needs_verification": []
}
],
"summary": {
"succeeded": 1,
"failed": 0,
"billable": 1
},
"profile_id": "similarity-v1",
"score_policy_id": "similarity-score-v3",
"pipeline_mode": "active"
}Pricing that supports a serious review workflow
Clear commercial terms. Built for scale.
Start with guided onboarding, meet a clear monthly minimum, and pay only for the image work your marketplace uses.
Standard onboarding
$1,500A 1 to 2 business day integration window for API setup, metadata mapping, and a first successful match.
Monthly minimum
$299Applied against usage. It is not an additional platform fee.
Usage
$0.01
per image indexed
$0.002
per image searched
Bulk indexing is available for large catalogs.
See full pricing and pilot termsData handling and control
Your marketplace data stays under your control.
Tenant-isolated data
Customer data is logically separated by tenant.
HTTPS in transit
Images are sent over HTTPS. Similarity AI does not retain original customer images after creating a visual vector.
Delete vectors and match records
Customers can request deletion of stored visual vectors and associated match records.
No training on customer data
Customer images and visual vectors are not used to train shared models.
FAQ
Questions from Trust & Safety and engineering teams.
How do we control false positives?
Similarity AI uses provisional near-exact and geometric verification rules, with a separate review band. Tune the rules on labeled examples from your marketplace before setting an operating target.
Can it find a photo after it has been cropped or edited?
Yes. Similarity AI is designed to identify exact copies as well as likely relationships across crops, resizes, mirrored images, color edits, watermark removal, and other common listing-image changes.
How fast are batch checks?
In current rental-marketplace use, a 10-image batch typically returns in 1 to 2 seconds. Actual timing depends on the image mix and workload.
Which image formats do you support?
Similarity AI currently supports JPEG, PNG, and WebP images.
How long do you retain vectors and match data?
Visual vectors and match records remain while your account is active and they are relevant to search, unless you request deletion.
How long does integration take?
A standard API setup typically takes 1 to 2 days, based on current implementation experience.
Can we start with a pilot instead of indexing our full marketplace?
Yes. Pilots use standard onboarding and usage pricing. Scope, success criteria, and the expansion path are agreed before the pilot begins. Start with a defined listing segment, region, or category.
What happens after a pilot?
Customers who continue expand to more listings or categories under the published onboarding, monthly minimum, and usage pricing.
Also used for
Short-term rental and travel listings
Insurance claim review
E-commerce catalog quality
Online communities and image-spam prevention
Similarity AI by NewByte Solutions
Find the scam listings your current checks miss.
Tell us about your marketplace and the image issues you want to solve. We will discuss the right next step with you.