Poetry in the Age of Artificial Intelligence

Allow ourselves to introduce… ourselves. In a world flooded with synthetic media, the ability to separate human‑written content from AI‑generated text is becoming essential. IDENTIF.AI is an open, zero‑shot text detection API built for developers, journalists, and security teams who need reliable AI content verification without complex setup or training data.

Soooo… how does this thing work?

IDENTIF.AI is powered by a technique known as perplexity analysis. It measures how “surprised” a large language model is by a given text. If a model predicts each word with high confidence, the text is likely AI‑generated; if the model struggles to predict the next word, it’s more likely written by a human.

Under the hood, our API loads a lightweight transformer and computes a normalized “authenticity score” — lower scores indicate more AI‑like writing. This statistical signal is scaled and compared to a configurable threshold (default 0.9) to produce a verdict:

• ai_generated — high probability of synthetic origin
• human_written — likely composed by a human author

Each detection includes confidence, score, and metadata such as text length and processing time, making the output easy to integrate into moderation or analytics pipelines.

Quickstart Demo

You can test IDENTIF.AI instantly using curl or Python. The API expects a JSON payload and an API key provided in either the Authorization or X‑API‑Key header.

Test with cURL

curl -s -X POST "https://api.identif.ai/api/v1/detect" \
  -H "Authorization: Bearer dev-key-12345" \
  -H "Content-Type: application/json" \
  -d '{
    "text": "This paragraph might have been written by an AI language model... BEEP BOP BOOP! Oh, uhh *ahem*... Sorry, excuse me. Sometimes I make those noises. Nothing to be concerned about. I swear I'm definitely NOT an Ai or a robot or a poorly trained machine learning model. I love fresh air, and hamburgers, and long walks on the...beach. Using my feet. Because I'm a human. And stuff. Anyway, let's talk about something else.",
    "threshold": 0.9
  }'
  

The API will respond with a structured JSON result like:

{
  "verdict": "ai_generated",
  "confidence": 0.42,
  "score": 0.37,
  "is_ai_generated": true,
  "model_used": "identif.ai",
  "processing_time_ms": 12.8,
  "metadata": {
    "text_length": 68,
    "word_count": 12,
    "threshold": 0.9
  }
}

Use the Python Client

With Python and the requests library:

import requests

API_KEY = "dev-key-12345"
text = "This paragraph might have been written by an AI language model."

response = requests.post(
    "https://api.identif.ai/api/v1/detect",
    headers={"Authorization": f"Bearer {API_KEY}"},
    json={"text": text, "threshold": 0.9}
)

print(response.json())

Batch Detection

Need to check multiple documents at once? Send them as an array:

{
  "texts": ["Text one.", "Text two."],
  "threshold": 0.9
}

The response will summarize AI vs. human counts and average scores for all items.

Interpreting Results

The key output fields are:

• verdict — either ai_generated or human_written
• confidence — strength of the verdict (0.0–1.0)
• score — normalized perplexity (lower = more likely AI)

Wanna give it a whirl? Sign up for a demo to get your own API key FREE!