An archive you cannot take with you is a subscription

Whether you own your blog shows up on the day you stop paying for it. If what comes back is a zip file of loose text, you were never building an archive. You were renting one, and the rent included the right to lose it. An engine that makes leaving expensive is selling lock-in, and a year of posts is only an asset if it survives the end of the contract intact.

The text is the easy part, and the part vendors point at

Every vendor will tell you the content is yours, and most mean it. The words are the cheapest thing to hand over. The European Commission's explainer on the Data Act, a regulation that has applied since 12 September 2025, says that providers of platform and software services must at a minimum export data in a commonly used, machine-readable format. Plain text in a folder clears that bar. It is also nearly useless on its own.

A post is more than its body copy. It has a publication date, a slug, a title that differs from its SEO title, a meta description, the links it makes to other posts, and a second language attached to the same document. An export that drops the dates gives you a pile of undated essays. One that drops the slugs breaks every inbound link on the day you import it. We argued in an earlier post that the archive is the asset, and an archive is exactly the structure that a careless export strips out.

So the question is not whether you can export your content. Ask whether an export can be re-imported somewhere else and come back as the same site.

Images are where the rights question hides

Generated illustrations raise a question that text mostly does not. The US Copyright Office concluded in its January 2025 report on copyrightability that copyright does not extend to purely AI-generated material, and that prompts alone do not make someone its author. That is a separate matter from ranking, where Google does not care who or what wrote the page. It does mean that your right to keep using a machine-made hero image after you leave rests largely on the contract, because copyright may not give you much to stand on.

Then there is the plainer problem of where the file actually lives. A hero image served from the vendor's storage is a broken image the morning the account closes. We found out how quietly that fails on our own blog. For five days most of our posts pointed at hero files that no longer existed, while the database insisted everything was fine. Nobody noticed until a check went looking. Images have to live in storage you control, or come with the export as files rather than as links.

The brief is the part nobody thinks to ask for

The most valuable thing a year of automated publishing produces is not any single post. It is the working description of your company that the engine writes from. That means what you sell and to whom, the claims you have evidence for, the ones you refuse to make, and the voice rules someone argued over in month two. That knowledge took effort to assemble, and it is the part that makes the next vendor, or the next hire, productive in a week rather than a quarter.

If that description exists only as settings inside the vendor's system, leaving means onboarding from zero. It should exist as a document a person can read and hand to someone else.

What legitimately stays behind

We are not arguing that a vendor owes you everything. The machinery is theirs: the pipelines, the prompts, the checks and the scheduling are what you paid for, and they stay behind when you leave. The rhythm stops too. An archive survives a vendor change and a publishing cadence does not, so the month after you leave will be quiet whoever you move to.

The law is also less settled than a confident vendor or an anxious buyer would like. The Data Act's switching rules were written with cloud services in mind, and how far they reach into a content engine is a question for a lawyer rather than a blog post. Treat them as a description of what a reasonable exit looks like, not as a guarantee you can enforce next Tuesday.

The line is simple enough to apply without one. Anything that describes how the vendor works stays with the vendor. Anything that describes you, or was published under your name, leaves with you.

Ask the exit question on the first call

The time to find out what leaving costs is before you sign, while the vendor still wants your business. Ask four things. In what format does the archive come out, and does it keep dates, slugs and both languages? Where do the images live, and what does the licence say after cancellation? Can you have the brief as a document? And who controls the address, which we covered in a separate post?

At Affinero, posts are published directly into your own site, so the text is in your system from the first morning rather than ours. A vendor that is confident in what it produces should make leaving cheap, because the archive it leaves behind is its best reference. One that makes leaving expensive is telling you what it expects you to want to do.