API for Chatbots

Your Chatbot Uses Claude API. Claude Adds Watermarks. We Strip Them.

Real-time watermark removal for chatbot responses. Add one middleware layer → protect all users. Streaming support included.

const response = await fetch(
  "https://api.gptwatermarkremover.com/v1/clean",
  {
    method: "POST",
    headers: {
      "X-API-Key": "YOUR_API_KEY",
      "Content-Type": "application/json"
    },
    body: JSON.stringify({ text: "Your AI-generated text here" })
  }
);

const { cleaned_text } = await response.json();

The Chatbot Watermark Problem

You use AI APIs for chatbot responses. AI providers embed watermarks. User's company runs AI check. Busted.

1. User Asks Question

Your chatbot receives query from sales rep, support agent, or employee.

2. Claude API Responds with Watermarks

You call claude.messages.create(). Response contains invisible tracking characters.

3. User Copies → Gets Detected

User pastes response into email/Slack. Company AI detector flags it. Trust lost.

The Fix: Middleware Watermark Removal

Add watermark removal between Claude API response and user display. User sees clean response in <2.1 seconds total (2 sec AI + 0.1 sec cleaning).

// Your chatbot streams response from Claude
const stream = await anthropic.messages.stream({
  model: "claude-3-5-sonnet-20241022",
  messages: [{ role: "user", content: "Help me..." }],
  max_tokens: 1024
});

let fullResponse = "";
for await (const chunk of stream) {
  if (chunk.type === 'content_block_delta') {
    fullResponse += chunk.delta.text;
  }
}

// Strip watermarks from complete response
const clean = await fetch(
  'https://api.gpt-watermark-remover.com/v1/remove',
  {
    method: 'POST',
    headers: { 'X-API-Key': 'your_api_key' },
    body: JSON.stringify({ text: fullResponse })
  }
);

const { clean_text } = await clean.json();
return clean_text;  // ✅ No watermarks

Middleware Pattern

Invisible layer between AI API and your users

User Query

Claude API

Has watermarks

Watermark Removal

<100ms

Clean Response to User

No watermarks, no detection risk

Streaming Response Support

For real-time chatbots using Claude/ChatGPT streaming APIs:

  • Collect stream chunks into full response
  • Strip watermarks after stream completes
  • <100ms added latency (imperceptible)
  • Users start seeing text immediately

Batch Processing

For chatbots with multiple response variations or async workflows:

  • Clean up to 50 responses in one API call
  • Webhook support for async processing
  • Perfect for A/B testing or multi-variant responses
  • Same <100ms latency per response
<100ms
Latency Added
3 req/s
Default Rate Limit
99.9%
Uptime SLA
50/batch
Responses/Call
99.9%
Uptime
<100ms
Response Time
All Known
Watermark Types
100K
Max Characters
HTTPS encrypted
No text stored
Edge-deployed globally

Simple Pricing

API access included with every premium subscription

Available Now

Premium Subscription

Everything you need to get started

  • 1,000 API calls per month
  • 100,000 characters per request
  • All web features included
  • Instant API key generation
View Documentation
Coming Soon

Developer Tiers

Higher volume plans for production apps

  • Higher volume tiers
  • Team and organization plans
  • Custom enterprise solutions

Watermark Types We Detect

Our API detects and removes all known invisible Unicode characters used by AI systems like ChatGPT, Claude, Gemini, and others to watermark generated text.

Zero-Width Characters

Invisible characters with no visual width, commonly inserted by AI systems to mark generated text.

Zero-Width Space
Most common AI watermark. Invisible space character inserted between words or characters.
Zero-Width Non-Joiner
Prevents character joining in scripts like Arabic. Used as hidden marker in AI text.
Zero-Width Joiner
Forces character joining. Commonly used in emoji sequences, exploited for watermarking.
Byte Order Mark
Originally for encoding detection. When embedded mid-text, serves as invisible watermark.
Word Joiner
Prevents line breaks. Invisible character used to mark AI-generated content.

Formatting Characters

Characters that affect text rendering or spacing without being visible.

Soft Hyphen
Marks where a word can break. Invisible unless line breaks at that point.
Combining Grapheme Joiner
Affects how characters combine. Used to invisibly mark text.
Function Application
Mathematical invisible operator. Repurposed as text watermark.
Invisible Times
Mathematical multiplication marker. Zero-width, used in watermarking.
Invisible Separator
Mathematical separator with no width. Hidden text marker.

Bidirectional Control

Characters that control text direction, often invisible and exploitable for watermarking.

Left-to-Right Mark
Forces left-to-right text direction. Invisible directional override.
Right-to-Left Mark
Forces right-to-left text direction. Hidden directional marker.
Directional Embeddings
Multiple characters that embed text direction. Invisible formatting controls.
Directional Overrides
Override text direction rendering. Hidden watermark characters.

Variation Selectors

Characters that modify how preceding characters render, often invisibly.

Standard Variation Selectors
Selectors that change character appearance. Often invisible modifications.
Extended Variation Selectors
Extended selectors for specialized use. Rarely used, suspicious when found in regular text.

Annotation Characters

Characters for interlinear annotations, invisible in most contexts.

Annotation Anchor
Marks annotation start. Invisible formatting character.
Annotation Separator
Separates annotation elements. Zero-width marker.
Annotation Terminator
Ends annotation. Hidden text marker.

Why Do AI Systems Add Watermarks?

Large language models like ChatGPT, Claude, and Gemini embed invisible watermarks to enable detection of AI-generated content. These watermarks use Unicode characters that have zero visual width but can be detected programmatically. Our API removes all known watermark patterns while preserving your text's meaning and formatting.

Chatbot Integration FAQ

Technical questions about real-time watermark removal

Production-ready chatbot protection in minutes

Every Chatbot Response, Watermark-Free

Add one middleware layer. Protect millions of conversations. Real-time watermark removal for Claude, ChatGPT, and Gemini chatbots.

<100ms
Added Latency
Streaming
Fully Supported
Real-time
Processing

1,000 API calls included • Claude, ChatGPT & Gemini supported