Computer Vision (Image Recognition)

Computer vision is the branch of artificial intelligence that enables machines to understand images and video. Its tasks include image classification (what is in this picture?), object detection (where is each object?), optical character recognition (what does the text say?), image captioning and finding visually similar images. Image recognition is the everyday name for the most common of these tasks.

How it works

Most modern computer vision uses deep learning. Convolutional neural networks, and more recently vision transformers, learn to detect edges, textures, shapes and eventually whole objects from millions of labelled images. Pre-trained models can be used as they are, through APIs such as Google Cloud Vision, or as feature extractors: you pass an image through a model such as ResNet50 and get a numeric vector you can cluster or compare. Multimodal models such as Gemini or GPT-4o handle images and text together.

How it’s used in SEO and marketing

• Image SEO: Google says it uses alt text “along with computer vision algorithms and the contents of the page” to understand images, so the picture itself is part of what is evaluated.
• Visual search: Google Lens and multimodal search match photos to products and pages.
• Alt text and captions at scale: generating drafts with image captioning models, then reviewing them.
• Organising image libraries: clustering product or blog images by colour or visual features.
• Moderation and quality control: flagging unsafe, low-quality or off-brand images.

Example

In MLforSEO’s image clustering practical, product images from a fashion data set are grouped by colour, first with simple colour features and then with a pre-trained CNN to extract richer features, reduced with PCA and clustered with k-means. The clusters help build colour-based navigation and spot mislabelled products, and their quality is checked with the silhouette score and by looking at samples.

Limitations and things to watch

• Models trained on general data can misread niche products, logos or cultural context.
• Deep models need significant compute and are hard to explain, which matters for critical decisions.
• Generated alt text needs human review for accuracy and accessibility.
• For many sites, structured product data already exists in a database; you do not always need computer vision to get it.

Learn it on MLforSEO

• Image SEO in 2026: The Complete Guide to Image Optimisation, Advanced Tactics and Multimodal Search
• Image Clustering by Color with Three Machine Learning Approaches for Visual Content Organization
• The Characteristics of a Machine Learning Problem: A Task, Data and Solution Framework for SEOs
• How to Find Machine-Learning-Enabled Automation in Your SEO Workflow
• Introduction to Machine Learning for SEO (course)