Agency Websites and SEO: Is Search Engine Optimization Coming to an End, or Is It Just Changing?

Imagine telling your assistant: “Find a lightweight laptop with long battery life within a certain price range.” Instead of browsing online stores yourself, the assistant will do it for you. A store that can’t clearly tell the assistant what it sells may be excluded from the selection. This concept is referred to as an “agent-friendly website”—that is, a website used not only by people but also by software agents. Some say this spells the end of traditional SEO. Available sources suggest something else entirely.

What Does “Agentic Web” Mean?

An agent is a program based on a language model that not only responds but can also take action: search for information, compare offers, and, in some cases, complete an order. When a user assigns it a task, the web agent doesn’t need to “read” like a human. It’s interested in information it can reliably process: prices, specifications, availability, and terms.

This is still in its early stages. One overview of protocols for agents explicitly acknowledges this: agent-based commerce is still new, and the protocols themselves are evolving. We have no basis for the claim that autonomous agents are currently “massively” surfing the web and that machines make up the majority of the economic web.

What Technologies Make This Possible

Instead of reading the page’s layout, standards are emerging that allow agents to communicate with websites via machine-readable interfaces. Overviews typically divide these into several layers.

Connecting the agent to tools and data

Protocols such as MCP allow an agent to connect to external data, APIs, and tools. Other protocols, such as A2A, are used to pass tasks between agents.

Connecting the agent to the web and e-commerce

Proposals such as NLWeb and WebMCP are designed to enable direct querying of the Web. For shopping, protocols such as ACP and UCP are being developed, which describe how an agent discovers products and completes a transaction, and AP2 for agent-initiated payments. The common thread is that the data is in a machine-readable, structured format.

Note on purchases: According to one description of a universal shopping cart, the transaction is still completed on the merchant’s site. This does not mean, therefore, that merchants will disappear from the process.

What the Search Engine Itself Says About SEO

On May 15, 2026, Google published its first official guide to optimization for generative search features, namely AI Overviews and AI Mode. Its key message: SEO best practices remain relevant because these features rely on the same ranking and quality systems.

• A page must be indexed and accessible to appear in search results—that is, it must meet basic technical requirements.
• The content must be crawlable.
• According to the guide, unique, useful content has the greatest impact.
• No special files for AI (such as llms.txt), rewriting content for AI, or special markup are required.

It is important to note that this guide pertains to Google’s features, not all AI assistants. Other tools may have different rules, and some documentation projects do use the llms.txt file. However, this is not a universal ranking factor. The guide also includes introductory recommendations regarding agents and references the UCP protocol.

What Actually Changes for Web Developers

The basics remain the same

Readability, speed, a clear structure, and content with intrinsic value remain important. Texts full of repetitive keywords were already useless before the advent of agents.

Data is becoming more important

For e-commerce sites and services, it will likely become increasingly important for prices, availability, specifications, and terms and conditions to be listed accurately and in a structured manner—ideally in a format that software can process. This is our conclusion based on how agent protocols are described—not a verified fact about what agents currently require.

Risks on the User’s Side

To say that an agent filters out “only verified truth” is an oversimplification. An agent can make mistakes or be influenced by manipulative data. Therefore, when making purchases, it remains important to verify exactly what terms the agent has actually approved. The responsibility for the decision still lies with the human.

Is SEO Dead?

According to Google’s official documentation, no. The AI features in its search are based on the same index and the same quality signals. However, SEO is expanding: in addition to human searches, agents and protocols are emerging that have not yet been implemented everywhere.

Conclusion

Agent-based web is a real trend, but it doesn’t spell the end of human readers or the end of SEO. The most sensible approach for creators is twofold: stick to proven fundamentals (high-quality, easy-to-navigate content) and keep an eye on emerging protocols for bots, especially in e-commerce. We don’t yet know which ones will catch on.