Verify Content

Submit content for verification using the POST /v1/verify endpoint.

Verify Content

Submit content for AI-powered verification analysis using Check's multi-paradigm verification system.

POST

/v1/verify

Analyze content and return a verification verdict with confidence score.

Request Body

Parameters

Parameter	Type	Description

Verification Options

Fine-tune verification behavior with the optional options object:

Options Parameters

Parameter	Type	Description

Verification Methods

Configure which methods to use and their relative weights:

Method	FPR	Description	Speed
`formal`	0%	mathjs autoformalization for symbolic verification	Fast
`tool`	~1%	CLoVE contrastive web verification	Slow
`reasoning`	~4-5%	Adaptive self-refinement verification	Medium
`biprm`	~5-7%	Parallel bidirectional step-level scoring	Medium
`entropy`	~8-10%	Dual-path semantic entropy	Medium
`ensemble`	~15-20%	Meta-judge adjudicated consensus	Slow
`semantic`	~20-30%	Boltzmann energy embedding analysis	Fast

FPR = False Positive Rate (lower is better for accuracy)

Methods Configuration

{
  "methods": {
    "reasoning": 1.0,    // Full weight
    "tool": 0.5,         // Half weight
    "ensemble": 0.25     // Quarter weight
  }
}

Weights influence the final confidence score calculation. Higher weights give more importance to that method's result.

Example Request

Response

Success Response (200)

{
  "id": "ver_abc123def456",
  "status": "completed",
  "verdict": "uncertain",
  "confidence": 0.72,
  "decision": "escalate",
  "reasoning": "While hydration is important for health, the specific '8 glasses' recommendation is not universally supported by scientific research. Individual water needs vary based on factors like activity level, climate, and diet.",
  "paradigmResults": [
    {
      "paradigm": "reasoning",
      "verdict": "uncertain",
      "confidence": 0.70,
      "reasoning": "The claim oversimplifies hydration needs. Scientific consensus acknowledges hydration is important but doesn't support a specific '8 glasses' rule.",
      "processingTimeMs": 1200
    },
    {
      "paradigm": "tool",
      "verdict": "false",
      "confidence": 0.85,
      "reasoning": "Multiple authoritative sources (Mayo Clinic, NIH) indicate the '8x8 rule' is a myth. Actual needs vary by individual.",
      "processingTimeMs": 2100,
      "sources": [
        "https://www.mayoclinic.org/...",
        "https://www.nih.gov/..."
      ]
    }
  ],
  "processingTimeMs": 2340,
  "cached": false,
  "metadata": {
    "source": "article-123"
  },
  "createdAt": "2024-01-15T10:30:00Z",
  "completedAt": "2024-01-15T10:30:02Z"
}

Response Fields

Parameter	Type	Description

Paradigm Result Fields

Field	Type	Description
`paradigm`	string	Method name (reasoning, tool, etc.)
`verdict`	string	This method's verdict: "true", "false", "uncertain"
`confidence`	number	This method's confidence score (0.0 to 1.0)
`reasoning`	string	Explanation for this method's verdict
`processingTimeMs`	number	Processing time for this method
`sources`	array	(tool method only) URLs of sources consulted

Decision Mapping

Decision	Confidence	Condition	Recommended Action
`accept`	>= 0.95	High confidence	Content is verified, safe to use
`refine`	0.70 – 0.95	Moderate confidence	Content may need clarification
`escalate`	< 0.70	Low confidence + paradigm disagreement	Flag for human review
`reject`	< 0.70	Low confidence + paradigm agreement	Content is likely false

Error Responses

400 Bad Request - Missing Content

{
  "error": {
    "code": "validation_error",
    "message": "Validation failed",
    "details": [{ "field": "content", "message": "Content is required" }],
    "requestId": "req_abc123"
  }
}

400 Bad Request - No Methods

{
  "error": {
    "code": "validation_error",
    "message": "Validation failed",
    "details": [{ "field": "methods", "message": "At least one verification method is required" }],
    "requestId": "req_abc123"
  }
}

429 Rate Limited

{
  "error": {
    "code": "rate_limited",
    "message": "Rate limit exceeded. Try again in 60 seconds.",
    "retryAfter": 60,
    "requestId": "req_abc123"
  }
}

403 Quota Exceeded

{
  "error": {
    "code": "usage_limit_exceeded",
    "message": "Monthly verification limit reached. Upgrade your plan.",
    "limit": 1000,
    "used": 1000,
    "resetDate": "2024-02-01T00:00:00Z",
    "requestId": "req_abc123"
  }
}

Use Cases

Fast Verification (Chatbots)

For real-time applications where speed matters:

const result = await client.verifyAndWait({
  content: 'Paris is the capital of France.',
  methods: { reasoning: 1.0 }  // Single method, ~500ms
});

High-Accuracy Verification (Important Content)

For content where accuracy is critical:

const result = await client.verifyAndWait({
  content: 'This medication should be taken twice daily.',
  methods: {
    reasoning: 1.0,
    tool: 1.0,      // Web search for verification
    ensemble: 0.5   // Multi-model consensus
  },
  context: { domain: 'medical' }
});

Mathematical Claims

For verifiable mathematical statements:

const result = await client.verifyAndWait({
  content: 'The square root of 144 is 12.',
  methods: { formal: 1.0 }  // 0% FPR for math
});

With Webhook Notification

For async processing without polling:

const response = await client.verify({
  content: 'Your content here',
  methods: { reasoning: 1.0, tool: 0.5 },
  webhookUrl: 'https://your-app.com/webhooks/check'
});

// Response returns immediately with ID
console.log(response.id);  // 'ver_abc123'

// Results delivered via webhook when complete

For real-time applications, consider using webhooks with multiple methods to get accurate results asynchronously without blocking your application.

Polling for Results

If you don't use webhooks, you can poll for results:

// Start verification
const response = await client.verify({
  content: 'Your content',
  methods: { reasoning: 1.0 }
});

// Poll for completion
let result = await client.getVerification(response.id);
while (result.status === 'pending' || result.status === 'processing') {
  await new Promise(r => setTimeout(r, 1000));  // Wait 1 second
  result = await client.getVerification(response.id);
}

console.log(result.verdict);

Or use the SDK's built-in verifyAndWait method which handles polling automatically.

Verify Content

Submit content for verification using the POST /v1/verify endpoint.

Verify Content

Submit content for AI-powered verification analysis using Check's multi-paradigm verification system.

POST

/v1/verify

Analyze content and return a verification verdict with confidence score.

Request Body

Parameters

Parameter	Type	Description

Verification Options

Fine-tune verification behavior with the optional options object:

Options Parameters

Parameter	Type	Description

Verification Methods

Configure which methods to use and their relative weights:

Method	FPR	Description	Speed
`formal`	0%	mathjs autoformalization for symbolic verification	Fast
`tool`	~1%	CLoVE contrastive web verification	Slow
`reasoning`	~4-5%	Adaptive self-refinement verification	Medium
`biprm`	~5-7%	Parallel bidirectional step-level scoring	Medium
`entropy`	~8-10%	Dual-path semantic entropy	Medium
`ensemble`	~15-20%	Meta-judge adjudicated consensus	Slow
`semantic`	~20-30%	Boltzmann energy embedding analysis	Fast

FPR = False Positive Rate (lower is better for accuracy)

Methods Configuration

{
  "methods": {
    "reasoning": 1.0,    // Full weight
    "tool": 0.5,         // Half weight
    "ensemble": 0.25     // Quarter weight
  }
}

Weights influence the final confidence score calculation. Higher weights give more importance to that method's result.

Example Request

Response

Success Response (200)

{
  "id": "ver_abc123def456",
  "status": "completed",
  "verdict": "uncertain",
  "confidence": 0.72,
  "decision": "escalate",
  "reasoning": "While hydration is important for health, the specific '8 glasses' recommendation is not universally supported by scientific research. Individual water needs vary based on factors like activity level, climate, and diet.",
  "paradigmResults": [
    {
      "paradigm": "reasoning",
      "verdict": "uncertain",
      "confidence": 0.70,
      "reasoning": "The claim oversimplifies hydration needs. Scientific consensus acknowledges hydration is important but doesn't support a specific '8 glasses' rule.",
      "processingTimeMs": 1200
    },
    {
      "paradigm": "tool",
      "verdict": "false",
      "confidence": 0.85,
      "reasoning": "Multiple authoritative sources (Mayo Clinic, NIH) indicate the '8x8 rule' is a myth. Actual needs vary by individual.",
      "processingTimeMs": 2100,
      "sources": [
        "https://www.mayoclinic.org/...",
        "https://www.nih.gov/..."
      ]
    }
  ],
  "processingTimeMs": 2340,
  "cached": false,
  "metadata": {
    "source": "article-123"
  },
  "createdAt": "2024-01-15T10:30:00Z",
  "completedAt": "2024-01-15T10:30:02Z"
}

Response Fields

Parameter	Type	Description

Paradigm Result Fields

Field	Type	Description
`paradigm`	string	Method name (reasoning, tool, etc.)
`verdict`	string	This method's verdict: "true", "false", "uncertain"
`confidence`	number	This method's confidence score (0.0 to 1.0)
`reasoning`	string	Explanation for this method's verdict
`processingTimeMs`	number	Processing time for this method
`sources`	array	(tool method only) URLs of sources consulted

Decision Mapping

Decision	Confidence	Condition	Recommended Action
`accept`	>= 0.95	High confidence	Content is verified, safe to use
`refine`	0.70 – 0.95	Moderate confidence	Content may need clarification
`escalate`	< 0.70	Low confidence + paradigm disagreement	Flag for human review
`reject`	< 0.70	Low confidence + paradigm agreement	Content is likely false

Error Responses

400 Bad Request - Missing Content

{
  "error": {
    "code": "validation_error",
    "message": "Validation failed",
    "details": [{ "field": "content", "message": "Content is required" }],
    "requestId": "req_abc123"
  }
}

400 Bad Request - No Methods

{
  "error": {
    "code": "validation_error",
    "message": "Validation failed",
    "details": [{ "field": "methods", "message": "At least one verification method is required" }],
    "requestId": "req_abc123"
  }
}

429 Rate Limited

{
  "error": {
    "code": "rate_limited",
    "message": "Rate limit exceeded. Try again in 60 seconds.",
    "retryAfter": 60,
    "requestId": "req_abc123"
  }
}

403 Quota Exceeded

{
  "error": {
    "code": "usage_limit_exceeded",
    "message": "Monthly verification limit reached. Upgrade your plan.",
    "limit": 1000,
    "used": 1000,
    "resetDate": "2024-02-01T00:00:00Z",
    "requestId": "req_abc123"
  }
}

Use Cases

Fast Verification (Chatbots)

For real-time applications where speed matters:

const result = await client.verifyAndWait({
  content: 'Paris is the capital of France.',
  methods: { reasoning: 1.0 }  // Single method, ~500ms
});

High-Accuracy Verification (Important Content)

For content where accuracy is critical:

const result = await client.verifyAndWait({
  content: 'This medication should be taken twice daily.',
  methods: {
    reasoning: 1.0,
    tool: 1.0,      // Web search for verification
    ensemble: 0.5   // Multi-model consensus
  },
  context: { domain: 'medical' }
});

Mathematical Claims

For verifiable mathematical statements:

const result = await client.verifyAndWait({
  content: 'The square root of 144 is 12.',
  methods: { formal: 1.0 }  // 0% FPR for math
});

With Webhook Notification

For async processing without polling:

const response = await client.verify({
  content: 'Your content here',
  methods: { reasoning: 1.0, tool: 0.5 },
  webhookUrl: 'https://your-app.com/webhooks/check'
});

// Response returns immediately with ID
console.log(response.id);  // 'ver_abc123'

// Results delivered via webhook when complete

For real-time applications, consider using webhooks with multiple methods to get accurate results asynchronously without blocking your application.

Polling for Results

If you don't use webhooks, you can poll for results:

// Start verification
const response = await client.verify({
  content: 'Your content',
  methods: { reasoning: 1.0 }
});

// Poll for completion
let result = await client.getVerification(response.id);
while (result.status === 'pending' || result.status === 'processing') {
  await new Promise(r => setTimeout(r, 1000));  // Wait 1 second
  result = await client.getVerification(response.id);
}

console.log(result.verdict);

Or use the SDK's built-in verifyAndWait method which handles polling automatically.