Opinion: SEOs Should Not Be Responsible For The "Agentic Web"

At what point does SEO stop being SEO and simply become responsible for everything that happens on a website?

That question has been creeping up on the industry for years. Every time Google, browsers, users, or the broader web ecosystem place more importance on a particular technology or experience, SEO teams somehow end up responsible for optimizing it. Sometimes that makes sense. Often it happens simply because nobody else owns the problem.

Page performance is a good example. Core Web Vitals measure loading performance, responsiveness, and visual stability. Those are fundamentally web development and user experience concerns. But because Google incorporates Core Web Vitals into its broader page experience systems, SEO teams frequently end up monitoring them, reporting on them, diagnosing them, and pushing engineering teams to fix them. Google itself makes an important distinction here: good Core Web Vitals can contribute to search success, but chasing perfect scores purely for SEO is not necessarily a good use of resources.

JavaScript followed a similar path. Rendering architecture is an engineering decision, but technical SEOs are routinely expected to diagnose client-side rendering failures, hydration problems, blocked resources, and content that search engines cannot reliably access. UX, information architecture, digital PR, internationalization, analytics, accessibility-adjacent concerns, and even parts of product strategy have all increasingly become things SEO teams are expected to understand.

There is usually a legitimate reason.

There also needs to be a limit.

The agentic web describes an emerging version of the internet where AI agents do more than retrieve and summarize information. They can take actions on a user’s behalf.

Instead of someone manually moving between websites, comparing products, filling out forms, accessing software, or coordinating multiple services, an AI agent can potentially perform those steps itself. That might mean researching a purchase, booking a reservation, interacting with a business system, retrieving information from several databases, or completing a multi-step workflow.

We are already seeing the technical infrastructure for this take shape. Anthropic's Model Context Protocol, or MCP, provides a standardized way for AI applications to connect with tools and data sources. Google has also introduced Agent2Agent, or A2A, to standardize communication between agents built by different organizations and frameworks. Google's broader documentation now describes an ecosystem of protocols covering data access, agent communication, commerce, payments, interfaces, and other agent interactions. This also includes WebMCP for Chrome.

The important shift is that the web is beginning to move from being primarily a place where humans find information toward one where software can potentially act on information.

For SEOs, there is an obvious connection.

If an AI agent is deciding which company offers a particular service, which product satisfies a user's requirements, or which source provides trustworthy information, many familiar SEO concerns still matter. Crawlability matters. Structured information matters. Clear site architecture matters. Consistent entity information matters. Content quality and authority matter.

That does not mean SEO should own the agentic web.

SEO has a habit of absorbing adjacent disciplines whenever they begin affecting organic visibility.

More recently, AI visibility and Generative Engine Optimization have started joining that list.

None of these connections are imaginary. Google explicitly considers Core Web Vitals within its ranking systems, for example. Search engines also need websites that can be crawled and rendered properly. Links and mentions affect discovery and authority. Localization affects which content appears for users in different countries and languages.

The problem begins when involvement becomes ownership.

A technical SEO should understand how JavaScript affects crawling. That does not make the SEO responsible for the entire frontend architecture.

An SEO should understand how site performance affects search visibility. That does not mean the SEO team should own the performance engineering program.

An SEO should understand how brand mentions, citations, and authority influence visibility. That does not make SEO responsible for the entire public relations department.

This distinction becomes even more important with the agentic web because the technical scope is considerably larger.

Making a website or business usable by autonomous agents goes far beyond helping a crawler understand a page.

Consider what happens when an AI agent moves from simply reading content to actually performing a transaction.

The system may need to authenticate the user. It may need permission to access particular data. It might call an API, check inventory, submit payment information, modify account settings, or communicate with another agent. Business rules need to determine what the agent is allowed to do. Security systems need to prevent abuse. Developers need to determine what happens when something fails halfway through a workflow.

Google's own explanation of the emerging agent protocol ecosystem illustrates how broad this becomes. MCP can provide access to tools and data. A2A can allow agents to communicate. Other protocols are being developed around commerce, payment authorization, interfaces, and user interaction.

Those are not SEO problems.

They are product, engineering, security, infrastructure, and business operations problems.

SEO may have opinions about how an agent discovers the service or understands what the company offers. But once the conversation turns to authorization, payments, APIs, transactions, or application state, pretending this is simply another branch of technical SEO creates the wrong expectations for everyone involved.

Unfortunately, there is a very good reason agentic optimization will probably end up on many SEO roadmaps.

SEOs already spend their careers thinking about how machines understand websites.

We understand crawling. We think about structured information. We care about whether important content is accessible without unnecessary technical friction. We work with metadata, internal linking, canonicalization, information architecture, entities, and machine-readable markup.

That makes experienced technical SEOs useful people to have in the room when organizations start thinking about agentic systems.

In many companies, they may also be among the first people asking the right questions.

Can an agent discover our products?

Can it understand the differences between our service tiers?

Is pricing clearly represented?

Are locations, availability, and product attributes presented consistently?

Can machines reliably identify the business entities behind the website?

Those are reasonable SEO concerns.

But there is a large difference between saying SEO should participate in agentic strategy and saying SEO should be responsible for agentic readiness.

The first makes sense.

The second is how SEO becomes an organizational junk drawer.

One useful boundary might be this:

SEO should help optimize how agents discover and understand information. SEO should not own the infrastructure that allows agents to operate the business.

That would leave plenty of meaningful work for search teams.

SEOs can help ensure important information is crawlable and accessible. They can improve structured data and entity consistency. They can help organize products, services, documentation, and content so machines can interpret them accurately. They can identify situations where critical information exists only inside interfaces that automated systems cannot reasonably access.

Those responsibilities align with what SEO has always done: improving discoverability and interpretation.

But if an AI agent needs to authenticate into an application, execute a purchase, query private inventory, modify customer data, or trigger a business process, that work belongs somewhere else.

SEO should be a stakeholder. We should not automatically become the owner.

There may eventually be a broader organizational lesson here.

Instead of continuing to expand SEO indefinitely, companies may need a new type of role that sits between technical SEO, product, engineering, digital experience, and AI systems.

Call it Web Optimization.

Call it Digital Experience Engineering.

Call it Machine Visibility.

Call it something entirely different.

The title is less important than recognizing that the scope has changed.

If someone is responsible for making digital systems work well for search engines, AI answer engines, autonomous agents, browsers, humans, APIs, and whatever comes next, that role has moved well beyond the traditional definition of SEO.

Creating a broader function could actually make SEO more effective because it would establish clearer boundaries. SEO specialists could concentrate on discovery, search visibility, content interpretation, and demand capture while working alongside people responsible for the broader technical ecosystem.

That is healthier than quietly adding another responsibility to the SEO team's project list every time the web changes.

None of this is an argument for ignoring the agentic web.

Quite the opposite.

SEOs should understand how AI agents work. We should understand the protocols being developed around them. We should pay close attention to how agents discover businesses, evaluate information, interact with websites, and select sources.

There will almost certainly be implications for organic visibility.

But understanding something does not mean we need to own it.

SEO works best when its responsibilities are connected to outcomes the team can reasonably influence. Once the discipline becomes responsible for every technical, content, marketing, and product decision that might indirectly affect visibility, it becomes nearly impossible to set priorities or establish reasonable expectations.

The agentic web may become an important part of how people interact with businesses online.

SEOs should absolutely have a seat at that table.

We just shouldn't be expected to build the table too.

Diffusion Models: Generative AI models that learn to create new data, such as images, by gradually reversing a process that adds random noise to training examples.

One more thing: AI is only as good as its operator, and if you are reading this newsletter, you’re better than most!

Till next time,

PS: Let me know what you think of this issue, or anything else here: [email protected]