AI Content Detector

AI Content Detector analyzes text to determine whether it was written by a human or generated by AI, using statistical analysis without relying on any external AI APIs. Fast, private, and accurate -- your text never leaves the processing pipeline.

What does AI Content Detector do?

This actor takes a piece of text and runs it through multiple statistical analyses to calculate the probability that it was AI-generated. It measures perplexity, burstiness, vocabulary diversity, sentence uniformity, repetition patterns, transition word density, punctuation regularity, and known AI phrase frequency. The result is a clear verdict -- likely_ai, likely_human, or uncertain -- along with a confidence score and a full breakdown of every metric.

Three modes are available. Detect analyzes a single text. Bulk Detect processes multiple texts at once. Stats returns basic text statistics (word count, sentence count, paragraph count, etc.) without running the AI detection.

The detection runs entirely on statistical methods. There are no calls to OpenAI, GPT, or any other external service. This means your text stays private and processing is fast.

What data do you get?

• ai_probability -- overall probability the text is AI-generated (0.0 to 1.0)
• verdict -- likely_ai, likely_human, or uncertain
• confidence -- how confident the model is in its verdict
• text_length -- character count of the input
• perplexity_score -- language predictability (lower = more AI-like)
• burstiness_score -- sentence length variation (uniform = AI-like)
• vocabulary_diversity -- type-token ratio
• repetition_score -- repeated n-gram frequency
• transition_word_density -- transition words per total words
• ai_phrase_count -- number of common AI-generated phrases detected
• sentence_uniformity -- how uniform sentence lengths are
• avg_sentence_length -- average words per sentence
• std_sentence_length -- standard deviation of sentence lengths
• unique_word_ratio -- unique words divided by total words
• punctuation_regularity -- regularity of punctuation patterns

Stats mode returns: word_count, sentence_count, paragraph_count, avg_word_length, char_count.

Who is this for?

• Content teams and editors -- verify that submitted articles are human-written before publishing
• Academic institutions -- screen student submissions for AI-generated content
• SEO professionals -- check whether competitor content or outsourced articles are AI-generated
• Publishers and media companies -- maintain editorial integrity by flagging AI content
• Developers -- integrate AI detection into your own content management tools

How to use it

1. Open the actor in Apify Console and choose a mode (detect, bulk_detect, or stats).
2. Paste your text into the text field, or provide an array of texts for bulk mode.
3. Click "Start" to run the analysis.
4. Review the results in the Dataset tab -- you'll get a verdict, probability score, and detailed breakdown.
5. Export results or pull them via the Apify API for automated workflows.

Input parameters

Parameter	Type	Default	Description
mode	string	detect	Operation mode: detect, bulk_detect, or stats
text	string	--	Text to analyze (for detect and stats modes)
texts	array	--	Array of texts to analyze in bulk (for bulk_detect mode)
language	string	en	Language of the input text (currently English only)

Sample output

{
  "text_length": 1847,
  "ai_probability": 0.82,
  "verdict": "likely_ai",
  "confidence": 0.76,
  "details": {
    "perplexity_score": 0.71,
    "burstiness_score": 0.85,
    "vocabulary_diversity": 0.43,
    "repetition_score": 0.38,
    "transition_word_density": 0.062,
    "ai_phrase_count": 7,
    "sentence_uniformity": 0.79,
    "avg_sentence_length": 18.4,
    "std_sentence_length": 3.2,
    "unique_word_ratio": 0.51,
    "punctuation_regularity": 0.67
  }
}

How much does it cost?

Each result costs $0.003. Analyzing 1,000 texts costs $3, and 10,000 texts costs $30.

Apify gives every new user $5 in free monthly credits, so you can analyze roughly 1,600 texts for free.

Common questions

How accurate is the detection? The detector uses multiple statistical signals combined into a weighted score. It works well for standard-length content (300+ words). Very short texts or heavily edited AI content may produce uncertain results, which is why the actor returns a confidence score alongside the verdict.

Does it send my text to any external AI service? No. All analysis is done locally using statistical methods. Your text never leaves the actor's runtime environment.

Does it support languages other than English? Currently, the analysis is optimized for English text. The statistical models (AI phrase detection, transition word lists, etc.) are English-specific. Support for additional languages may be added in the future.

Contact & Custom Solutions

Need a custom scraper, higher volume, or a specific integration? We're here to help.

If anything isn't working right or you need support, don't hesitate to reach out.

• Telegram: t.me/novashield_dev
• Email: novashield.dev@gmail.com