Multiple Image AI Analysis

Multiple Image AI Analysis and Regression Testing

The Multiple Image AI Analysis feature can integrate AI-based image analysis directly into your workflow. You can submit up to 5 image URLs and 5 AI Prompts.

Prompts can analyze any image in the array or compare multiple images (e.g., “Compare the second and third images"). Prompts can also extract specific details like product prices, or simply describe the content of an image.

Each image and each prompt sent for analysis counts as an individual "AI Image Analysis Credit" for your PagePixels account.

AI Analysis Request Body Schema

AI Image Analysis Endpoint

Use the post ai/analyze endpoint to submit an array of images and an array of prompts to go along with those images for an AI to analyze.

Example

curl -X POST "https://api.pagepixels.com/ai/analyze?access_token=your-api-key" \
  -H "Content-Type: application/json" \
  -d '{
    "urls": [
      "https://cdn.pagepixels.com/4cc8ebae-ea0d-41e3-b766-11d2f805c235/a77787747d21ba-1789523042538.jpg",
      "https://cdn.pagepixels.com/2c9f6245-0fcd-42d7-bf06-6c551148d3f2/d1f9855cab709c-1789523093784.jpg"
    ],
    "prompts": [
      "What is the second image of?",
      "What is the first image of?",
      "Compare and contrast the two images."
    ]
  }'

{
  "success": true,
  "result": {
    "urls": [
      "https://cdn.pagepixels.com/4cc8ebae-ea0d-41e3-b766-11d2f805c235/a77787747d21ba-1789523042538.jpg",
      "https://cdn.pagepixels.com/2c9f6245-0fcd-42d7-bf06-6c551148d3f2/d1f9855cab709c-1789523093784.jpg"
    ],
    "prompts": [
      "What is the second image of?",
      "What is the first image of?",
      "Compare and contrast the two images."
    ],
    "responses": [
      "The second image promotes the features of...",
      "The first image is about the pricing plans...",
      "Comparison:\n - Both images focus on..."
    ]
  }
}

Example: using your own AI provider

With the Bring Your Own AI addon and a key stored for the provider, add ai_provider and ai_model and the same request runs on your account instead. Nothing else changes. The example below sends the same two images and three prompts to OpenRouter, so it uses no AI Image Analysis credits. Your key is never sent on this request; it is looked up from the one you stored with the store a key endpoint.

curl -X POST "https://api.pagepixels.com/ai/analyze?access_token=your-api-key" \
  -H "Content-Type: application/json" \
  -d '{
    "urls": [
      "https://cdn.pagepixels.com/4cc8ebae-ea0d-41e3-b766-11d2f805c235/a77787747d21ba-1789523042538.jpg",
      "https://cdn.pagepixels.com/2c9f6245-0fcd-42d7-bf06-6c551148d3f2/d1f9855cab709c-1789523093784.jpg"
    ],
    "prompts": [
      "What is the second image of?",
      "What is the first image of?",
      "Compare and contrast the two images."
    ],
    "ai_provider": "openrouter",
    "ai_model": "google/gemini-3.8-flash"
  }'

{
  "success": true,
  "result": {
    "urls": [
      "https://cdn.pagepixels.com/4cc8ebae-ea0d-41e3-b766-11d2f805c235/a77787747d21ba-1789523042538.jpg",
      "https://cdn.pagepixels.com/2c9f6245-0fcd-42d7-bf06-6c551148d3f2/d1f9855cab709c-1789523093784.jpg"
    ],
    "prompts": [
      "What is the second image of?",
      "What is the first image of?",
      "Compare and contrast the two images."
    ],
    "responses": [
      "The second image promotes the features of...",
      "The first image is about the pricing plans...",
      "Comparison:\n - Both images focus on..."
    ],
    "ai_provider": "openrouter",
    "ai_model": "google/gemini-3.8-flash"
  }
}

Leave out ai_model to use the default model stored with your key. If no key is stored for the provider the request fails with ai_provider_not_configured (412), and a model without image input fails with ai_model_not_supported (400). Claude works the same way with "ai_provider": "anthropic", and OpenAI with "ai_provider": "openai".

Response Breakdown

Structured Data for Multiple Images

The same Structured Data you can use on a screenshot works here too. Define the fields you want and the analysis returns JSON that matches them, so a regression check or a comparison across several images drops straight into your code. Set ai_prompt_mode=structured and pass ai_structured_fields. prompts becomes optional; anything you do send is added to the instructions.

Pick a scope with ai_structured_scope: request (the default) answers once for the whole set of images, which suits comparisons. image answers once per image, in the order you sent them, which suits checking many pages against the same checklist.

Credits work as before: one per image and one per prompt, with the schema counting as one prompt when you send none. With your own provider, no credits are used: add ai_provider and ai_model to either example below exactly as in the own-provider example above, and the response gains the same two keys.

Example: one result per image

curl -X POST "https://api.pagepixels.com/ai/analyze?access_token=your-api-key" \
  -H "Content-Type: application/json" \
  -d '{
    "urls": [
      "https://cdn.pagepixels.com/4cc8ebae-ea0d-41e3-b766-11d2f805c235/a77787747d21ba-1789523042538.jpg",
      "https://cdn.pagepixels.com/2c9f6245-0fcd-42d7-bf06-6c551148d3f2/d1f9855cab709c-1789523093784.jpg"
    ],
    "ai_prompt_mode": "structured",
    "ai_structured_scope": "image",
    "ai_structured_instructions": "These are pages from our marketing site.",
    "ai_structured_fields": [
      {"name": "page_loaded", "type": "boolean", "description": "Did the page render without errors?"},
      {"name": "status", "type": "enum", "options": ["ok", "warning", "broken"]},
      {"name": "headline", "type": "string", "description": "the main headline"}
    ]
  }'

{
  "success": true,
  "result": {
    "urls": [
      "https://cdn.pagepixels.com/4cc8ebae-ea0d-41e3-b766-11d2f805c235/a77787747d21ba-1789523042538.jpg",
      "https://cdn.pagepixels.com/2c9f6245-0fcd-42d7-bf06-6c551148d3f2/d1f9855cab709c-1789523093784.jpg"
    ],
    "prompts": [],
    "responses": [],
    "ai_prompt_mode": "structured",
    "ai_structured_scope": "image",
    "ai_structured_fields": [
      {"name": "page_loaded", "type": "boolean", "description": "Did the page render without errors?"},
      {"name": "status", "type": "enum", "description": "", "options": ["ok", "warning", "broken"]},
      {"name": "headline", "type": "string", "description": "the main headline"}
    ],
    "ai_structured_response": [
      {"page_loaded": true, "status": "ok", "headline": "Scheduled Screenshots"},
      {"page_loaded": true, "status": "warning", "headline": "Pricing"}
    ]
  }
}

Example: one result for the whole set

Leave out ai_structured_scope (or set it to request) and describe the set as a whole. This example compares two captures of the same page:

curl -X POST "https://api.pagepixels.com/ai/analyze?access_token=your-api-key" \
  -H "Content-Type: application/json" \
  -d '{
    "urls": ["https://cdn.pagepixels.com/.../before.jpg", "https://cdn.pagepixels.com/.../after.jpg"],
    "ai_prompt_mode": "structured",
    "ai_structured_instructions": "The first image is yesterday, the second is today.",
    "ai_structured_fields": [
      {"name": "changed", "type": "boolean", "description": "Is there any visible difference between the two?"},
      {"name": "changes", "type": "multi", "options": ["price", "headline", "layout", "images", "navigation"]},
      {"name": "summary", "type": "string", "description": "one sentence describing the difference"}
    ]
  }'

{
  "success": true,
  "result": {
    "urls": ["https://cdn.pagepixels.com/.../before.jpg", "https://cdn.pagepixels.com/.../after.jpg"],
    "prompts": [],
    "responses": [],
    "ai_prompt_mode": "structured",
    "ai_structured_scope": "request",
    "ai_structured_fields": [ ... ],
    "ai_structured_response": {"changed": true, "changes": ["price"], "summary": "The monthly price changed from $8.99 to $9.99."}
  }
}

A simple request (prompts only) returns exactly what it always has. The structured keys appear only when you ask for them.