The noise around AI search has been at a high pitch for the past two years. Cutting through it to understand what's actually happening is harder than it should be. Here's a grounded read on where things stand.
The numbers that matter
Google's AI Mode has passed one billion monthly users. Queries are reportedly doubling every quarter. Those are meaningful numbers — not proof that traditional search is dying, but proof that AI-mediated search is no longer experimental. It's a mainstream behaviour for a significant share of the population.
Perplexity has established itself as the go-to research interface for a specific but valuable user profile: people doing longer, more complex information gathering. ChatGPT's search mode is used by millions of people who wouldn't describe what they're doing as "AI search" — they're just asking questions. Copilot is embedded across Microsoft's product suite. The surface area of AI retrieval is now very large.
The agentic shift
The more significant signal from I/O 2026 isn't the new input modalities — it's the direction Google is building toward. Persistent agents that monitor topics, retrieve on your behalf, book services, gather information from businesses. The paradigm isn't "better search"; it's "delegated discovery."
If that plays out, the implication for brands isn't just about visibility — it's about agent-readability. Can an agent extract what it needs from your site in order to confidently include you in a recommendation? Pricing, service scope, availability, contact method: structured, accessible, unambiguous.
What's still in flux
Citation behaviour varies across models and changes as they're updated. Best practices from six months ago aren't always best practices now. Measurement is immature — there's no standard attribution model for AI-driven discovery, and most teams are tracking this through manual audits rather than analytics.
The competitive landscape is also unsettled. The models are developing fast. The market shares between ChatGPT, Perplexity, Gemini, and others are still moving. Optimising heavily for one model's specific behaviour is a risky bet.
What's stable enough to act on
Make your content clear and direct. First paragraph answers, structured information, specific claims over vague ones.
Build genuine third-party presence. Editorial coverage, reviews, forum mentions from real users solving real problems.
Measure your AI visibility. Not as a proxy, but directly — which models mention you, for which queries, in what context.
The brands that are doing well at GEO in 2026 aren't running exotic strategies. They're doing the content and PR fundamentals with more deliberate attention to how AI systems read and retrieve them.
The V-Score gives you a direct measure of where you stand across ChatGPT, Claude, Perplexity, and Gemini. Run your free audit and get your baseline.