Industry-leading accuracy
Advanced AI detection model trained on 2 billion+ samples. Catches ChatGPT, Claude, Gemini, Llama, Mixtral, DeepSeek, Qwen, GLM, and paraphrased rewrites. Confidence band returned on every response.
The AI content detection API engineering teams will keep
Industry-leading accuracy. Trained on 2 billion+ samples. Real-time analysis. OpenAPI 3.1 spec, official JavaScript + Python SDKs, webhook fan-out. Public API launching soon — sign up to be first in line.
import { createClient } from '@deepaidetector/client';
const client = createClient({ apiKey: process.env.DEEPAIDETECTOR_KEY });
const r = await client.detect({ text: 'Paste the prose you want to check...' });
console.log(r.score, r.band);
// 0.873 'definitely_ai'Per-sentence verdicts
Sentence-level breakdown returned on every response. Show users exactly which lines look AI-generated. Defensible in academic, editorial, and HR contexts.
20+ languages, automatic
Submit any supported language — we detect the source and score it. No flag, no per-language quota. The verdict is for the original text.
Multi-source plagiarism
Plagiarism detection across the open web, academic databases, and reference works. One JSON call returns matched passages with similarity scores.
Webhooks
detection.completed, plagiarism.completed, subscription.*, balance.low. HMAC-SHA256 signed, 7 retries over 72h.
Pricing quote on every call
Every response shows per-word vs per-detection cost — pick the cheaper mode automatically.
SDKs
| Language | Package | Install | Status |
|---|---|---|---|
| JavaScript / TypeScript | @deepaidetector/client | npm i @deepaidetector/client | GA |
| Python | deepaidetector | pip install deepaidetector | GA |
| Go | — use net/http | See API docs sample | Sample |
| Ruby | — use Net::HTTP | See API docs sample | Sample |
| PHP | — use cURL | See API docs sample | Sample |
| C# | — use HttpClient | See API docs sample | Sample |
What you can build
- Classroom integrity tooling. Bulk-scan student submissions; surface the per-passage breakdown to make academic appeals defensible.
- Content marketplaces. Vet contributor submissions before they hit the editorial queue. Webhook fan-out into Slack / Linear.
- Editorial QA. Score every article before it ships; flag anything above your AI-percent threshold for human review.
- Affiliate / SEO platforms. Check syndicated articles against your AI policy in real time. Cache hits are free.
- Customer support quality. Detect canned AI-generated agent replies in QA samples.
- Trust & safety pipelines. Score user-generated content; pair with plagiarism to identify scraped/repurposed material.
Why teams pick the Deep AI Detector API
Designed for AI content detection workflows, not bolted on
Deep AI Detector is an AI content detector first. The REST API exposes the same advanced detection model that runs in the web dashboard — a model trained on 2 billion+ samples covering ChatGPT, Claude, Gemini, Llama, Mixtral, DeepSeek, Qwen, and GLM, plus paraphrased and humanized AI text that legacy detectors classify as 97%+ human. The API was built for product teams who need per-sentence verdicts inside their own surface, not for a vendor who needed to put "API access" on a feature list.
Stable contracts, conservative versioning
v1 is frozen against breaking changes for at least six months before deprecation.
Additive minors (new fields, new endpoints, new event types) ship every few weeks and are
documented in the changelog. When a major bumps, we ship v2 shoulder-by-shoulder with the prior version and email every API key holder on the
deprecation timeline. You can integrate against the spec today and trust the contract for at least
18 months.
Pricing quote on every response
Every detection response includes a pricing_quote object showing both billing modes
(per-word and per-detection) and which one is cheaper for the document just submitted. This lets
you build cost preview UIs without round-tripping through a separate pricing endpoint.
Privacy-first by default
Submitted text is deleted from logs and cache within 24 hours unless the account explicitly
opts into history. Cache keys are sha256(text) — even our cache never stores plaintext.
We do not train on customer submissions without per-product opt-in.
How to integrate the Deep AI Detector API
Most integrations land in production in under a day. The path below is what we recommend for new builds.
- Get a Pro tier or higher key. Pro ($34.99/month) is the entry tier for API access. Studio, Business, Scale, Premier, and Enterprise scale up the quota and word limits.
- Drop in an SDK.
@deepaidetector/clientfor JavaScript / TypeScript,deepaidetectorfor Python. Both wrap the OpenAPI spec, ship type definitions, and handle retries plus error mapping. - Call
/v1/detect. Pass the text, get back a score, a band, a per-sentence breakdown, and the pricing quote. - Wire up webhooks. Configure a webhook URL in your dashboard to receive
async events:
detection.completed,plagiarism.completed,subscription.*,balance.low. - Cache aggressively. Identical text returns a cached verdict at no rate-limit cost. Re-checking edits is free.
Spec-first
The OpenAPI 3.1 spec at /openapi.yaml is the source of truth. Both SDKs are generated from it; the human docs, Postman collection, and SDK types are kept in lockstep.
Versioning is conservative: v1 is frozen for at least 6 months before
deprecation, and we ship a v2 shoulder-by-shoulder with the old version.
Subscribe to the changelog for additive changes.
Public API launching soon
We're finishing the rate-limit infrastructure and SDK polish. Drop your email and we'll send the launch announcement (with starter credit) the moment keys go live.
Today: the web dashboard, exports, per-passage highlights, and word-pack credits all work across the paid tiers.