Skip to content
TheChatArchiveThe Chat Archive
Geminigemini-2.5-pro-exp-03-25-grounding· GoogleHugging Face LMArena Search Arena 24K

Do you have OCR capabilities?

From a Chatbot Arena battleThe same person sent this prompt to two anonymous models at once. Compare it with Perplexitysonar-reasoning-pro-high’s answer.
  1. User

    Do you have OCR capabilities?

  2. Gemini Assistant

    As a large language model, I don't have built-in Optical Character Recognition (OCR) capabilities myself in the way that a dedicated OCR software application does. I cannot directly "see" an image file or scan a document and extract the text from it.

    However, I can process and understand text that has already been extracted using an OCR tool. If you use an OCR tool to convert the text within an image or document into digital text format, you can then provide that text to me, and I can:

    • Analyze it
    • Summarize it
    • Translate it
    • Answer questions about it
    • Rewrite it, etc.

    So, while I don't perform the OCR action itself, I can certainly work with the results of OCR.

  3. User

    Can you use ocr to interpret image data from a website?

  4. Gemini Assistant

    Okay, let's clarify this based on my capabilities:

    1. Direct OCR on Website Images: I, as a language model, cannot directly browse a website, locate an image within it, and then perform OCR on that specific image myself. I don't have "eyes" to see the visual layout of a webpage or the capability to run OCR software on image files directly from a URL.
    2. Processing Text Associated with Images: When I access web information (usually via search tools), I primarily get the text content of the page. This might include:
      • The main text of the article or page.
      • Image captions.
      • ALT text (alternative text) associated with images, which is supposed to describe the image content for accessibility and search engines.
      • Sometimes, search engines themselves might have already performed OCR on images they've indexed, and some of that text might be available in the search results I receive.
    3. Working with Text After OCR: If you or another tool performs OCR on an image from a website and provide me with the extracted text, then I can absolutely interpret, analyze, summarize, translate, or answer questions about that text.

    In summary: I cannot perform the OCR action on an image hosted on a website myself. That step needs to happen externally first. However, I can work with the text once it has been extracted by an OCR tool, or sometimes interpret text related to the image (like captions or alt text) if it's present on the webpage's text content that I can access.

Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.