Vertex AI Search: AI search for your own website explained

Vertex AI Search is Google Cloud's search product for your own website and app. Google is currently renaming it Agent Search.

In Germany, most people still search for the old name: around 170 monthly searches for “vertex ai search” versus around 20 for “agent search.” That's according to our own analysis of Google Ads data via DataForSEO (average from September 2025 to August 2026, retrieved October 2026).

This article is an overview. It covers what the product can do according to Google, what an online store needs to use it, and which parts of the work we handle.

Agent Search answers questions on your website with AI

According to Google, Agent Search lets you build AI-powered search for public websites and apps. It runs on your own content: pages, documents and structured data.

Features Google highlights include:

• AI answers: Search summarizes content and answers questions conversationally, including follow-ups.
• Semantic search: Search understands natural-language queries, not just individual keywords.
• Result controls: filters, autocomplete suggestions and targeted boosting of specific results.
• Personalization and recommendations: personalized results while browsing, plus custom recommendations (still in preview).
• Integration: a ready-made search widget for your own website.

Online stores get a separate version

For retail, Google offers a dedicated version, formerly Vertex AI Search for commerce and now AI Commerce Search. It searches your product catalog, which you import specifically for this purpose.

Three points from Google matter for store owners:

1. Update daily: Google recommends re-importing the catalog every day if possible. Without regular imports, model quality degrades over time.
2. Hourly for large changes: If a lot changes, Google recommends incremental imports every one to two hours.
3. No new Merchant Center connections: Google lists the Merchant Center connection as deprecated and available to existing users only. New projects load the catalog via the Catalog API or BigQuery.

If you also run Google Shopping, your product data goes to Google twice: once to Merchant Center and once to your store search. Price, stock and attributes need to be in sync across both.

Google survey: 77% avoid websites after a poor search experience

In late 2022, The Harris Poll surveyed almost 13,500 adults in 14 countries on behalf of Google Cloud. 77% of respondents worldwide said they avoid websites where they've had search problems before; in the US, the figure was 82%.

The study estimates that retailers worldwide lose more than US$2 trillion in revenue each year to abandoned searches. It does not give a separate figure for Germany.

What we do and what we don't

We don't set up Agent Search or AI Commerce Search. That's a project for your IT team or a Google Cloud partner.

We work on the product data, landing pages and content that bring customers to you from ads and AI answers:

• Product data: We write product titles that match real search queries and add material, size and intended use. At Jomos Schuhe, sales rose 128% after we revised titles, rebuilt tracking and adjusted campaigns (Jomos Schuhe case study, product feed management).
• We map search queries to the landing pages that best explain each offer. Anyone searching for an offer lands on exactly that offer. We implemented this for fensterkonzepte.de for windows, doors and sun protection (fensterkonzepte.de case study).
• AI answers beyond your website: Agent Search only works on your own site. To get mentioned in ChatGPT, Perplexity and Google, we structure your pages so these systems can pull answers directly from them. That brought ChargeOne 202 visits from AI assistants (ChargeOne case study, GEO).

Our article on Generative Engine Optimization explains in detail how to become visible in AI answers.