AI-assisted Content Migration to Storyblok in weeks, not months

Over the last twenty years I have been helping large brands govern, orchestrate and move their content from one content management system to the next. Long enough to know how often the content migration gets underestimated. The frontend gets rebuilt more or less on schedule while the website content follows over the months after, and that split is where the pain sits. Your content teams end up working across two systems at once, with no clear overview and no clean cut-over, and for brands operating in markets around the world that is exactly where a project starts to hurt.

Storyblok's recent Content Debt report puts a price on this: outdated and hard to find content is now a 4.63 trillion dollar liability, and AI is pulling that hidden content into the open. A migration is the natural moment to act, because your content debt has to move too. You can either fix it on the way or drag it across untouched.

Because sooner or later, every platform change arrives at the same central challenge. The old content lives here, the new site needs it there, and somebody has to move all of it, without the website having temporary blind spots, and ideally without a small army copying and pasting for months. The content has to arrive fast, intact and correct. It also, and this is the part people underestimate, never arrives one to one. There is always a transformation of how the content is stored and organised in the new platform that needs to be done on the way.

Think of it as moving from an old house to a new house. The new house is built to modern standards and has a new floor plan, meaning that the rooms for your content are shaped differently. Your furniture, appliances and inventory still have to come with you, but not everything fits where it used to. So, some of it gets rebuilt, some gets thrown out, and a few things you finally get around to fixing. Nobody would try to put every item and box back in exactly the spot it held in the old house. Content migration is the same challenge, at scale, under a deadline.

In the past the honest answer to "can we speed this up" was always "a little, if we reduce the scope or increase the manpower". Then AI came along and changed it. This is the story of a project where I put this to the test, on a real content migration into Storyblok.

The same four boundaries, every time

Strip away the client specifics and almost every content migration from a platform change runs into the same four problems.

Content mapping. The source content model and the target content model are rarely the same shape. A hero in the old world might become a banner in the new one, a teaser might split into two components, and some old modules simply have no home, so their content has to be rehoused somewhere sensible. This is especially underestimated when moving to a headless CMS. Getting that mapping right, module for module, page after page, is most of the work.

Asset handling. Images, documents and downloads have to move into the new system's asset manager or DAM, with every reference rewired so nothing points back at the old site. Miss one and you get broken images or dead links on your new website.

Multilingual content. The moment more than one language is involved, complexity multiplies. Language sites are seldom perfect mirrors of each other. One market has an extra section, another translated only half a component, a third quietly left some English in place. A migration has to decide, section by section, what maps to what and how to deal with localisation.

SEO continuity. The new website inherits the old site's hard-won search equity only if you protect it. URLs change, metadata needs to come across, redirects have to be planned, and alt text and structured signals should ideally improve rather than regress on the move.

None of this is new. What is new is how much of it a machine can now do for us.

Why Storyblok is a destination worth the move

A quick word on the destination CMS Storyblok. After having worked with several different CMS platforms over the past two decades, I must say I really enjoy working in Storyblok. It is a headless, component-based CMS that treats content as structured, reusable blocks rather than fixed pages, which is exactly what makes content migration manageable and keeps the content useful afterwards for multichannel delivery and AI discoverability. The visual editor keeps content editors productive without a developer on standby, the Management API gives engineers a clean way to create and update everything programmatically, and the built-in internationalisation handles multi-market content properly rather than as an afterthought. And it offers a variety of native AI tools and experimental labs alongside third-party integrations for digital optimisation which really empower teams to build flexible, automated workflows directly inside the CMS.

Independent analysts back that up. Storyblok is a Leader in the IDC MarketScape for AI-enabled Content Management Systems, a Strong Performer in the Forrester Wave for Content Management Systems, and G2 ranked it the number one enterprise CMS in EMEA in its Summer 2026 report. MSQ DX is a Storyblok Platinum Partner and certified Storyblok Enterprise Expert, so we know the platform inside out. It is a target CMS of choice for many of our clients, and one we work with every day, not just for a single project.

AI-assisted content migration into Storyblok

The solution we used for this project is our content migration accelerator Nimbus. This article focuses more on the Storyblok integration rather than elaborating the full AI-assisted toolchain. But first, a few basics on how Nimbus works before we get to the Storyblok part.

Nimbus follows a scrape-first approach. This means that rather than integrating it deeply within the source CMS, it visits the live pages the way a customer's browser does and captures the rendered HTML. That sidesteps a whole class of problems, because you do not need credentials, exports, or deep knowledge of the platform you are leaving. You basically just need the pages to be publicly accessible and reasonably consistently marked up.

From there it runs a repeatable pipeline. It discovers the in-scope URLs, scrapes the HTML, and extracts it into structured JSON against a schema, using XPath, CSS or, where the markup fights back, custom or LLM-based extractors. That structured JSON is the pivot point of the whole operation. Once the content is clean, predictable data rather than messy markup, everything downstream becomes programmable.

The Storyblok Connector is the step that turns the finalised JSON into real Storyblok content. Nimbus maps that JSON onto the target Storyblok content model and then loads it through the Storyblok Management API. Stories, fields, metadata and assets go in as first-class Storyblok content, not as a flat dump. Images are downloaded at their highest resolution, re-uploaded into the Storyblok Asset Manager, sorted as needed into localised market folders, and the image fields on each story are rewired to the new internal assets automatically.

Two Storyblok-specific things are worth highlighting, because they are usually where migrations get stuck.

The first is internationalisation. We used English as the master language and layered additional languages on top using Storyblok's internationalisation (i18n), with field-level localisation wherever the content model allowed it. The hard part was not the translation, but the alignment. Around a third of pages had structural differences between languages, a section present in one market and missing in another. So we built an AI agent whose only job was to compare each master section against the localised versions and decide, section by section, what genuinely matched and what did not. Where a master section had no counterpart in a given language, the migration set that section's visibility to hidden for that market rather than inserting wrong content or leaving a gap. That gave us a reliable, automated basis for field localisation and section visibility instead of reviewing hundreds of blocks by hand.

The second is enrichment on the way through. Because the content passes through structured JSON, we can improve it during the transfer rather than after launch. Nimbus used AI vision to generate alt text for images that were missing it, describing what is actually in each picture instead of leaving an empty but required attribute. It generated improved, search-oriented page titles and descriptions and produced a report of every change for review. And it flattened a redundant path segment out of the URL structure to give cleaner, shallower, more SEO-friendly slugs on the new content tree.

What was delivered in Storyblok

We migrated one hundred pages of a large, multi-market enterprise site into Storyblok, with English as the master language and additional fully localised language variants, all within four weeks and ready for client approval and go-live.

Twelve components were mapped to the new Storyblok content model, covering every content type in use on the migrated section. Deprecated modules were dropped and their content rearranged in the nearest sensible new component.

All images and files were moved into the Storyblok Asset Manager, in localised market folders, with references rewired. Missing alt text was generated with AI vision. A redundant URL segment was removed to clean up the tree, and the redirect information was prepared so that, once the redirects are in place, inbound links and search equity carry over rather than break. SEO reports were produced per URL, and improved metadata was generated and added, with logs for a simple review.

But the point is not the raw page count. It is that all of this was delivered by one migration engineer running a repeatable pipeline and iterating quickly, with the machine doing the volume work and the human making the judgment calls. This is not only faster but also demands fewer people and produces far more accurate results than the traditional cycle of copy, paste, check, repeat.

The benefits we experienced

A few key takeaways out of this project that will hold for any potential Storyblok customer facing a similar challenge.

• Scrape-first is a feasible strategy. Not needing deep access to the platform you are leaving removes a dependency that usually consumes the first weeks of a project.

• The structured-JSON base is where the leverage is. Once content is clean data, the transformation, localisation alignment and enrichment stop being manual chores and become manageable configurations.

• The direct and proven connection into Storyblok is the difference between an experimental approach and a highly efficient migration, and that gives us a real advantage. Loading real stories, real i18n and real assets through the Management API is what turns "we scraped the content" into "the content is in Storyblok and ready to publish".

• As a digital impact company, impact only counts for us when you can measure it. Weeks instead of months is measurable. A complete content-type mapping is measurable. Content enhancements such as alt text present instead of absent, metadata improved instead of copied, URLs cleaned instead of inherited, all of it is measurable, all of it delivered on the way in one go rather than as a weary and time-consuming cleanup project afterwards. And this is not only internal efficiency, it is faster time-to-value for the client, which is where the commercial impact of a migration actually sits.

AI does not make content migration disappear. It moves the human effort off the repetitive volume and onto the decisions that actually need judgment, and it makes the whole thing fast enough to stop being that part of the project everyone dreads.

Content migration is here to stay

The thing about content migration is that every few years a better platform arrives, an acquisition forces consolidation, or a brand outgrows the system it is on, and the content has to move again. As long as organisations change systems, somebody will be packing the boxes and loading up the van.

What changes now is how the packing gets done. And the direction of this journey is clear. Less copy and paste by hand, more pipeline orchestration with the enrichment, accessibility fixes and SEO improvements happening automatically in transit, so a new website launches better than the old one rather than merely equal to it. The migration engineer's job shifts from doing the moving to designing and supervising it, and to solving the genuinely hard, site-specific problems no pipeline can guess at, like how multiple language sites that disagree with each other should be reconciled.

Moving from an old house to a new house will always be a bit of an upheaval. But there is a real difference between doing it alone with a stack of cardboard boxes and doing it with a good crew, a plan for every room, and someone who has done it a hundred times before. That is where AI-assisted migration has its place. Same job, far less complexity, and a result you can actually measure.

And this is where the story goes on. A migration is the moment your content finally becomes clean, structured and consistent. But structured content is worth far more than a tidy launch. A follow-up article will look at what migrated content becomes once it lives in Storyblok, and how turning it into structured, semantic content starts converting yesterday's content debt into something valuable that the business, and the AI systems now reading it, can actually use.