Adobe Is Testing a Tool That Personalizes Web Pages for Individual Visitors
Adobe is testing an experimental tool called Project Marshvale that could allow brands to tailor the language and selected elements of a web page to individual visitors. Rather than showing every person the same static homepage or landing page, the system is designed to use visitor signals and AI agents to make subtle changes that make the content feel more relevant to the person viewing it. The project remains experimental and has not been announced as a generally available Adobe product.
The idea was reported by Marketing Brew on November 12, 2026, following an early demonstration by Adobe. According to the report, Project Marshvale was developed through Adobe’s internal Sneaks program, which encourages employees to create and showcase technologies that may eventually become part of Adobe’s products.
What Project Marshvale Does
Project Marshvale is intended to make “subtle changes to the language of a page” based on the visitor’s persona. It uses a small amount of JavaScript, along with AI and AI agents, to adjust certain elements of a web page according to what Adobe believes will resonate with that visitor.
The reported example concerns Adobe’s own homepage. The standard version includes language about understanding information from documents using AI summaries, audio overviews, presentations, and interactive visual reports. Under the personalization concept, a healthcare marketer might instead see wording such as “Understand what matters from your patient education documents with AI summaries [and] audio overviews.” The page’s underlying product or message remains the same; the framing and examples are adjusted to fit the visitor’s likely professional context.
This is different from a website simply offering several unrelated versions of itself. Adobe’s stated aim is to reduce the burden on marketing teams, which often create many versions of landing pages, homepages, and campaigns to address different audiences. Joshua Hailpern, Adobe’s director of engineering for enterprise GenAI experience, told Marketing Brew that the goal is to let marketers focus on the core message while the technology helps make it feel relevant to different visitors.
How the Personalization Works
The reported system relies on visitor “personas” and signals to determine how a page should be adapted. Adobe has not publicly provided a complete technical specification for Project Marshvale, so it would be inaccurate to claim that it uses a particular named data set, model, or decision process beyond what Adobe and the reporting have disclosed.
What is clear from the report is that the tool is meant to adjust page language rather than fundamentally replace the page with a different product or service. In the demonstrated case, the changes were focused on making examples and wording more relevant to the visitor’s professional role. A small-business owner might see content connected with social video, while a healthcare marketer might see content connected with the healthcare sector.
The use of AI agents is notable because it suggests that Adobe is exploring a more automated approach to web personalization. Instead of requiring marketers to manually build every possible audience-specific variation, the system could potentially generate or select contextually appropriate wording for different visitor types. However, the available reporting does not establish that the tool independently writes entire pages, makes unsupervised decisions, or is available for customers to deploy today.
Experimental Status
Project Marshvale is currently described as experimental. It was created as part of Adobe’s Sneaks program, an internal initiative that surfaces employee-developed concepts and technologies.
Adobe principal evangelist Eric Matisoff told Marketing Brew that roughly 30% to 40% of tools developed through the Sneaks program eventually appear in Adobe products “in some form.” That figure describes the program’s historical pattern as reported by Adobe; it does not mean that Project Marshvale has a 30% to 40% chance of becoming a product, nor does it establish a launch date, pricing model, or customer availability.
Adobe’s own documentation confirms that it often enables new Target features for selected customers for testing and feedback before broader release. But the reporting on Project Marshvale does not say that it is a beta feature within Adobe Target, nor does it identify a customer-testing group, rollout schedule, or product name under which it might eventually ship.
How It Fits Adobe’s Existing Tools
Adobe already sells personalization technology through Adobe Target, part of Adobe Experience Cloud. Target supports personalization across websites, mobile sites, apps, social media, and other digital channels. Its capabilities include A/B testing, multivariate testing, rules-based targeting, Experience Targeting, Auto-Target, Automated Personalization, and Recommendations.
Adobe Target’s Automated Personalization uses machine learning to deliver personalized experiences based on visitor activity and profiles. Auto-Target identifies high-performing marketer-defined experiences and delivers the most tailored experience to each visitor based on individual profiles and the behavior of similar previous visitors. Recommendations can display products or content based on a visitor’s prior activity.
Project Marshvale appears to extend this broader direction by emphasizing AI-generated or AI-assisted changes to page language. The distinction matters: Target’s established personalization features generally select among experiences, offers, content, or recommendations that marketers have defined, while the reported Marshvale demonstration specifically describes AI and agents adjusting page language based on visitor personas.
Adobe’s existing product materials also describe the ability to personalize same-page and next-page experiences in real time, tailor content and product selections, and use AI-powered experimentation to identify effective content for audiences and individuals. These capabilities provide context for why Project Marshvale is plausible as an Adobe research direction, but they do not themselves prove that Marshvale is commercially available or functionally equivalent to Target.
Why It Matters for Marketers
For brands, the potential appeal is efficiency. Companies commonly invest heavily in landing pages, homepage design, campaign messaging, and audience-specific variations. If a system can adapt wording to a visitor’s likely needs, it could reduce the amount of manual variation work required while preserving a brand’s central message.
For example, a software company might want a homepage to speak differently to a small-business owner, a healthcare marketer, a financial-services professional, and an enterprise technology buyer. Each group may care about the same underlying product, but they may respond to different examples, priorities, and terminology. Project Marshvale’s reported approach is to make those differences visible in the page language rather than requiring a separate page for every audience.
The concept also reflects a wider shift in marketing technology toward AI-assisted content and campaign automation. Adobe has recently introduced or expanded several AI-related offerings, including Brand Concierge, LLM Apps, and CX Enterprise Coworker. According to Marketing Brew, Brand Concierge uses agents to build customer-facing chat features for brand surfaces, while CX Enterprise Coworker is an agentic AI tool intended to help with campaign automation.
Potential Concerns
Personalized web pages can improve relevance, but they also raise practical questions around privacy, consent, transparency, accuracy, and brand control. The available reporting does not state which visitor signals Project Marshvale uses, how personas are inferred, what data is retained, whether users can opt out, or how Adobe would prevent inaccurate or inappropriate personalization. Those details are therefore unknown and should not be assumed.
There is also a difference between personalizing a page and manipulating it. A visitor may find it useful when a software page uses examples relevant to their industry. But if the wording becomes overly targeted, misleading, or inconsistent with the brand’s actual offering, it could undermine trust. Adobe’s reported framing emphasizes “subtle” changes that feel resonant “without being overtly preachy,” which suggests that the company is aware of this balance, though that does not establish how the final product would manage such risks.
Regulatory and platform-level privacy rules may also affect any eventual deployment. Personalization systems that rely on behavioral data, identity signals, or cross-site context can face constraints depending on the jurisdiction, the data involved, and the way consent is obtained. The public reporting reviewed here does not provide evidence about Project Marshvale’s data practices, so no specific compliance claims can responsibly be made.
What Is Not Known
Several important questions remain unanswered:
Whether Project Marshvale will become an Adobe product, a feature inside an existing product, or remain an internal experiment.
Which visitor signals and persona data it would use in a production environment.
How brands would review, approve, limit, or audit AI-generated page changes.
Whether the tool would be integrated with Adobe Target, Adobe Experience Platform, Adobe Journey Optimizer, or another Adobe product.
When, if ever, it would become available to Adobe customers.
At this stage, the accurate description is that Adobe is testing an experimental concept, not that it has launched a broadly available personalized-web-page product.
The Bigger Picture
Project Marshvale is significant because it points toward a possible next stage of web personalization: pages that adapt not only their offers or layout, but also their language, to the person reading them. If developed responsibly, such technology could make digital experiences more useful by reducing generic messaging and surfacing more relevant examples.
But its value will depend on execution. Brands will need clear controls over what AI can change, safeguards against inaccurate personalization, and transparent data practices. For now, Project Marshvale should be understood as an early look at Adobe’s thinking about AI-driven web personalization—not as a confirmed product release.
Sources
Marketing Brew — “Adobe is testing a tool that tailors web pages to individual visitors,” November 12, 2026: https://www.marketingbrew.com/stories/adobe-tailoring-web-pages-project-marshvale
Adobe Experience League — “What is Target? | Adobe Target,” updated October 1, 2026: https://experienceleague.adobe.com/en/docs/target/using/introduction/intro
Adobe — “Adobe Target | A/B Testing Software”: https://business.adobe.com/in/products/target.html
Adobe — “Personalised Interactions | Adobe Target”: https://business.adobe.com/in/products/target/personalized-interactions.html
Adobe — “Experimentation and Optimization | Adobe Target”: https://business.adobe.com/products/target/experimentation-optimization.html
Adobe — “Adobe Target | Product description,” updated August 7, 2026: https://helpx.adobe.com/legal/product-descriptions/adobe-target.html
Adobe Experience League — “Create web experiences | Adobe Journey Optimizer,” updated October 19, 2026: https://experienceleague.adobe.com/en/docs/journey-optimizer/using/channels/web/create-web
Adobe — “Adobe Target Announces Enhanced Analytics Measurement for AI-Powered Testing and Personalization,” July 29, 2020: https://business.adobe.com/blog/the-latest/adobe-target-announces-enhanced-analytics-measurement-for-ai-powered-testing-and-personalization
Adobe — “Adobe Announces Same Page Enhanced Personalization with Adobe Target and Real-time Customer Data Platform,” October 5, 2021: https://blog.adobe.com/en/publish/2021/10/05/adobe-announces-same-page-enhanced-personalization-with-adobe-target-real-time-customer-data-platform
Comments