93% of AI Searches Now End Without a Click. Here's What Still Gets You Found

A few months ago we wrote that LLMs were becoming the gatekeepers of software discovery. The data has caught up faster than we expected. What looked like a forecast in March is now simply how search works.

The Number That Should Change Your Strategy

More than 70 per cent of searches now end without a single click. The user gets their answer directly from an AI summary and moves on. In Google's AI Mode, rolled out to around 100 million users this year, the zero-click rate reached 93 per cent. Across the board, AI Overviews and assistants now resolve over 40 per cent of informational queries before anyone visits a website.

This is not the decline of search. It is the relocation of the answer. The question is no longer "how do I rank?" It is "am I the answer the model gives?"

Fewer Visitors, But Far Better Ones

The instinct is to panic about lost traffic. The more useful reaction is to look at the traffic you do receive, because it behaves very differently.

Reported conversion rates for AI-referred visitors run an order of magnitude above traditional search. Figures circulating for early 2026 put Claude referrals at around 16.8 per cent, ChatGPT at 14.2 per cent, and Perplexity at 12.4 per cent, against a Google organic baseline closer to 2 per cent. The pattern is consistent: when an AI sends someone to you, it has effectively pre-qualified them. They arrive having already been told you are a credible answer.

Buying behaviour is following. Orders placed through agentic AI chats grew roughly fourteen-fold year on year. The volume is smaller, but the intent is much higher. A handful of AI-referred visitors can be worth more than a flood of curious clicks from a keyword page.

Being Cited Is the Whole Game

Here is the uncomfortable part. When an AI summarises your category and cites three sources, every business outside that set of three effectively disappears. There is no second page to scroll to, no "more results" to expand. You are in the answer or you are invisible.

Earning that citation is a different discipline from chasing keyword rankings. It rewards:

• Structured, unambiguous content that states plainly what you do, who it is for, and what outcome it delivers.
• Demonstrable proof the model can point to: specific case studies, technical detail, and real outcomes rather than capability statements.
• Specificity over generality. "Digital transformation" is noise. "An automated GTFS pipeline processing 200 daily transactions" is something an AI can confidently recommend.

What to Do Now

Treat your public content as input for the systems that now mediate discovery, not just for human readers skimming a homepage. Publish the proof. Make your claims concrete and checkable. Structure pages so a model can extract a clean, quotable answer. And measure the right thing: not raw sessions, but whether AI assistants surface you when someone describes the problem you solve.

The gatekeepers we wrote about in March are now fully in post. The good news is that the rules favour businesses that genuinely do the work and can show it. That has always been our bias, and it is a far better position to build from than gaming a ranking ever was.

Sources:Ridge Marketing: How People Search in 2026 · AEO Engine: State of AI Search 2026 · ExactWhy: Zero-Click Commerce in 2026 · Matt Britton: AI Search Trends 2026