AI Agents as the New Search Engines: What This Means for Traffic

For two decades, Google was the front door to the internet. If you wanted customers, you optimised for Google. That model is fracturing.

A growing number of consumers now ask their AI assistant to research, compare, and sometimes purchase on their behalf. The assistant does the searching, filtering, and shortlisting. The human just makes the final call, or in some cases, delegates that too.

This changes the rules for how websites attract visitors. And most businesses haven't updated their playbook.

From Click-Through to Agent-Through

Traditional search works in a predictable loop. A user types a query, scans results, clicks a link, visits your site, and either converts or bounces. You can measure every step.

Agent-mediated discovery works differently. A user asks their agent a question like "find me a good espresso machine under £400 with low maintenance." The agent queries multiple sources, sometimes visiting websites directly, sometimes pulling from APIs and structured data feeds. It synthesises the information and presents a recommendation.

Your website might be consulted during this process without ever registering a traditional visit. The agent may read your page, extract the relevant data, and present it to the user alongside competitor offerings. If the user then clicks through, it arrives as direct or referral traffic with little context about the agent's role.

The Traffic Numbers Are Already Shifting

Several data points suggest this shift is material, not speculative.

Web analytics firms tracking user-agent strings report that identifiable AI agent traffic grew between 300% and 500% across most commercial sectors during 2025. That growth hasn't slowed.

Search engine click-through rates for informational queries have dropped measurably as users get answers directly from AI assistants. For commercial queries, the drop is less severe but accelerating. This shift also distorts traditional engagement metrics, as we explore in why your bounce rate might be an agent problem.

A European travel agency shared anonymised data showing that 18% of their bookings in Q4 2025 involved at least one touchpoint with an AI agent during the research phase. In Q4 2024, that figure was under 4%.

What Agents Value vs. What Google Values

Google's ranking algorithm considers hundreds of signals including backlinks, page speed, content depth, and user engagement metrics. AI agents care about different things.

Data clarity over content length. Google rewards thorough, long-form content. Agents want specific, structured answers. A 3,000-word buying guide matters less to an agent than a clean product schema with accurate specifications.

Machine readability over visual design. A beautifully designed product page means nothing if the underlying HTML is a tangle of divs with no semantic meaning. Agents need headings, lists, and schema markup they can parse reliably.

API access over page rendering. Some agents prefer to pull data from APIs rather than scraping web pages. If you have a public or semi-public product API, agents can access your data more efficiently and are more likely to include you in results.

Data freshness over domain authority. An agent comparing prices will always prefer up-to-the-minute pricing data over cached information from a high-authority domain. If your competitor's API returns live stock levels and your site shows yesterday's data, the agent trusts your competitor more.

Two Traffic Strategies, Not One

Smart businesses are running parallel traffic strategies now. The first is their existing SEO and paid search programme for human visitors. The second is an agent accessibility programme for AI-mediated traffic.

The agent accessibility programme typically involves:

Structured data investment. Going beyond basic Product schema to include detailed specifications, reviews, FAQs, and availability data in machine-readable formats. JSON-LD is the preferred format because it's cleanest for agents to parse. Our guide on Schema.org markup for AI agents covers the specifics.

Server-side rendering or static generation. Ensuring that the HTML response from your server contains complete, useful content without requiring JavaScript execution.

Consistent data across channels. Making sure your website, API, Google Shopping feed, and any other data sources all reflect the same prices, stock levels, and product details.

Agent-specific monitoring. Tracking AI agent traffic separately and understanding which agents visit, what content they access, and whether they get complete information.

The SEO Parallel

This mirrors what happened with SEO in the early 2000s. Initially, most businesses ignored search engine optimisation because search traffic was a small percentage of their total. By the time search became dominant, the early movers had built massive advantages in rankings, content, and authority.

The businesses investing in agent accessibility now will be in a similar position. When agent-mediated transactions become a significant revenue channel, and current trends suggest that's a matter of when, not if, the sites that agents already trust and prefer will capture the majority of that traffic.

The window for early adoption is still open. But it's closing faster than most businesses expect.