Vertex AI

Vertex AI was Google Cloud’s unified platform for building, training, deploying and using machine learning models, including Google’s Gemini models and many third-party and open models. In April 2026, at Google Cloud Next, Google launched Gemini Enterprise Agent Platform as “the evolution of Vertex AI”, and said all Vertex AI services would be delivered through the new platform from then on. Many tutorials, including older MLforSEO material, still use the Vertex AI name.

How it works

Vertex AI brought several services together under one Google Cloud project:

• Model Garden and generative AI: access to Gemini and other models through an API or the Vertex AI Studio interface.
• AutoML and custom training: train classification, vision or forecasting models on your own data (this is where Google Cloud AutoML moved).
• Embeddings and vector search: text and multimodal embedding models that place images, text and video in the same space.
• Vertex AI Search: enterprise and retail search, with features such as layout-aware chunking and product search autocomplete.
• Agent tools: Agent Builder and Agent Engine, now renamed within Agent Platform.

How it’s used in SEO and marketing

• Running Gemini on large batches of content for classification, summarisation or extraction with more control than a chat app.
• Generating multimodal embeddings for image search, image SEO audits and product matching.
• Testing how Google’s own models read your content, for example checking how your homepage text is interpreted.
• Site search for ecommerce, including autocomplete trained on your catalogue and shoppers’ searches.
• Training custom models where pre-built APIs do not fit, such as product image detection.

Example

An ecommerce team wants autocomplete for its own site search. Google’s commerce search product (now documented as AI Commerce Search) builds suggestions from the retailer’s product catalogue and the search events it records from its own shoppers. That makes it a site search tool rather than a window into Google Search or Google Shopping queries, so it complements keyword research rather than replacing it.

Limitations and things to watch

• Names and menus have changed; look for Gemini Enterprise Agent Platform in current documentation.
• Usage is billed per call, token or node hour, so set budgets before batch jobs.
• Generative outputs vary between runs. For repeatable entity or sentiment analysis, MLforSEO recommends task-specific APIs.
• Check data governance settings before sending client or personal data.

Learn it on MLforSEO

• Comparing NLP APIs for Entity and Sentiment Analysis (vs Generative AI)
• Image SEO in 2026: The Complete Guide to Image Optimisation, Advanced Tactics and Multimodal Search
• How to use Google Autocomplete API and Places API for Keyword Suggestions with Python
• The Semantic AI-powered SEO Keyword Research course, which covers autocomplete APIs across Google products