Optimizely’s Summer ’26 Roadmap: The CMS Is Starting to Look Less Like a Publishing Tool and More Like Marketing Infrastructure

Optimizely’s Summer ’26 Product Roadmap event was not just a list of product updates. At least, that is not the part I found most interesting.

The more useful takeaway was the way Optimizely is framing the next phase of digital experience: not only around marketers creating content, or customers consuming it, but also around AI agents discovering, retrieving, interpreting, and eventually acting on that content.

That may sound abstract at first. In practice, it is very real.

Most marketing and digital teams are already using AI in some form. Someone drafts copy in ChatGPT, Claude, Gemini, or Copilot. Someone else uses AI to summarize research, produce metadata, create campaign variations, or generate a first version of an article. Then the work moves back into the CMS, where the real operating process begins: page structure, metadata, approvals, brand review, localization, publishing, analytics, optimization, and eventually the next round of revisions.

The AI helps, but it usually sits beside the workflow. It does not really understand the CMS. It does not know the content model. It does not have reliable memory of the brand’s taxonomy, page patterns, permissions, approval rules, or performance history. So the team still carries the coordination burden.

That gap — between AI as a useful side tool and AI as part of the operating model — was the thread that ran through the sessions I attended.

The event opened with Optimizely’s product and AI leadership sharing recent product highlights and what is coming next, especially around AI. Optimizely Academy describes the Summer ’26 Product Roadmap as a 90-minute event with live demos, product deep dives, and an education track across CMS, CMP, Experimentation, Commerce, and Opal AI. From there, the event moved into product breakouts and practical education sessions. I joined the CMS breakout, followed by the SEO, GEO, AEO, and AI Visibility session, which focused on how teams can understand what AI agents are doing on their websites and how to act on that information.

For digital leaders, marketing VPs, and MarTech decision makers, the event raised a practical question: what happens when the CMS is no longer only the place where teams publish pages, but the system that helps humans, customers, and AI agents interact with the brand?

That is where this roadmap gets interesting.

The three experiences Optimizely is organizing around

One of the clearest messages from the opening session was that Optimizely is organizing its product thinking around three experiences: the marketer experience, the customer or digital experience, and the agent experience.

The marketer experience is the day-to-day reality of the teams inside the platform: content creators, campaign managers, product marketers, digital producers, SEO teams, analysts, and developers. These are the people dealing with content requests, campaign deadlines, approval cycles, localization needs, personalization rules, and reporting pressure.

The customer experience is the more familiar DXP conversation: what customers, prospects, members, partners, or buyers see when they interact with the brand. This includes websites, landing pages, product pages, personalized journeys, account-based marketing pages, forms, search experiences, and commerce flows.

The new piece is the agent experience. In the opening session, Optimizely described agents as a new type of actor in the system, one that needs the right access, visibility, and governance. That point matters. AI agents are not just another reporting segment. They are becoming part of the discovery and interaction layer around digital experiences. Optimizely’s message was that brands need to understand and optimize for them without losing control of governance, reliability, or trust.

That is a useful framing because it avoids treating AI as a generic productivity add-on. It also avoids the common mistake of talking about AI only as a content generation tool. The larger issue is not whether AI can draft a paragraph. It can. The larger issue is whether AI can work inside the same operating model that enterprise marketing teams already depend on.

And that means governance, context, workflows, permissions, measurement, and accountability.

The CMS is being asked to do a different job

The CMS breakout made this shift more concrete.

Optimizely’s CMS team described the current reality many teams face: AI is often layered on top of the CMS, not built into the workflow. A marketer drafts something in an AI tool, copies it into the CMS, manually optimizes for SEO or GEO, publishes it, looks at insights somewhere else, then comes back to make updates. There is nothing inherently wrong with those steps. The issue is that the work remains fragmented.

The AI does not know which page it is working on. It does not know the content type. It does not know the approval flow. It does not know whether the content follows the organization’s structure, metadata standards, or brand guidelines.

That is why the CMS discussion centered on the idea of an “agentic CMS.” In the session, Optimizely described this as a CMS where AI does not just sit next to the platform, but operates inside it, using context, memory, governance, and structured organizational knowledge.

The phrase “agentic CMS” can sound like product marketing language, so it is worth translating.

In plain terms, the CMS is being positioned less as a page factory and more as the structured knowledge layer for the brand. Historically, teams treated the CMS as the place where the output lived: pages, articles, landing pages, campaign content, and digital experiences. In the direction Optimizely described, the CMS increasingly becomes the place where the inputs are governed: content models, taxonomy, metadata, brand facts, assets, design systems, localization rules, and reusable content structures.

That is a meaningful distinction.

If the CMS only manages output, AI helps generate more things to publish. If the CMS manages structured knowledge, AI can potentially help create, adapt, localize, optimize, and govern digital experiences using the same context the organization already depends on.

This is also where expectations need to be realistic. The shift does not happen by turning on an AI feature. It depends on whether the organization has clean content models, disciplined metadata, usable taxonomies, clear ownership, and governance rules that can be translated into workflows. If the content operation is messy, AI may expose that mess faster than it fixes it.

For leaders who want to get ahead of this, the first move is not necessarily to launch a large AI program. It is to inspect the foundation. Are key content types clearly defined? Is metadata consistently applied? Do categories and tags make sense for both human users and AI agents? Is ownership clear when content needs to be reviewed, updated, localized, or retired? And are content creation and approval workflows documented well enough that an agent-assisted process could follow them without creating confusion?

Starting there helps teams surface weaknesses early and build a stronger foundation for AI-assisted operations. That is not a reason to avoid AI. It is a reason to clean up the operating model before scaling it.

Virtual teammates are the more interesting idea

The most memorable concept from the opening session was the idea of virtual teammates.

Optimizely described virtual teammates as more than single-purpose agents. The example shared was closer to a persistent role that can collaborate with humans, use tools the user has access to, deploy agents or workflow agents, and participate in a team process. In the demo narrative, a virtual teammate proactively created a Q&A-style asset for an article, involved human peers in the discussion, incorporated feedback, and then required human approval before publishing.

That last part is important. The useful version of agentic marketing is not “AI publishes whatever it wants.” That may be exciting in a demo, but it is not how serious enterprise teams operate.

Marketing work is collaborative by nature. There are brand reviews, legal considerations, SEO inputs, product messaging changes, accessibility checks, regional market needs, and approval workflows. The more regulated or complex the organization, the more important those controls become.

This is why the human-in-the-loop workflow conversation stood out. Optimizely discussed bringing workflows into Opal so agents and virtual teammates can operate with the right checks in place before publishing. The event also referenced brand packs and channel packs, which are intended to help organize brand information and adapt assets for different channels.

That is a healthier framing of AI adoption. It is not about replacing the marketing team with agents. It is about identifying which parts of the workflow can be delegated, accelerated, or monitored — while keeping humans responsible for judgment, risk, quality, and final approval.

For a marketing VP, this should be the real evaluation question: can this reduce operational drag without creating governance debt?

The answer should not be based only on whether the demo looks impressive. Teams should look at practical indicators: workflow cycle time from brief to publication, where approval bottlenecks appear, how often content needs rework after review, how many exceptions still require manual intervention, and whether agent-assisted outputs have a clear audit trail. These are not glamorous metrics, but they reveal whether AI is improving the operating model or simply adding another layer of activity.

If the answer is yes, there is business value. If the answer is “we can generate more content, but nobody knows who approved it or why,” then the organization has only moved the bottleneck.

CMS 13 and Opal are becoming part of the same conversation

The CMS breakout also spent time on CMS 13 and how it connects to the broader Opal and agentic CMS direction. In the transcript, CMS 13 was described as fully available and positioned as a release that brings customers closer to the newer Opal and Visual Builder experience. Optimizely’s CMS 13 documentation describes it as a cloud-first content management system with architecture and capabilities intended for content managers, administrators, and developers. Optimizely also describes Opal in CMS 13 as an agent orchestration platform embedded in CMS 13 that helps users explore questions, automate common tasks, and make decisions without leaving the CMS.

For decision makers, the point is not to memorize every feature. The more useful takeaway is that Optimizely is trying to connect content creation, content operations, workflow, search, analytics, and AI assistance into a more coherent operating environment.

The Page Builder agent is a good example. In the session, Optimizely described it as an agent that can look at existing pages and content types, understand how pages are usually composed, and help create a first version of a page from different inputs such as a prompt, brief, CMP task, or design.

That could be useful, but not because marketers are desperate to avoid drag-and-drop interfaces. The real value is reducing the distance between campaign intent and publishable experience. In many organizations, a campaign brief still has to travel through too many disconnected steps before it becomes a page. If AI can help produce a strong first version inside the CMS, using the right content types and patterns, the team can spend more time refining the experience and less time assembling the basics.

The same is true for CMS and CMP integration. A better connection between planning, task management, preview, approval, and publishing is not glamorous, but it matters. Content operations usually break in the handoffs. Any roadmap that reduces those handoffs deserves attention.

Still, there is a caution here. “Available,” “in beta,” “coming soon,” and “roadmap” are not the same thing. The event included a mix of recent releases, current capabilities, and future direction. Any team planning around these capabilities should validate availability, licensing, dependencies, and timing directly with Optimizely before building a business case or delivery plan around them.

That is not skepticism. That is just good MarTech governance.

AI visibility is becoming part of the discoverability conversation

The SEO, GEO, AEO, and AI Visibility session was probably the most practical part of the event for marketing leaders.

The session started from a reality many teams are already noticing: discovery is changing. The traditional model was simple enough to understand. A person searched in Google, saw links, clicked through to a site, and then the website had a chance to educate, persuade, or convert. That model still exists, but it is no longer the only path.

More people are asking questions inside AI platforms. AI systems may retrieve information from a website, summarize it, cite it, or use it to answer a user without the user ever visiting the site directly. The session framed this as a world where every website has two audiences: human visitors and AI agents.

That is a useful way to explain the issue to leadership teams. Traditional analytics tools are still valuable for human behavior, but they may not tell the full story of server-side AI agent activity. If an AI system retrieves your content to answer a question, that interaction may not look like a normal website visit in client-side analytics.

This is where Optimizely’s Agent Visibility Analytics discussion came in. The session described two types of AI visibility tooling. The first is CDN log-level data, which is factual data showing AI agent requests to specific pages. The second is Conductor-style AI visibility data, which helps show how the brand appears across AI search experiences compared with competitors. The session was careful to explain that the Conductor-style data is simulated or panel-style, while log-level data is factual request data.

Optimizely’s support documentation describes the Agent Visibility Analytics dashboard as a way to view AI platform traffic and intent, powered by CDN log data and Opal-generated insights. It also states that the dashboard can help teams track AI traffic trends and intent, identify which AI platforms visit the site most often, and see which webpages are most popular among those platforms. Optimizely Academy’s SEO, GEO, AEO, and AI Visibility session description also frames the topic around seeing what AI agents are doing on a website and integrating technical optimization into content creation and publication workflows.

That distinction is worth keeping. There is already plenty of noise in the AEO and GEO space. Some tools make strong claims about “AI visibility” without making it clear whether they are measuring real traffic, simulated prompts, competitive panels, citations, or something else. A mature strategy probably needs more than one signal.

The way I would explain it is this: log data helps you understand what AI agents are doing on your site; AI visibility data helps you understand how your brand is showing up in answer experiences. One tells you about behavior on your owned property. The other gives you a view into market visibility.

Both matter, but they answer different questions.

From visibility to action: the part teams should care about

The more valuable part of the AI Visibility session was not the dashboard itself. Dashboards are easy to admire and hard to operationalize. The interesting part was the workflow from visibility to action.

In the demo, Optimizely showed how Agent Visibility Analytics can help identify which AI platforms are visiting a website, which pages they are hitting, and why those requests may be happening. The session described intent categories such as training, search/indexing, and user action or real-time retrieval. The example then showed Opal being used to ask questions of the underlying data, such as identifying pages that are being trained on heavily but not indexed. From there, the workflow could point to agents that help with next steps, such as metadata implementation, FAQ creation, or content refresh.

That is where the business value starts to appear.

A marketing team does not need another isolated AI report. They need a way to decide what to fix first. Which pages are being crawled but not surfaced? Which pages are invisible to agents? Which topics are competitors winning in AI answers? Which content needs clearer structure, better metadata, stronger factual answers, or FAQ formatting? Which product pages are important to the business but not showing up in AI-driven discovery?

A practical prioritization model starts with business impact. Revenue-generating product pages, high-intent comparison pages, critical support documentation, and campaign landing pages should usually receive more attention than low-value archive content. From there, teams can weigh effort against potential gain: some fixes may be as simple as improving metadata, adding clearer Q&A structure, or refreshing outdated content; others may require deeper content modeling, technical SEO, or workflow changes. The last filter is strategic timing. If a major campaign, product launch, or competitive push is coming, AI visibility work should align with that moment rather than sit as a disconnected optimization task.

For example, imagine a B2B software company where pricing and product feature pages are major lead drivers. The team might start by making sure those pages are structured with clear metadata, direct answers, and FAQ sections so both human users and AI systems can interpret them more easily. At the same time, they might notice that a set of long-tail support articles are often surfaced by AI platforms but contain outdated information. Because updating those articles requires relatively low effort and may improve how AI systems interpret them, those become a sensible next priority. Then, with a product launch planned for the next quarter, the team shifts focus to refreshing content related to the new features so that AI-driven answers and competitive comparisons are less likely to rely on stale information.

That kind of prioritization helps teams avoid spreading effort too thin. It also keeps AEO and GEO from becoming another quarterly audit that produces a deck and then disappears into backlog purgatory. The better model is a continuous operating loop: detect, prioritize, improve, measure, and repeat.

That is easier said than done, of course. Teams still need ownership. They need content standards. They need technical support. They need to know which recommendations are worth acting on and which ones are noise. But the direction is right: visibility alone is not enough. Actionability is the real test.

Why this matters beyond the product roadmap

The practical implication is that this is not just a CMS or AI feature conversation. It touches how marketing teams plan work, how content gets governed, how digital experiences are measured, and how brands remain discoverable when AI systems increasingly sit between the customer and the website.

That is why the roadmap matters beyond product teams. It gives marketing, digital, and technology leaders a reason to revisit some operating-model questions that are easy to postpone until they become painful.

The first is whether AI in MarTech is still being treated as task assistance or whether it is starting to become workflow orchestration. The distinction matters. Task assistance helps someone write faster. Workflow orchestration helps the team move work through planning, creation, approval, publishing, measurement, and optimization with less friction.

The second is whether the CMS is being treated as infrastructure or merely as a publishing destination. There has been a tendency in some organizations to treat the CMS as plumbing: important, but not strategic. That view is becoming harder to defend. If the CMS becomes the structured knowledge layer that supports pages, personalization, AI agents, search, localization, and content governance, then CMS maturity becomes a business capability, not just a technology concern.

The third is whether discoverability is being measured broadly enough. Traditional SEO is still relevant, but it is no longer the whole picture. Brands need to understand how AI systems retrieve, interpret, and cite their content. They also need to know where competitors are gaining visibility in AI-generated answers.

The fourth is whether governance is strong enough to support agentic workflows. Human-in-the-loop approvals, brand rules, role-based access, auditability, and clear ownership are not administrative overhead. They are what make AI-assisted work usable in an enterprise environment. Without them, teams may move faster, but not necessarily better.

And the final point is probably the most practical one: readiness matters. The organizations that will get the most value from this direction are not necessarily the ones that adopt the newest AI capability first. They are the ones with cleaner content models, clearer governance, stronger content operations, and a realistic path for adoption.

A few questions marketing leaders should ask now

Before jumping into agentic content workflows, I would start with a few plain questions.

Is our CMS structured well enough for AI to use it responsibly? If page types, metadata, taxonomies, and reusable content are inconsistent, AI will have a weaker foundation to work with.

Do we know which content actually matters for AI discoverability? Not every page deserves the same optimization effort. Product, pricing, support, documentation, comparison, and educational content may play very different roles depending on the business.

Can we tell the difference between human traffic and AI agent activity? If not, the organization may be missing a growing part of the discovery picture.

Do we have approval workflows that can support AI-assisted work? If an agent proposes metadata updates, FAQ content, or page changes, who reviews them? Who approves them? Who is accountable?

Are our teams ready to work differently? This may be the least technical question, but it is usually the hardest one. AI adoption changes responsibilities, not just tools.

Change management should start with a clear definition of what AI is expected to improve, which teams are affected, how approvals will work, and how success will be measured. From there, teams should pilot the new workflow in a controlled area, gather feedback, adjust roles and responsibilities, and only then scale it more broadly. Most importantly, leaders should communicate early and often about what is changing and why, while making space for the people doing the work to point out where the process still needs adjustment.

Final thought

What stood out from Optimizely’s Summer ’26 roadmap was not one individual feature. It was the larger direction: CMS, Opal, analytics, content operations, experimentation, personalization, and AI visibility are being pulled into a more connected story.

That story is not “AI will do marketing for you.” Good. That would be too simplistic.

The better story is that marketing teams are being pushed to rethink how digital experiences are created, governed, discovered, and optimized when both humans and AI agents are part of the audience.

For some organizations, that will feel early. For others, especially those already struggling with content scale, localization, campaign velocity, or SEO/AEO uncertainty, it will feel overdue.

Either way, the conversation has moved. The CMS is no longer just where content gets published. Increasingly, it is where the brand’s digital knowledge, governance, and AI readiness come together.

That is the part worth paying attention to.

References / Sources

Optimizely Summer ’26 Product Roadmap notes, July 9, 2026.

Optimizely Academy, “Summer ’26 Product Roadmap Virtual Event.”
https://academy.optimizely.com/student/page/3469092-summer-26-product-roadmap-virtual-event

Optimizely Academy, “SEO, GEO, AEO, and AI visibility: Using agentic workflows and intelligence to optimize your content.”
https://academy.optimizely.com/student/page/3499549-seo-geo-aeo-and-ai-visibility-using-agentic-workflows-and-intelligence-to-optimize-your-content

Optimizely Developer Documentation, “CMS 13 overview.”
https://docs.developers.optimizely.com/content-management-system/v13.0.0-CMS/docs/cms-13-overview

Optimizely Developer Documentation, “Optimizely Opal in CMS 13.”
https://docs.developers.optimizely.com/content-management-system/v13.0.0-CMS/docs/optimizely-opal-in-cms-13

Optimizely Support Help Center, “Agent Visibility Analytics dashboard in Optimizely Analytics.”
https://support.optimizely.com/hc/en-us/articles/45787354895757-Agent-Visibility-Analytics-dashboard-in-Optimizely-Analytics