How to migrate to SitecoreAI: what AI automates and what it doesn't

Migrating to SitecoreAI means moving from a content management system running on your own servers to one managed by Sitecore in the cloud.
Whether you are coming from on-premise XP or XM, or another CMS, AI takes some of the workload off your plate: Pathway reads your installation, proposes where each piece of content should go, and imports it. What it doesn't take off your plate is the rest of the site.
We’ll walk you through what the AI automates, what it doesn't touch, and what you need to decide before getting started.
Before you begin: SitecoreAI is not a new product
SitecoreAI is XM Cloud, rebranded and expanded. You aren't choosing between two different Sitecore platforms.
You are moving from a system that runs on your servers to one managed by Sitecore, where the public website and the content manager are separated. The fundamental change isn't just a version update; it’s a shift in architecture.
And the clock is ticking, though it doesn't tick the same for everyone.
Source: Sitecore Product Support Lifecycle page
If your installation is fifteen years old, you’re likely not on 10.4; you’re on 9.x or 10.1. The problem there isn't just that you’ll stop receiving new features in 2027. If you’re on 9.x, you are already without security patches at any price; if you’re on 10.1, that window closes in December 2026.
What SitecoreAI Pathway migrates and what remains human work
Pathway is a content migration tool.
Three practical facts about access
- It is included in the Sitecore 360 subscription at no additional licensing cost.
- It is not enabled by default. You must request access from your Sitecore account manager.
- It is currently on version 1.4 (July 2026). Since version 1.3 (April 2026), it also migrates from non-Sitecore websites, though this is still in beta.
Pathway exclusions
- code, renderings, and the presentation layer
- images and files from the media library
- page designs and partial designs
- xDB data, the XP visitor behavior database
- email marketing content
- personalization
- workflow states
- custom modules and integrations
Pathway Limits
To that list, we must add three limits that are not listed as exclusions but still condition the entire project.
- Pathway does not build the destination structure: it places content into templates and components that must already exist. Sitecore documentation states it plainly: templates must be recreated by the client in the SitecoreAI environment. If no one has modeled the destination, the AI has nowhere to put anything.
- Nor you correct the field-by-field mappings: You do edit the extracted data and the instructions you give the AI, and then you re-run the migration. That is the real loop: you execute, review, adjust instructions, and repeat. Version 1.4 improved exactly this part, because migration metrics reset when re-running with new instructions instead of accumulating.
- Inherited fields are not parsed during the destination export: and that affects component mapping. If your template tree has been inheriting for fifteen years, this is not a footnote. It is planning.
Pathway moves the content and your team builds the site where that content lands.

Run a pilot before committing to a schedule
Some clients come to us because they had failed migrations even on single-page test sites, unable to determine if the cause was the configuration, a mismatch in a component's mapping, or a version limitation.
Add to that the fact that it doesn't correct things manually, and the conclusion is clear: test Pathway with a real section before setting dates in a plan. A section with real content, its templates, and its languages. What you learn there is what tells you what the rest is worth.
Start with a serious assessment
Pathway migrates, it doesn't audit. It doesn't tell you what you have set up, what you actually use, and what you can turn off.
You create that inventory yourself, beforehand, and every line ends in one of three decisions: migrate, replace, or retire.
- xDB history is archived externally or retired. Your visitors' behavioral data in XP does not carry over to the new platform. The CDP data model is different, and there is no bridge. Decide what you export to your own warehouse and which segments you rebuild from scratch. This is the point that almost no one talks about and that everyone discovers too late, with the project already underway and marketing asking about their audiences.
- Images and files migrate using the XM→XMC migration tool. They are outside of Pathway. Someone has to move them and, in the process, decide whether they end up in the media library or a DAM. It is the best time to throw away what you no longer use. A nuance from the latest version: Pathway 1.4 added image migration in July 2026, but only for the "Any website" source type. If you are coming from XM or XP, it still does not apply to you.
- Page designs and partial designs are rebuilt. Excluded as well. They are redone as part of the template work.
- Personalization is replaced. XP rules do not automatically convert into Sitecore Personalize experiences. Each rule is translated by hand: who you are talking to, what you show them, and the criteria used to decide.
- The internal search engine is replaced. Sitecore Search indexes differently. It is not just a matter of changing a setting: it is rethinking what content goes into the index and how it is queried.
- Forms and emails are redone. EXM is retired, and email moves to the marketing tool you already use. Sitecore Send remains in the portfolio, with a nuance that is worth clarifying before your client asks: in June 2025, Sitecore sold Moosend, the underlying technology, to Constant Contact, and Moosend continues to provide service to Sitecore Send.
- Automations and approval flows, one by one. Marketing Automation has no direct migration. Some cases are redone with Agentic Studio Flows and Personalize decisioning, and others are rethought. Approval workflows are rebuilt, and this is the time to simplify the ones no one was using. This is where half the work that wasn't in the initial budget appears.
In ThielmannBefore touching the website, we hold a workshop with heads of various departments to define their buyer personas and map out each customer journey. Without this step, you end up migrating content no one asked for and keeping sections that are only there because they’ve been there for six years.
Finish this inventory before requesting access to Pathway. The tool accelerates the part you’ve already solved and doesn't touch the part you haven't.
Design a coexistence strategy
Almost no one moves a large website in a single weekend. The standard pattern has a name: the strangler fig.
You build the new site alongside the old one and move sections over one by one, using edge-based routing to decide which platform serves each URL. The old site gradually loses traffic until it can be shut down.
Sitecore applies this same pattern in its own transition from XP to modern .NET.
What you need to resolve while both environments coexist
- Session and login. If there is a private area, decide where authentication lives. Prevent the user from crossing the border between platforms without realizing it.
- Analytics. Two platforms, one report. Define your tracking before the first cutover, not after.
- SEO. Canonicals, sitemaps, and redirects are managed at the border. The real risk of migrating in phases is having duplicate URLs for weeks.
An order that works: start with a section that has enough traffic to learn from but few dependencies. The blog or resource area is usually a better first cut than the product catalog.
Take care of code governance
When moving to the cloud, you lose the server. What you cannot lose is control over changes.
The practice Sitecore recommends is SCS (Sitecore Content Serialization): templates, components, and configuration become files in your Git repository. It stops being something someone tweaked in an environment on a Tuesday afternoon. It becomes a change with an author, a reviewer, and a date.
Avoid this mistake when migrating in SitecoreAI
A common mistake in projects coming from XP: forcing Helix onto everything, including the front-end.
The Sitecore Accelerate Cookbook for XM Cloud recommends the opposite. Helix is still valid for .NET code, serialization, and SXA modules. In the web application, follow the framework's conventions.
This carries twice the weight when you aren't the only one making the decision.
In The Caregiver's Corner, by the Hartmann Group, the challenge was fitting into the global IT structure led from Germany and their deployment pipelines. With structural content in Git, that alignment is just a code review. Without it, it’s a list of manual changes that someone has to replicate across three environments and verify by eye.
.jpg)
Think about GEO, not just SEO
More and more people are looking for providers by asking an assistant instead of typing into Google. A migration is the perfect time to prepare your content for that, since you'll be touching the templates anyway.
SitecoreAI comes with an AEO/SEO Researcher agent out of the box. Use it, but don't expect it to do the heavy lifting for you.
Structured data markup (JSON-LD) is half the battle. The other half: content that can be cited, consistent entities (the same product name, the same spec sheet, and the same data across the entire site), and measuring how much traffic is already coming from Perplexity or ChatGPT.
Without measurement, GEO is just a trend. With measurement, it’s a project you can budget for.
What you lose when moving to the cloud
Everything mentioned above is what you gain. It’s also worth mentioning what goes away.
- You lose control over the calendar. SaaS means Sitecore updates when Sitecore decides. There are no maintenance windows to negotiate or versions to stay on for three years. Your team shifts from planning upgrades to monitoring changes that have already happened.
- You lose the "this is where my server ends" boundary. Platform, API, environment, and volume limits appear, which you used to solve by adding more hardware. Now, they are solved by redesigning or talking to Sitecore.
- You lose access to the machine. No more checking server logs or applying a patch by hand on a Friday night. You gain stability, but you lose room for maneuver.
None of the three are reasons not to migrate. They are all reasons to know about them beforehand rather than discovering them during the first incident.
What really drives the cost
Pathway handles the mechanical work of moving content, but not the validation. It is reviewed by template groups, with human validation at every level, so more content and more templates mean more validation. Volume counts. What really drives the figure up, however, is something else:
- Integrations. Every connection to an ERP, PIM, CRM, or search engine must be rebuilt against new APIs. This isn't just our estimate: custom modules and integrations are on the official list of what Pathway does not migrate. They define the project's critical path.
- Languages and markets. Six languages don't cost six times as much, but they do multiply the validation required. Pathway 1.4 improved multi-language migration from XM and XP, and it maintains language context during extraction, mapping, and import. It helps, but it doesn't solve it entirely.
- Active personalization. If you have rules running in production today, add the cost of translating them and measuring them all over again.
- The current state of the code. A well-organized project has a head start over one that has been patched for years.
When you ask for a quote, request that these four lines be itemized separately. A single lump sum gives you no room to negotiate.
And there is a fifth line that rarely appears in any budget: the cost of staying where you are. Paid extended support will be a real expense starting in 2028 for anyone still on 10.4. For those on 9.2 or 9.3, there is no invoice that can fix it: the cost is running without security patches.
Migrating costs money. Not migrating does too.
We are here to help
We are a Sitecore partner in Spain and work with the platform on B2B projects.
Tell us where your project is starting from, and we will tell you what you really need.
