Sitecore announced MarketingIQ this morning at Symposium 2026 in Orlando. The message on screen was clear: “Marketing that builds, runs, and improves itself.”
That's quite a promise. I recorded the session and went back through it afterwards, looking beyond the polished presentation to understand what Sitecore is actually building.
My main takeaway? MarketingIQ is not really about adding more AI agents. It's about giving those agents the right context, coordinating their work and making sure they operate within the rules of your organisation.
And that is a much more interesting problem to solve.
MarketingIQ's four pillars, as presented at Sitecore Symposium 2026: access, context, action and control.
His argument is that we're moving towards another shift. Today, someone asks an assistant which product best fits their needs. Tomorrow, they might ask that assistant to book an appointment, request a demo or register for an event. In that scenario, your website is no longer serving just human visitors. It also needs to support agents acting on their behalf.
Scrunch already addresses part of this through its Agent Experience Platform (AXP), serving content in a format that AI agents can more easily understand. The next step presented on stage was WebMCP, moving beyond reading content towards interacting with websites. That last part is important, but it's also still on the roadmap.
I've been experimenting with agent-ready rendering myself, and this is where things become interesting from an architectural perspective. Making content easier for agents to consume is one thing. Allowing those agents to interact with systems, execute actions and respect the user's permissions is a different challenge altogether.
The keynote illustrated how AI assistants increasingly mediate the relationship between people and brands.
The customer perspective came from Oliver Pilgerstorfer, CMO of IFS, who participated in the early access programme. He attributed €11 million in pipeline to AI search as a first-touch channel. His ambition for MarketingIQ was straightforward: respond to a market signal in two hours instead of two weeks. That's a useful business objective. The question is how Sitecore intends to deliver it.
The inputs feeding MarketingIQ include campaign briefs, approved assets, market signals and AI search visibility. One slide represented this information as a graph connecting four domains: customer, brand, content and performance. It included everything from personas and product claims to compliance rules, tone of voice, A/B test results and SEO/AEO metrics.
Sitecore's context graph connects customer, brand, content and performance information.
What I found interesting wasn't the graph itself. A diagram doesn't tell us much about the underlying implementation. The important question is whether MarketingIQ can reliably determine which information is authoritative, current and relevant to a particular task.
An agent finding a product document is easy enough. Knowing whether that document contains the approved claims for the German market, applies to the correct product variant and hasn't been superseded is considerably harder. That's where the actual value of this context layer would be.
A quick fix: A page agent identifies an opportunity, prepares updated copy and moves the proposed change through the required approval process.
A new campaign: A brief agent creates the campaign brief and involves the team. A content agent drafts copy, while an assets agent prepares materials for the different channels. A designer remains responsible for the hero concept, and a brand lead provides final sign-off.
That's quite a promise. I recorded the session and went back through it afterwards, looking beyond the polished presentation to understand what Sitecore is actually building.
My main takeaway? MarketingIQ is not really about adding more AI agents. It's about giving those agents the right context, coordinating their work and making sure they operate within the rules of your organisation.
And that is a much more interesting problem to solve.
It starts with Scrunch, not Sitecore
Chris Andrew, co-founder of Scrunch, opened the session with an observation about his own behaviour. He visits far fewer websites these days because his AI assistant does much of the browsing for him. By the time someone reaches your website, assuming they visit at all, an AI assistant may already have helped them create a shortlist.His argument is that we're moving towards another shift. Today, someone asks an assistant which product best fits their needs. Tomorrow, they might ask that assistant to book an appointment, request a demo or register for an event. In that scenario, your website is no longer serving just human visitors. It also needs to support agents acting on their behalf.
Scrunch already addresses part of this through its Agent Experience Platform (AXP), serving content in a format that AI agents can more easily understand. The next step presented on stage was WebMCP, moving beyond reading content towards interacting with websites. That last part is important, but it's also still on the roadmap.
I've been experimenting with agent-ready rendering myself, and this is where things become interesting from an architectural perspective. Making content easier for agents to consume is one thing. Allowing those agents to interact with systems, execute actions and respect the user's permissions is a different challenge altogether.
The keynote illustrated how AI assistants increasingly mediate the relationship between people and brands.
The customer perspective came from Oliver Pilgerstorfer, CMO of IFS, who participated in the early access programme. He attributed €11 million in pipeline to AI search as a first-touch channel. His ambition for MarketingIQ was straightforward: respond to a market signal in two hours instead of two weeks. That's a useful business objective. The question is how Sitecore intends to deliver it.
What is MarketingIQ actually doing?
Danny Robinson, Sitecore's Chief Product Officer, described MarketingIQ as the contextual intelligence and orchestration brain behind SitecoreAI. Strip away the marketing terminology and two capabilities remain: contextual intelligence and agent orchestration.Contextual intelligence
MarketingIQ connects information that usually lives across multiple systems:- Campaign briefs, brand guidelines and approved messaging
- Product information and audience data
- Existing content and digital assets
- Performance data, experiments and market signals
- AI search visibility and competitive insights
The inputs feeding MarketingIQ include campaign briefs, approved assets, market signals and AI search visibility. One slide represented this information as a graph connecting four domains: customer, brand, content and performance. It included everything from personas and product claims to compliance rules, tone of voice, A/B test results and SEO/AEO metrics.
Sitecore's context graph connects customer, brand, content and performance information.
What I found interesting wasn't the graph itself. A diagram doesn't tell us much about the underlying implementation. The important question is whether MarketingIQ can reliably determine which information is authoritative, current and relevant to a particular task.
An agent finding a product document is easy enough. Knowing whether that document contains the approved claims for the German market, applies to the correct product variant and hasn't been superseded is considerably harder. That's where the actual value of this context layer would be.
Agent orchestration
The second part coordinates work across agents, systems and people. Sitecore demonstrated two possible routes.A quick fix: A page agent identifies an opportunity, prepares updated copy and moves the proposed change through the required approval process.
A new campaign: A brief agent creates the campaign brief and involves the team. A content agent drafts copy, while an assets agent prepares materials for the different channels. A designer remains responsible for the hero concept, and a brand lead provides final sign-off.
The resulting content can be prepared for email, social, web, display and paid search. It's also intended for two audiences: people consuming the experience and AI agents that need to interpret the underlying content.
What matters here is that Sitecore isn't presenting every task as something an AI agent should complete independently. Human involvement remains part of the process, which seems like a sensible direction.
But coordinating agents and people is only part of the challenge. Those agents also need access to the right information, a way to interact with existing systems, and clear boundaries around what they're allowed to do.
That's where Sitecore's four pillars come in.
Four pillars: access, context, action and control
Sitecore organised MarketingIQ around four capabilities: access, context, action and control. Together, they describe how marketers interact with the platform, how agents get the information they need, how work gets done and how organisations remain in control.
The individual concepts aren't particularly new. What's interesting is how Sitecore intends to bring them together into a single experience.
I like that approach. But making functionality available through MCP is relatively straightforward compared to maintaining consistent permissions, user identities and approval rules across different interfaces. That's where the implementation becomes interesting, especially when agents start performing actions on behalf of users.
And regardless of where those actions originate, they need to be based on information the organisation can actually trust.
The demo placed Claude alongside SitecoreAI, both working with the same MarketingIQ context.
The feedback loop shown on stage, including a fictional +4.1-point improvement in citation rate.
Sources:
The individual concepts aren't particularly new. What's interesting is how Sitecore intends to bring them together into a single experience.
Access: work where you already work
The connection between Claude and SitecoreAI is enabled through MCP and APIs, allowing marketers to work with MarketingIQ without being tied to a particular interface. Whether someone prefers Claude, ChatGPT, Copilot or SitecoreAI itself, the ambition is to provide access to the same business context and capabilities.I like that approach. But making functionality available through MCP is relatively straightforward compared to maintaining consistent permissions, user identities and approval rules across different interfaces. That's where the implementation becomes interesting, especially when agents start performing actions on behalf of users.
And regardless of where those actions originate, they need to be based on information the organisation can actually trust.
Context: not every document is the right document
This was probably the strongest part of Robinson's presentation. MarketingIQ doesn't just need access to information; it needs to understand which information is relevant, current and approved for a particular task.
Consider something as simple as updating a product page. Which products does a particular claim apply to? Has the wording been legally approved for the intended market? And is new content even necessary, or could an existing campaign or page be reused?
The same principles apply to digital assets. An image might look perfect but have an expired licence, be restricted to a particular market or use the wrong rendition for the intended channel. Sitecore is bringing the Content Hub DAM experience into SitecoreAI as SitecoreAI DAM, making approved assets and their metadata available within the same workflow.
That makes sense. Context isn't limited to text, and giving an agent access to thousands of documents and images isn't particularly useful if it cannot determine which ones it should use.
Of course, having the right information is only half the challenge. MarketingIQ also needs to turn that information into something useful.
Action: orchestration is more than generating content
MarketingIQ can start from a business objective or react to a market signal, such as a change in AI search visibility. Based on the available context, it can recommend actions and coordinate the work required to complete them.
Some tasks can run in parallel, others depend on previous steps, and certain decisions require human involvement. The demo also showed how marketers can subscribe to signals and receive updates through tools such as Microsoft Teams, Slack and email.
This addresses something I think is frequently overlooked in discussions about AI productivity. Generating a campaign brief in seconds isn't particularly useful if someone still spends the afternoon coordinating five people, three applications and several approval rounds.
If MarketingIQ can reduce that overhead, it could deliver considerably more value than content generation alone. But the more work you automate, the more important it becomes to understand what happens when something goes wrong.
Control: what happens when the source is wrong?
My favourite example from the session involved an incorrect product claim. Imagine a source document stating that a particular sofa cover fits every model in a range, when it actually fits just one. An AI agent uses that information to generate convincing product descriptions, and suddenly customers are ordering covers that don't fit.
MarketingIQ is intended to address this by allowing the original source to be corrected, identifying where the incorrect claim has been reused and coordinating updates across the affected content. It's a compelling proposition because generative AI can spread incorrect information remarkably quickly, particularly when the output looks perfectly credible.
There is an architectural challenge here, though. For this to work reliably, MarketingIQ needs to maintain relationships between generated content and its original sources. That requires provenance and content lineage, not just a record of which documents an agent retrieved.
And then there's governance. If a brand requires human approval before publishing German translations, that rule needs to apply regardless of whether someone starts the process in SitecoreAI or Claude. Using a different interface shouldn't provide a way around the approval process.
Taken together, these four pillars describe something more ambitious than a collection of AI agents. Sitecore wants to connect trusted information, coordinated execution and governance into a process that continuously learns from its results.
Which brings me to what I think is the most interesting part of MarketingIQ.
The loop is the product
The real ambition isn't to generate campaigns faster. It's to connect market signals, content creation and business outcomes into a continuous improvement process.
Signal → Context → Plan → Create → Approve → Measure → Improve
The Forma Lux demonstration showed how Sitecore envisions this working.
Scrunch identifies growing interest in small-space sofas through AI search. Competitors are appearing in AI-generated answers, but Forma Lux isn't getting mentioned. MarketingIQ recognises the opportunity and recommends a campaign to improve visibility.
Using the available business context, it prepares a brief containing the target audience, objectives and proposed messaging. Nora reviews the recommendation and approves the next steps, after which agents and people collaborate on the campaign, reusing existing materials where possible.
Rather than treating the campaign as finished once the content is published, Nora subscribes to the original signal to monitor its performance.
A few weeks later, she receives an update through Microsoft Teams. In the fictional demonstration, the AI citation rate has improved by 4.1 percentage points. MarketingIQ uses that result to recommend the next steps: creating a comparison page and refreshing the FAQ content.
And the cycle starts again.
The feedback loop shown on stage, including a fictional +4.1-point improvement in citation rate.
In the demo, Nora can start that next round of work with a single click, completing the loop from market signal to measurable outcome and back again. Of course, this was a fictional demonstration, not evidence of a customer achieving a 4.1 percentage point improvement.
Robinson also addressed a less successful scenario. What happens when a campaign increases traffic but fails to generate additional sales? In that situation, the results should inform the next brief rather than simply triggering another round of content production.
And that's where I think MarketingIQ becomes genuinely interesting. Individual AI agents that generate content aren't particularly new, but connecting market signals, approved business knowledge, execution and measurable outcomes into a continuous process is a much bigger ambition. Whether Sitecore can deliver that consistently in real-world environments remains to be seen.
What I would want to validate
I like the direction Sitecore is taking, but a keynote is still a keynote. The end-to-end experience was presented through a video rather than a live demonstration, and there's a difference between showing how something should work and proving that it does.
Before recommending MarketingIQ as part of a customer architecture, there are several areas I'd want to explore in more detail.
1. Your context needs to be reliable before your agents can be
If an organisation has three conflicting versions of a product claim, MarketingIQ needs a reliable way to determine which one is correct. That requires someone to establish ownership, identify authoritative sources and maintain the metadata that makes those decisions possible. The same applies to licensing information, product relationships, localisation requirements and content lifecycle states.
For many organisations, the biggest challenge won't be configuring AI agents. It will be getting their content and assets into a state where those agents can safely use them. MarketingIQ might help expose inconsistencies, but it won't magically resolve years of fragmented content management.
2. Governance belongs in the platform, not the prompt
Sitecore repeatedly emphasised that accessing MarketingIQ through an external assistant shouldn't allow users to bypass established permissions or approval processes. I completely agree with that principle, but it only works when those rules are enforced server-side, regardless of which agent or interface initiates an action.
I'm particularly interested in how MarketingIQ handles identity and authorisation through MCP. Does the platform operate on behalf of the individual user, and how is that identity propagated across systems? What happens when an agent attempts an action the user isn't authorised to perform, and can we inspect the resulting audit trail?
These aren't minor implementation details. They're fundamental requirements for enterprise adoption, especially when agents move beyond generating recommendations and start executing actions across connected systems.
3. Claim tracing requires reliable lineage
The ability to correct an incorrect claim and identify every piece of content that reused it sounds excellent. However, that depends on maintaining reliable relationships between the source information and the generated output.
For content created entirely through MarketingIQ, retaining those relationships seems achievable. But what happens to content created over the last ten years, or material generated in other systems and imported afterwards? I'd like to understand how far Sitecore can trace those dependencies and where the limitations are.
There's also a difference between identifying affected content and safely correcting it. Even when MarketingIQ knows where a claim originated, changing every occurrence automatically might not always be the right response.
4. AI search attribution remains difficult
Chris Andrew made a strong claim about AI search attribution, suggesting it could become a channel with complete attribution. Yet Oliver Pilgerstorfer explained that links are frequently missing and estimated that the actual impact could be ten times greater than what IFS can directly measure.
Those statements illustrate the challenge rather well. The €11 million pipeline figure is based on observable first-touch attribution, but it doesn't necessarily capture the full influence of AI-assisted discovery.
Someone might use an AI assistant to research a product, visit the website independently several days later and eventually convert through an entirely different channel. Tracking AI citations, visibility and referral traffic provides useful signals, but connecting those signals to actual commercial outcomes remains considerably more difficult.
I'd therefore be careful about treating improved AI visibility as direct evidence of additional revenue. It's an important metric, but only one part of the picture.
5. Keep the workflow predictable
Another area I'd like to explore is how Sitecore separates AI reasoning from process orchestration. Understanding market signals, preparing campaign briefs and recommending content improvements are tasks where generative AI can add considerable value.
Task dependencies, approval rules, permissions and publishing decisions are different. Those should remain predictable and auditable, regardless of how intelligent the underlying agents become.
I don't want an agent deciding that an approval step is unnecessary simply because it believes the content looks fine. Let the model interpret the information and recommend actions, while the platform enforces the rules. That separation is important if we're going to trust these systems with increasingly autonomous workflows.
6. Understand what is available today
One particularly interesting part of the announcement was the commercial positioning. According to Sitecore's newly published [MarketingIQ product information](https://www.sitecore.com/platform/marketing-iq), MarketingIQ is included with SitecoreAI CMS and SitecoreAI DAM, and existing SitecoreAI customers will receive access.
However, MarketingIQ isn't available as a standalone product, and an existing XM or XP licence alone doesn't provide access. That distinction matters for organisations still running Sitecore's traditional platforms.
During Symposium, my understanding was that MarketingIQ would become available to SitecoreAI customers the following week, without an additional MarketingIQ licence fee. Sitecore also discussed options for existing XM and XP customers running version 10.2 or later.
One route involved committing to SitecoreAI while continuing to operate the existing platform during migration. Another option mentioned was through Scrunch, which would also provide access to MarketingIQ.
The exact eligibility and commercial details of these legacy-platform routes still need confirmation, as Sitecore's public product information doesn't document them in that level of detail. Nevertheless, the broader architectural direction is relevant.
MarketingIQ is designed to connect to existing systems through MCP and other integrations. According to Sitecore, organisations don't necessarily need to replace their current CMS, DAM or CRM before they can begin using these capabilities. That could make adoption considerably more accessible, particularly for customers who aren't ready for a complete platform migration.
There's also an important distinction between what MarketingIQ can do for marketers today and what Sitecore ultimately wants to enable for customer-facing AI agents. According to [CMSWire's coverage](https://www.cmswire.com/digital-experience/sitecore-places-its-bet-on-answer-engine-visibility-with-marketingiq/), generated content still requires human involvement before publishing, while allowing customers' own AI agents to request demos or register for events remains a planned capability.
That's an important distinction when evaluating the announcement. The longer-term vision is compelling, but I'd want to understand exactly which capabilities are available now before planning an implementation around them.
My conclusion: start with the context, not the agents
MarketingIQ is probably the most coherent story Sitecore has told about AI so far. Last year, the focus was largely on introducing agents into SitecoreAI. This year, Sitecore is addressing the harder questions: what those agents know, which information they can trust, what they're allowed to do and whether their work actually delivers results.
I think that's the right direction, particularly because it acknowledges something we often overlook when discussing AI. Generating content is relatively easy. Making sure that content is correct, relevant, approved and measurable is where the real complexity lives.
It also means organisations shouldn't wait for MarketingIQ before addressing these challenges. Get authoritative product information in place, make sure digital assets have the correct metadata and usage rights, and define which decisions require human approval across different brands and markets. Structure content so that both people and AI agents can understand and use it.
None of that work depends on MarketingIQ delivering on its promise. It's simply good architecture for the way digital experiences are evolving, and it creates a foundation that other technologies can benefit from as well.
If Sitecore manages to deliver what it demonstrated, organisations that already have this foundation will be in a much better position to take advantage of it. They might actually reduce the time needed to respond to a market opportunity from two weeks to two hours.
The others risk achieving something considerably less useful: automating their existing inconsistencies.
Sources:
- Personal recording and photos from the Sitecore Symposium 2026 keynote, Orlando, 11 November 2026
- MarketingIQ: Official Product Overview
- Sitecore Places Its Bet on Answer Engine Visibility With MarketingIQ, CMSWire, 11 November 2026
- Sitecore Acquires Scrunch, Sitecore, 3 June 2026
Due to speed, written with AI. May contain flaws. Will be updated later.