Check any text for ChatGPT and GPT-4 output with a single POST request. You get a verdict, a confidence score, and a score for every sentence — so your reviewers see which passages triggered the flag instead of a number they have to take on faith.
1,000 free credits · no credit card · same endpoint as the text detection API
/api/v1/detect/textcurl -X POST https://www.wasitaigenerated.com/api/v1/detect/text \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"content": "Certainly! Here is a comprehensive overview..."
}'{
"isAI": true,
"confidence": 0.94,
"model": "tropa-3",
"patterns": ["AI patterns detected"],
"analysis": {
"likelihood": "AI-generated",
"reasoning": "Overall AI score: 94.1%"
},
"sentences": [
{
"text": "Certainly! Here is a comprehensive overview...",
"isAI": true,
"confidence": 0.98,
"scores": { "ai": 0.98, "human": 0.02 }
},
{
"text": "I added the bit about the 2019 outage myself.",
"isAI": false,
"confidence": 0.86,
"scores": { "ai": 0.14, "human": 0.86 }
}
]
}No detector can prove a specific product wrote a specific paragraph. What this endpoint measures is whether the text carries the statistical fingerprint of a large language model — and since ChatGPT is what most people are using, that is the question behind the search term.
Build on that honestly: treat a high score as a reason to look closer, not as a verdict. The sentences array exists precisely so a human can see the evidence and make the call.
Every response breaks the text down sentence by sentence, so mixed human/AI documents show up as what they are.
JSON in, JSON out, Bearer auth. Any language that can make an HTTP request is about fifteen lines from done.
1 credit per word, no seat licences, no minimum contract. One-time credits never expire.
Copy, paste your key, ship it
curl -X POST https://www.wasitaigenerated.com/api/v1/detect/text \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"content": "Certainly! Here is a comprehensive overview..."
}'content — string, required. text works as an alias.Authorization — Bearer YOUR_API_KEY.401 — bad or missing key.402 — out of credits.403 — website-only plan.413 — text too long, chunk it.Screen essays and discussion posts on submission and surface the flagged sentences in the instructor's existing grading view.
Check written assessments and cover letters before a recruiter spends time on a candidate who outsourced the task to a chatbot.
Catch bulk-generated forum answers and help-desk replies that look plausible and are confidently wrong.
Enforce a "human-written" guarantee by running every freelancer deliverable through the API as a QA gate.
Score reviews at write time, before generated praise dilutes the trust signal your buyers depend on.
Log a detection result alongside published content to support AI-disclosure obligations.
1 credit per word. A 500-word answer costs 500 credits.
1,000
credits on email verification
≈ 1,000 words of testing. No card required.
$5 – $29
one-time, never expires
25,000 credits for $5, or 200,000 for $29 — roughly 200,000 words.
$299/mo
no per-word cost
Unlimited calls, whitelabel PDF reports, bulk processing.
It tells you the text was machine-generated and how strongly, not which product produced it. There is no reliable way to separate ChatGPT from Claude or Gemini output after the fact — the models are trained on overlapping data and converge on similar statistical patterns. In practice this rarely matters: the question teams actually need answered is "did a human write this", and that is exactly what isAI and confidence answer.
Partially, and this is the honest limitation. Light edits usually leave the signal intact. Heavy rewriting, or a pass through a paraphrasing tool, degrades it — the more a human reworks the text, the more it reads as human, which is arguably correct. The sentence-level breakdown helps here: mixed documents typically show blocks of high-confidence AI sentences next to genuinely human ones.
All covered. Detection works on statistical properties of the text rather than a lookup of known outputs, so newer OpenAI models, custom GPTs and API-driven wrappers are handled without waiting for a model update on our side.
1 credit per word. A 500-word answer costs 500 credits. Verified accounts start with 1,000 free credits; after that 25,000 credits are $5 and 200,000 are $29 as one-time purchases that never expire, or $299/month for unlimited API calls.
Bearer token. Send Authorization: Bearer YOUR_API_KEY with every request. Your key is created with your account and shown in the dashboard. Keep it server-side — it spends your credits.
No. Detection is probabilistic, and a false positive against a real student or a real applicant is expensive. Use the score to route a submission to a human, show that person the flagged sentences, and let them decide. Products that auto-punish on a confidence number generate complaints you cannot defend.
Yes. The API endpoint runs the same tropa-3 model behind the ChatGPT detector on the website, so a result you see in the browser is the result your integration gets.
Yes. All paid plans include commercial use. Contact us if you need volume pricing, an SLA or dedicated rate limits.
Create an account, verify your email, and your API key is waiting in the dashboard.