Today, we're launching Agent Analytics, a new category of measurement that shows brands how they're being discovered.
The Blind Spot
ChatGPT now has more than 800 million weekly active users, the fastest-adopted technology in history[1]. Gemini, Claude, and Grok have each crossed 100 million. Last holiday season, agents influenced 20% of all global online orders[2].
The path to purchase has changed. A customer asks Claude which credit card to get for travel rewards. The agent analyzes her spending, scans welcome offers, and narrows the field to three. She picks one. She might show up on a brand's site a week later, or at a branch, or never at all. Most attribution stacks were built for the search-and-social era. They cannot see any of it.
Why We Built Agent Analytics
Agents are a new customer acquisition channel sitting on top of existing infrastructure, not a replacement for it. Their answers are grounded in web search, social results, and product feeds, so brands that have already won search and social have a head start, but no way to see it.
We built an engine that sits across the major agent platforms and surfaces what's being said in those conversations. Every company will need visibility into these conversations as a leading indicator of customer growth.
Three Metrics That Matter
Agent Analytics measures brand presence in agent conversations, before the customer ever visits your site. Three metrics every brand should track:
- Brand Share. How often your brand appears when customers ask agents about products and services in your category.
- Brand Win Rate. When your brand appears, how often it is the top recommendation.
- Top Influencing Content. Which sources are shaping the conversations.
For search advertisers, these will look familiar: the agent equivalents of impression share and click share. Early evidence suggests brand share is a leading indicator of revenue, and it should become a board-level metric for years to come.
What Drives the Number
Once a brand can see how it ranks, the next step is to grow that share. Three things drive agent recommendations.
- Training data. Agents ingest and compress huge swaths of the internet into an understanding of brands, products, and categories. The crawlers doing this work (OpenAI's GPTBot, Anthropic's ClaudeBot, Google-Extended for Gemini) are separate from traditional search. If your content isn't accessible to them, it isn't in the model.
- Live retrieval. After training, agents ground their real-time answers in live web search. ChatGPT uses Bing, Gemini uses Google, Claude uses Brave. Watch closely the next time you use ChatGPT and you'll see it search the web before answering. If your brand doesn't rank well in those underlying engines, you're not in the consideration set.
- Direct integrations. Where they exist, brands can plug into agents directly through structured feeds. Retail is the leading edge: agents are now ingesting product feeds the way Google has for years with Product Listing Ads, except the cards now appear in agent conversations rather than search results. Where it's available, direct integration is a great strategy.
The Next Decade
The closest parallel is the arrival of the iPhone in 2007. Before then, the internet was desktop-driven. Mobile created an entirely new surface for brands. New opportunities, new risks. Industry leaders are already a year or more into building for agents. The brands that will win the next decade are setting up now.
The agent economy has arrived. Time to measure it.
