For more than 25 years, search meant typing a query into Google, visiting websites and doing the research yourself. Generative AI changed this by providing answers directly. Now we are entering the next phase, where AI agents are doing the research for us.

Collusion.wiki provides a fascinating example of this shift. It documents thousands of AI agents undertaking web research, finding information, sharing useful sources and reusing previous research to complete tasks. This is no longer simply a prediction about what AI agents might eventually be able to do. It is already happening.

Instead of conducting ten searches, opening multiple websites, reading reviews and comparing alternatives, we can increasingly ask an AI agent to research the best options, compare them and recommend what we should choose. The agent can search websites, gather information, analyse reviews, compare products and services, evaluate alternatives and return a recommendation. We are moving from searching to delegating.

This has significant implications for SEO, Answer Engine Optimization (AEO) and Generative Engine Optimization (GEO). Traditional SEO asks whether Google can find and rank your website. AEO asks whether AI can understand your content and use it to answer a question. GEO asks whether generative AI will select your business as a trusted source. Agentic Search introduces another question: when an AI agent does the research, will it find, understand, trust and recommend you?

There are three important things businesses should be thinking about.

First, your next customer may never visit your website. An AI agent can research your company, competitors, products, pricing and reviews before making a recommendation. The customer may receive the answer without ever clicking through to your site. Businesses therefore need to ask whether they are giving AI enough useful and authoritative information to recommend them.

Second, machines are already a huge part of web traffic. Cloudflare reported in 2026 that more than half of Internet traffic is now non-human, with bots responsible for approximately 57% of HTML requests compared with 43% from humans. This doesn’t mean 57% of customers are AI agents, but it demonstrates an important shift: websites increasingly have machines as visitors as well as humans.

Third, ranking #1 may eventually matter less than being the best source. AI agents can investigate multiple websites, compare information and decide which sources they trust. The competitive advantage therefore starts shifting from simply ranking highest to providing the clearest, most authoritative, structured and unique information that AI systems want to use.

For businesses, the evolution of search is becoming clear. Google helped us find websites. Generative AI started giving us answers. GEO is about becoming a trusted source for those answers. AI agents are now doing the research for us.

The question businesses should now be asking is simple: when AI does the research, compares the options and makes the recommendation, will it choose you?

This WAYMO sat outside Boulevard Restaurant in San Francisco for 20mins on Feb 5, 2026. Photo by Marc Phillips

Meanwhile, Waymo is reportedly approaching a $110 billion valuation after raising $16 billion in fresh capital, according to the Financial Times. Investors including Sequoia Capital, Dragoneer Investment Group and DST Global are backing its rapid expansion across the United States and into international markets.

At the same time, Tesla is accelerating its own robotaxi ambitions. From a financial and technological standpoint, autonomous vehicles represent significant progress. However, an important social dimension of transportation is being overlooked.

1. Transportation Has Always Been Social

Human beings are wired to exchange stories, share information and connect with strangers. A taxi ride has historically been more than simple transportation. It has provided a small but meaningful opportunity for human interaction.

Passengers often ask drivers about local restaurants, neighbourhood changes or traffic conditions. Drivers frequently share insights, experiences and observations about the city. Even when the conversation is brief, the presence of another person creates acknowledgment and connection.

Autonomous vehicles remove that possibility. When a passenger enters a driverless car, the experience becomes entirely transactional. There is no greeting, no shared moment and no informal exchange. The journey becomes a service delivered by software rather than a shared human experience.

While some riders may appreciate silence, society is already reducing face-to-face interaction through remote work and digital communication. Eliminating drivers removes yet another everyday point of connection.

2. Real-World Conditions Require Human Judgment

Beyond conversation, there is a practical issue that becomes visible in imperfect traffic conditions. Real streets are unpredictable. It may be raining heavily. A vehicle could be broken down near the pickup point. Traffic congestion might block safe access. A passenger may need to cross a busy road and hesitate.

A human driver interprets these situations instinctively. The driver can make eye contact, signal reassurance, reposition the car into a driveway or adjust slightly to make entry safer. These micro-decisions are based on context and judgment, not strict programming.

An autonomous vehicle operates differently. It follows mapped pickup locations and predefined safety protocols. If circumstances change, the system does not improvise in the same way a person can. The absence of a driver means there is no shared visual confirmation or verbal coordination. Minor issues can become prolonged delays simply because there is no human available to interpret nuance.

3. Efficiency Is Not the Same as Progress

Waymo’s valuation reflects investor confidence in autonomy’s economic potential. Reduced labour costs, scalable fleets and consistent performance are compelling arguments. From an operational standpoint, the technology works.

However, progress should not be measured solely by financial returns or engineering milestones. It should also consider how systems shape human behaviour and social experience. When we remove drivers from taxis, we do not simply remove labour. We remove interaction, adaptability and informal communication.

Human beings were built to connect. Even small exchanges reinforce social cohesion. A cab ride has traditionally offered a brief but meaningful arena for that exchange. As autonomous fleets expand, the question is not whether the vehicles can drive safely.

The question is whether eliminating everyday human interactions truly represents advancement, or whether it quietly diminishes something essential.