Squarespace won't export the pages you can see

Before quoting a migration off Squarespace, I wanted to know how much content was actually there. Not “how many pages does the nav show” — how many words, how many images, how many of them are real.

You can get most of that without an account. Squarespace answers ?format=json-pretty on any URL and hands back the page’s own data. Append it to a collection URL and you get every item in the collection:

curl -s "https://example.com/events?format=json-pretty" > events.json

For an events collection that returns upcoming and past arrays, each item carrying full body HTML, startDate/endDate as epoch milliseconds, urlId, tags, and assetUrl for the cover image. Everything a migration needs. ?format=rss and ?format=ical work too, which makes a decent cross-check on dates.

So I wrote the export script, ran it against the static pages, and got this:

"mainContent": "<div class=\"sqs-layout\"></div>"

Empty. On every single page. Pages that render several screens of text in a browser.

Fluid Engine doesn’t serialize

Squarespace’s newer page editor — Fluid Engine, the drag-on-a-grid one — doesn’t put its content where the older layout engine did. The JSON endpoint still responds 200 with a well-formed document, still has a mainContent key, and that key is a stub.

The failure is quiet in the worst way. You don’t get a 404 or an error field. You get a valid-looking response with nothing in it, which is exactly what a page with no content would also return. If your script logs “exported 8 pages” you will believe it.

The built-in Settings → Import/Export WordPress XML has the same gap, and it skips Events entirely, so it isn’t the fallback either.

Check one page by hand before you build anything. Fetch the JSON, look at the actual bytes of mainContent, and compare against what the page renders. Length is enough: 143 bytes is a stub.

In my case it didn’t matter much. Stripped of markup, all eight pages together came to under 11 000 characters — the longest was 913. Rewriting them by hand in Markdown was faster than any scraper would have been, and produced better copy than a faithful port of three-year-old placeholder text. But that was luck. A content-heavy site would have turned a scripted afternoon into a manual week, discovered halfway through.

Two more things worth checking from the outside

Empty event addresses still carry coordinates. Eleven of the twelve events I pulled had a blank address string and map coordinates of 40.7207559, -74.0007613. That’s Squarespace’s default drop pin, in Lower Manhattan. Every one of those events happened in Sweden.

Nothing about the data says “unset”. mapLat and mapLng are populated floats sitting next to an empty addressTitle, and a naive importer will happily emit Event JSON-LD placing a Stockholm meetup in New York. Treat an empty address as the signal, and drop the coordinates when it’s blank.

Template stock photos hide in plain sight. Squarespace image URLs are shaped https://images.squarespace-cdn.com/content/v1/<site-id>/..., and that <site-id> is the site the asset belongs to — not necessarily the site serving the page. Get the id from the JSON (website.id), then bucket every image URL by it:

grep -oE 'images\.squarespace-cdn\.com/content/v1/[a-f0-9]+' page.html \
  | sort | uniq -c | sort -rn

Seven of the forty-three images on this site came from two other site ids. Those are the demo photographs that ship with the template — the ones with names like AroHa_02.jpg and 20140301_Trade-151_0124-copy.jpg. They had been sitting on the live About page for three years.

That turned out to be the most useful number of the whole exercise. It isn’t really a technical finding; it’s a measurement of how finished the site is, and it takes one grep. Combined with an empty “Upcoming Events” block and a job posting whose start date had passed, it reframed the question from how do we move this to is anyone actually running it — which is a much better question to ask before writing a line of migration code.

The short version

Squarespace will tell you a lot about a site you don’t have access to. Just don’t trust it to tell you when it’s told you nothing.