People are increasingly asking AI, not Google, to help them discover products. A recent shopping report says Americans, this holiday season, will likely turn to large language models this season to find gifts, deals, and sales instead of traditional search.
Retailers could see up to a 520% increase in traffic from chatbots and AI prompts in 2025 compared to 2024, according to the report. For brands, that means figuring out how to show up in AI-generated recommendations, and fast.
This surge in AI-driven traffic is the bet behind The Prompting Company, a YC-backed startup helping products get mentioned in AI apps through GEO (generative engine optimization), a strategy designed for a future where AI agents browse the internet on users’ behalf.
The four-month-old startup, founded by Kevin Chandra, Michelle Marcelline and Albert Punama, has raised $6.5 million in seed funding and already counts Rippling, Rho, Motion, Vapi, Fondo, Kernel, and Traceloop as customers.
“Over the past year, most of the growth on websites has come from AI bots, not people,” co-founder and CEO Chandra told TechCrunch in an interview. “We’re already seeing developers ask AI tools for product recommendations inside their workflows, and we think people, over time, will be less involved in parts of the purchasing funnel.”
As AI becomes the first touchpoint for product discovery and agents eventually transact on a user’s behalf, The Prompting Company believes brands must learn how to market to agents as well as humans.
What that means, according to Chandra, is that brands will need an AI-facing website, a version of their site made for agents without navigation bars, pop-ups, or marketing fluff. “Most businesses still design websites only for humans,” Chandra told TechCrunch. “But the fastest-growing segment of users on the internet today is AI agents and they need a completely different interface.”
Techcrunch event
San Francisco
|
October 27-29, 2025
Here’s how the platform works: First, it identifies and analyzes the questions AI agents are asking by probing models to uncover specific purchase-inte …