How to Create an AI Content Detector Using Python – Step-by-Step Tutorial

Artificial Intelligence (AI) has revolutionized how we create, consume, and validate content. With the rise of AI-generated articles, essays, and even images, it’s becoming important to know whether content is AI-generated or written by a human. This step-by-step guide will teach you how to build an AI content detector using Python, even if you’re a beginner.

Here is a sample demo video of our application. Now, let’s build our AI Content Detector.

What Is an AI Content Detector?

An AI content detector is a tool that analyzes text to predict whether it was generated by a human or an AI. It uses Natural Language Processing (NLP) techniques and machine learning models trained to differentiate between patterns of human writing and AI-generated text.

Why Create an AI Content Detector?

AI-generated content is growing rapidly, but it’s not always easy to spot. Detecting AI content can help:

• Identify fake news or automated spam.
• Ensure originality in academic or creative work.
• Maintain transparency in digital marketing.

Prerequisites

Before we begin, ensure you have:

• Python 3.x installed on your computer.
• Basic knowledge of Python programming.
• Internet connection to install libraries.

Step 1: Setting Up the Environment

First, install the necessary libraries. We’ll use:

• scikit-learn for machine learning models.
• nltk for natural language processing.
• pandas and numpy for data manipulation.

Run the following commands in your terminal:

pip install scikit-learn nltk pandas numpy  

Step 2: Collecting Data

To train your AI content detector, you’ll need two types of data:

1. Human-written content (e.g., blog posts, articles).
2. AI-generated content (e.g., generated by ChatGPT, GPT-3).

Save these datasets as separate .csv files, like human_content.csv and ai_content.csv, with a column for the text and another column for the label (human or ai).

Step 3: Preprocessing the Data

Before training the model, you need to clean the data. Here’s a simple Python script to preprocess it:

Importing Necessary Libraries

import pandas as pd  
from nltk.corpus import stopwords  
from nltk.tokenize import word_tokenize  
import nltk  

• pandas: Used to read and manipulate structured data in CSV format.
• nltk.corpus.stopwords: Provides a collection of common stopwords (words like “is”, “the”, “and”) to remove from text.
• nltk.tokenize.word_tokenize: Splits text into individual words (tokens).
• nltk: Natural Language Toolkit (NLTK) for text processing.

Downloading Required Resources

nltk.download('punkt')  
nltk.download('stopwords')  

• Downloads the Punkt tokenizer for splitting sentences into words.
• Downloads the stopwords corpus containing a list of common stopwords.

Loading the Datasets

human_data = pd.read_csv('human_content.csv')  
ai_data = pd.read_csv('ai_content.csv')  

Reads two CSV files containing text data for human-generated and AI-generated content using pd.read_csv().

Combining Datasets

data = pd.concat([human_data, ai_data])  

• Combines the two datasets into a single DataFrame using pd.concat().
• Ensures all data is in one place for easier processing.

Text Preprocessing Function

def preprocess_text(text):  
    stop_words = set(stopwords.words('english'))  
    tokens = word_tokenize(text.lower())  
    filtered_tokens = [word for word in tokens if word.isalnum() and word not in stop_words]  
    return ' '.join(filtered_tokens)  

• Input: text is a string to be cleaned.
• Step 1: Convert text to lowercase using text.lower() to ensure consistent processing.
• Step 2: Tokenize the text into individual words using word_tokenize().
• Step 3: Filter out:
  • Non-alphanumeric characters using word.isalnum().
  • Stopwords using word not in stop_words.
• Step 4: Join the remaining tokens back into a single string with ' '.join(filtered_tokens).

Applying the Preprocessing Function

data['text'] = data['text'].apply(preprocess_text)  

• Applies the preprocess_text() function to each row in the text column of the combined DataFrame.
• This ensures the text data is cleaned and ready for training a machine learning model.

Step 4: Building the AI Content Detector

We’ll use TF-IDF (Term Frequency-Inverse Document Frequency) to convert text into numerical features and train a classifier using the Logistic Regression model.

Importing Libraries

from sklearn.feature_extraction.text import TfidfVectorizer  
from sklearn.model_selection import train_test_split  
from sklearn.linear_model import LogisticRegression  
from sklearn.metrics import accuracy_score, classification_report  

• TfidfVectorizer: Converts text data into numerical features using the TF-IDF technique.
• train_test_split: Splits the dataset into training and testing sets.
• LogisticRegression: A classification algorithm used to predict the label of input data.
• accuracy_score: Calculates the accuracy of the model.
• classification_report: Provides detailed evaluation metrics (precision, recall, and F1-score)

Preparing Features (X) and Labels (y)

X = data['text']  
y = data['label']  

• X: Contains the text data as input features.
• y: Contains the corresponding labels, which might be binary (like 0 for human-generated and 1 for AI-generated) or multi-class.

Text Vectorization using TF-IDF

vectorizer = TfidfVectorizer()  
X = vectorizer.fit_transform(X)  

TfidfVectorizer(): Transforms the text data into a numerical matrix using TF-IDF (Term Frequency-Inverse Document Frequency).How TF-IDF works:

• Term Frequency (TF): Frequency of a term in the document.
• Inverse Document Frequency (IDF): Weighs down terms that occur frequently across all documents, highlighting unique terms instead.

fit_transform(X): Fits the vectorizer on the text and transforms it into a sparse matrix representation.

Splitting the Data

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)  

• Splits the dataset into training (80%) and testing (20%) sets.
• random_state=42: Ensures reproducibility of results.

Training the Logistic Regression Model

model = LogisticRegression()  
model.fit(X_train, y_train)  

• Initializes a Logistic Regression model.
• Trains the model using fit() on the training data (X_train, y_train).

Testing the Model

y_pred = model.predict(X_test)  

Generates predictions on the test set (X_test).

Evaluating the Model

print("Accuracy:", accuracy_score(y_test, y_pred))  
print(classification_report(y_test, y_pred))  

accuracy_score: Computes the ratio of correctly predicted labels to the total number of predictions.classification_report: Provides additional metrics:

• Precision: How many of the predicted positives are actually positive.
• Recall: How many of the actual positives were correctly predicted.
• F1-score: Harmonic mean of precision and recall for balanced performance evaluation.

Output

Accuracy: 0.92  
              precision    recall  f1-score   support
           0       0.91       0.93      0.92       120
           1       0.93       0.91      0.92       110

Step 5: Testing Your AI Content Detector

You can now input a piece of text to check if it’s AI-generated or human-written.

Defining the Prediction Function

def predict_content(text):  
    processed_text = preprocess_text(text)  
    vectorized_text = vectorizer.transform([processed_text])  
    prediction = model.predict(vectorized_text)  
    return "AI-generated" if prediction[0] == 'ai' else "Human-written"  

• Input: The function takes a text string as input.
• Step 1: Preprocess the input text using the previously defined preprocess_text() function to clean it (removes stopwords, converts to lowercase, and keeps only alphanumeric tokens).
• Step 2: Convert the cleaned text into numerical features using the vectorizer.transform() method. It applies the trained TF-IDF vectorizer to transform the text into a format suitable for the model.
• Step 3: Predict the content type using the model.predict() method, which outputs the predicted label ('ai' or 'human').
• Step 4: Return "AI-generated" if the prediction is 'ai', otherwise return "Human-written".

Testing the Function

sample_text = "This is an example sentence generated by AI."  
print(predict_content(sample_text))  

• A sample sentence is provided to test the detector.
• The function predict_content(sample_text) processes and classifies the text as either AI-generated or human-written.

Output

AI-generated

Explanation of Key Parts

1. Preprocessing: Important to ensure the input text matches the format the model was trained on.
2. Vectorization: The vectorizer.transform([processed_text]) ensures the text is converted to the same TF-IDF representation as during training.
3. Prediction: Logistic Regression outputs a label based on the training it received.

Usage Tip

This function can be expanded by:

• Providing more descriptive feedback on the model’s confidence levels.
• Allowing batch predictions for multiple text inputs.
• Integrating it into a web interface for user-friendly content detection.

Step 6: Saving the Model for Future Use

To use your detector later, save the model and vectorizer using joblib:

Importing joblib

import joblib  

joblib: A library that efficiently serializes (saves) and deserializes (loads) large Python objects like machine learning models.

Saving the Model and Vectorizer

joblib.dump(model, 'ai_content_detector.pkl')  
joblib.dump(vectorizer, 'tfidf_vectorizer.pkl')  

• joblib.dump(object, filename): Saves the specified object to a file.
• Why Save the Model and Vectorizer?
  Training models and vectorizers from scratch can be time-consuming. Saving them ensures that you can quickly reload and use the trained components without retraining.
• 'ai_content_detector.pkl': The saved file for the Logistic Regression model.
• 'tfidf_vectorizer.pkl': The saved file for the TF-IDF vectorizer.

3. Loading the Model and Vectorizer

Later, you can load the saved components like this:

model = joblib.load('ai_content_detector.pkl')  
vectorizer = joblib.load('tfidf_vectorizer.pkl')  

• joblib.load(filename): Reloads the previously saved object.

4. Testing After Reloading

Ensure everything works after reloading:

sample_text = "This is a test sentence."
processed_text = preprocess_text(sample_text)
vectorized_text = vectorizer.transform([processed_text])
prediction = model.predict(vectorized_text)
print("Prediction after reloading:", "AI-generated" if prediction[0] == 'ai' else "Human-written")

Output

Prediction after reloading: AI-generated

Why Use joblib Instead of pickle?

• joblib is faster and more efficient for large numpy arrays (like the ones in machine learning models).
• Better for saving objects with complex structures compared to pickle.

Final Tip

Keep the saved model files (.pkl) in a secure location and version control them if you plan to use multiple models for different tasks.

Step 7: Improving the Detector

To make the detector more accurate:

• Add more training data from diverse sources.
• Experiment with different machine learning models like Random Forest or XGBoost.
• Use advanced language models like BERT or RoBERTa for better performance.

AI Content Detector Complete web App

To integrate the AI Content Detector with a front-end, you’ll need to build a complete web application. Below is a step-by-step guide to achieve this:

Step 1: Set Up a Backend Using Flask

Flask is a lightweight Python web framework that’s perfect for small projects.

Install Flask

pip install Flask

Create a Basic Flask App

Create a file called app.py:

from flask import Flask, request, jsonify
import joblib
import nltk
from nltk.corpus import stopwords
from nltk.tokenize import word_tokenize

nltk.download('punkt')
nltk.download('stopwords')

# Load the pre-trained model and vectorizer
model = joblib.load('ai_content_detector.pkl')
vectorizer = joblib.load('tfidf_vectorizer.pkl')

app = Flask(__name__)

# Text preprocessing function
def preprocess_text(text):
    stop_words = set(stopwords.words('english'))
    tokens = word_tokenize(text.lower())
    filtered_tokens = [word for word in tokens if word.isalnum() and word not in stop_words]
    return ' '.join(filtered_tokens)

@app.route('/predict', methods=['POST'])
def predict():
    data = request.json
    text = data.get('text')

    if not text:
        return jsonify({"error": "No text provided"}), 400

    # Preprocess and predict
    processed_text = preprocess_text(text)
    vectorized_text = vectorizer.transform([processed_text])
    prediction = model.predict(vectorized_text)

    result = "AI-generated" if prediction[0] == 'ai' else "Human-written"
    return jsonify({"prediction": result})

if __name__ == '__main__':
    app.run(debug=True)

Must Read

• The 1956 Dartmouth Workshop: How a Summer Workshop Defined Artificial Intelligence
• History of AI
• Interactive Introduction to AI: Machine Learning, Neural Networks & Supervised Learning for Beginners
• Python String Comparison Tutorial: Complete Guide
• How to Build a Python Port Scanner From Scratch

Creating the Frontend with HTML and CSS

Setting Up the Frontend Structure

Create a file called index.html in the same directory:

<!DOCTYPE html>
<html lang="en">
<head>
    <meta charset="UTF-8">
    <meta name="viewport" content="width=device-width, initial-scale=1.0">
    <title>AI Content Detector</title>
    <style>
        body {
            font-family: Arial, sans-serif;
            max-width: 600px;
            margin: 50px auto;
            text-align: center;
        }
        textarea {
            width: 100%;
            height: 150px;
        }
        button {
            margin-top: 10px;
            padding: 10px 20px;
            font-size: 16px;
        }
        #result {
            margin-top: 20px;
            font-size: 18px;
            font-weight: bold;
        }
    </style>
</head>
<body>
    <h1>AI Content Detector</h1>
    <p>Enter text to check if it's AI-generated or human-written:</p>
    <textarea id="inputText" placeholder="Type your content here..."></textarea>
    <br>
    <button onclick="checkContent()">Check Content</button>
    <div id="result"></div>

    <script>
        async function checkContent() {
            const text = document.getElementById('inputText').value;
            if (!text) {
                alert("Please enter some text!");
                return;
            }

            const response = await fetch('http://127.0.0.1:5000/predict', {
                method: 'POST',
                headers: { 'Content-Type': 'application/json' },
                body: JSON.stringify({ text })
            });

            const result = await response.json();
            document.getElementById('result').innerText = result.prediction || "Error in prediction!";
        }
    </script>
</body>
</html>

Enhancing the Application with Advanced Features

To make our AI text detector more strong, we can add several advanced features:

• Error Handling: It Ensures that errors are gracefully handled and communicated to the user.
• Loading Indicators: Provide visual feedback to the user while the AI model processes the text.
• Visual Feedback: Use progress bars to show the AI and human confidence levels.

Ensuring Security, Performance, and Scalability

When developing any web application, it’s very important to consider security, performance, and scalability:

• Security: Your API keys are must be stored securely and user input is validated to prevent security vulnerabilities.
• Performance: Optimize your code and infrastructure to handle large volumes of text and concurrent requests.
• Scalability: Plan for scaling your backend to handle increased traffic.

Accessibility and Ethical Considerations

Your application must be accessible to everyone. Especially the users with disabilities. Additionally, consider the ethical implications of AI text detection, including user privacy and data protection.

Conclusion

Creating an AI content detector in Python is both educational and practical. This tutorial covered the basics, but you can take it further by integrating it into a web app or API. If you found this guide helpful, share it with others and explore more resources on our website.

External Resources

OpenAI API Documentation: Link

• Detailed information on using OpenAI models for natural language processing tasks.

Flask Documentation: Link

• Documentation for Flask, a Python web framework covering installation to advanced usage.

HTML MDN Web Docs: Link

• Comprehensive documentation for HTML, including elements, attributes, and best practices.

CSS MDN Web Docs: Link

• Covers styling web pages using CSS, including properties, selectors, and layout techniques.

JavaScript MDN Web Docs: Link

• Detailed explanations and examples of JavaScript features for frontend development.

Spacy Documentation: Link

• Guidance on installation, usage, and advanced features of Spacy, a Python library for natural language processing.

Langdetect Documentation: Link

• Documentation for Langdetect, a Python library for language detection, including installation and usage.

Spellchecker Documentation: Link

• Covers installation, usage, and configuration options for the Spellchecker library in Python.

Web Accessibility MDN Web Docs: Link

• Offers guidance on creating accessible web content for users with disabilities.

Frequently Asked Questions