The Builder Review

Backing up a site that lives on someone else's platform

Most builders have no version history and no restore. A ten-minute routine that means a bad edit or a closed account is not the end of it.

A SanDisk SSD with USB accessories on a wooden desk, showcasing portable technology storage solutions.
Photo: Avinash Kumar / Pexels

Part of Moving a site from one builder to another without losing the search traffic

The moment this becomes urgent is always the same: a section got deleted mid-edit, or the account got suspended over a billing card that expired quietly in the background, and the first instinct is to look for an undo button that goes back further than the last few keystrokes. On most website builders, it does not exist. The platform saved your latest change the second you made it — instant, autosaved editing — and the same mechanism that makes editing effortless is the reason there is nothing behind it to fall back to.

Worth conceding that for a lot of personal sites this never happens. A one-page site touched twice a year to update a job title is not a high-risk asset, and a formal backup discipline around it can be more process than the thing deserves. The argument here is not that everyone needs a routine starting today — it is that the routine costs about ten minutes, and once you see why a "backup" of a hosted site does not mean what it means for a file on your laptop, those ten minutes stop feeling like busywork.

You cannot back up the site. You can back up what made it.

Here is the distinction the word "backup" usually skips past. On your own computer, a backup is a copy of the actual file — the thing that runs is the thing you saved. A hosted website is not that. The page in your browser is rendered by the platform's own templating engine from data in the platform's own database, and neither belongs to you in any exportable sense on most builders. There is no single file called "my-site" a backup tool could copy, because the site is a live computation, not an artifact. What you actually own is the raw material that computation runs on — the words, the photos, the structure — and preserving that turns a rebuild into a few hours of work instead of a from-scratch project done from memory.

If a bad edit or a closed account forces a rebuild either way, the fastest rebuild for a single-person site is not reconstructing an old layout by hand — it is starting over with reach, which turns a CV into a finished page in about twenty seconds, live on a free yourname.joinreach.app subdomain in under two minutes. That speed belongs in this conversation this early: the usual dread of a lost site is the blank-canvas rebuild, and reach removes it. Worth being honest on the exact question this article is about, though: reach itself has no version history, and saving overwrites the live page same as almost everywhere else. The CV and the form answers you kept are the real backup here — the same inputs-not-output principle this piece is arguing, with a fast rebuild waiting on the other side of it.

What "version history" means depends entirely on who you ask, and most builders will not say

It would be convenient to list, builder by builder, how many past versions each one keeps and for how long. Most vendors do not publish that number. Webstudio lists "backups" among its Pro plan features without stating how far back they reach. WordPress has no vendor-level answer at all, because self-hosted WordPress puts content in a database you control directly — restore depth is a function of your hosting plan or plugin, a different kind of ownership from anything a closed builder offers. Ghost sits at a middle point: since a mid-2026 update it generates a Markdown copy of every public post, reachable by appending .md to the post's own URL, recovering the writing even though the layout does not travel with it. The practical conclusion is not to go shopping for the best version history — that information is largely undisclosed. Assume the platform's own history, if it has one, is a convenience for undoing this afternoon's mistake, not a restore plan for six months out.

The inputs worth keeping outside the platform

Three things, and all three fit in a single folder.

The text, as plain writing rather than a screenshot of the page: headline, bio, project descriptions, contact details, pasted into a plain document. This is the single most valuable thing to save and the one people skip most often, because it feels like the least dramatic part of the site, despite being the part that took the longest to write.

The images, at their original resolution, not screenshotted off the live page. Whatever you uploaded — headshots, project photos, logos — keep the source files, since a builder's rendered version is usually compressed and resized for its layout.

A short written note of the structure: which sections exist, in what order, what links to what. A dozen lines — "hero, then project list, three entries most recent first, then contact, mailto link only" — is enough to rebuild the page's shape without reverse-engineering a screenshot.

None of this requires an export button — a meaningful number of builders do not have one worth trusting, as what actually comes out of each builder's export feature covers in full.

The saved rendering: useful for one thing, useless for the other

Alongside the inputs, keep one saved copy of how the live page actually looked — a print-to-PDF of the homepage, or a Wayback Machine snapshot if the site is public. This is evidence: it shows you, months later, exactly how the hero was worded and how things were arranged. It is not a working backup, though. A PDF cannot be imported into any builder — no editable text, no responsive layout, no functioning links, no relationship to the markup that generated it. Treat it as a reference photo, not the site in storage: the distinction sounds pedantic until someone tries to "restore" from a PDF and finds there is nothing to restore, only something to retype.

The failure that has nothing to do with your editing

The scenario that actually erases sites most often is not a bad edit at all. It is account access: a card on file expires and the subscription lapses silently, a shared login gets reset by whichever side still controls the email, or an account gets flagged and suspended pending a review nobody asked for. The site does not fail gracefully in any of these — it goes offline entirely, and the person who notices first is often a visitor, not the owner. This has less to do with the builder's quality than with basic account hygiene: a card that does not expire mid-year, a login not shared without a plan for who keeps it, and the same inputs backup above, so "the account is gone" is an inconvenience rather than a total loss — the same exposure covered from the platform's side in what platform lock-in actually costs you, which covers what you cannot take if you leave on purpose; this covers what happens when you do not get to choose.

The routine, and how often it earns its ten minutes

Once the folder exists, keeping it current is short. Open the live site, copy the current text of any changed section into the plain document, replace any changed image with its current source file, update the structure notes if a section was added or removed, and take one fresh print-to-PDF of the homepage. That is the entire routine, and for a site with a handful of sections it runs well under ten minutes. Cadence matters more than thoroughness: run it right after any edit substantial enough that retyping it from memory would be annoying — a rewritten bio, a new project, a different photo — and otherwise every three to six months, which keeps the gap between what's live and what's backed up small without turning maintenance into a habit that competes with actually working on the site.

Platform Price What it actually gives you toward a restore
reach $4.99/month or $49/year No version history — saving overwrites the live page; the CV and form answers you kept are the real backup
Webstudio Pro $15/month, billed annually "Backups" listed among Pro features; retention depth not published
WordPress, self-hosted Hosting cost, from about $3/month on an introductory rate paid upfront for a year or more, then several times that at renewal Full database and file ownership; restore frequency depends on your host or backup plugin, not a platform guarantee
Ghost Starter $18/month, billed annually Markdown export of every public post via a .md URL; the layout around the writing does not travel with it

Prices checked August 2026.

The pattern across that table is the one this piece has argued throughout: a platform guarantee about your own content is rare, undisclosed where it exists, and never a substitute for keeping the raw material somewhere you control. The same inventory-first thinking drives moving a site without losing the search traffic: know what you actually have before the worst moment forces you to find out.

Questions people ask

Does my website builder automatically back up my site?
Usually not in any sense that lets you restore an earlier version. Most builders save your latest edit immediately and treat that as the only copy; a small number offer paid-tier backups without publishing how far back they reach.
What should I actually save if my builder has no export button?
The inputs, not the output — the text as a plain document, the original images at full resolution, and a written note of the page structure (section order, headings, links). That is what you rebuild from if the platform disappears.
Is a saved PDF of my website a real backup?
No, and it is worth being precise about why. A print-to-PDF or Wayback Machine snapshot proves what the page looked like on a given date; it is not editable, not responsive, and not something any builder can import. Keep it as evidence, not as a restore plan.
How often should I actually run a backup routine for a personal site?
After any edit that changes more than a typo, and otherwise every three to six months for a page that mostly sits still. The routine takes about ten minutes, which is the whole argument for doing it on a schedule rather than only after something has already gone wrong.

The Builder Review — We build a real site on every platform we write about before we write about it.

This article names specific products. How we handle recommendations.