The hidden cost of "free online" — ads, signups, file harvesting

When a website processes a file for you on its servers, somebody is paying for the compute, the storage, the bandwidth, the engineers who keep it running, and the cost of any liability that attaches to the operation. If the user isn’t paying, it is worth asking who is.

The answer, almost always, is the user. The transaction just looks different from a credit-card transaction.

There are four currencies “free online” tools tend to accept. None of them is illegitimate. All of them are real money flowing from the user to the operator, and the user usually isn’t told what they’re handing over.

The first is your email. The operator asks for it before the second operation, before the result is delivered, before the file can be saved — wherever the friction is highest. Once collected, an email is a marketing channel, a join key against any other database that has it, an asset that can be sold or licensed depending on the policy, and a nudge into upgrading. Nothing about that is hidden in the strict sense. The policy says it; the user can read it. The practical effect is that “free” really means “I’ll trade you indefinite re-engagement contact for one PDF operation.”

The second is your attention. Ads on the page, video pre-rolls, interstitials between operations. The ad network pays for impressions and clicks; the user pays in time and in the trackers most ad networks plant. Loft will eventually run ads on listing pages — planned, not yet shipped as of 2026-05, and called out in the pillar when they do. They sit on browsing surfaces, not on tool surfaces. That’s the line we draw; other operators draw it differently.

The third is your file’s metadata. Even when the file itself is deleted within an hour, the operator usually retains the surrounding information much longer: file name, file size, the operation performed, your IP at the time, the time stamp. Aggregated across millions of users this is real data — what kind of documents people work on, which keywords keep appearing in filenames, which tools need investment. No operator I know openly sells this. Several use it internally for product analytics. It is real, it exists, and it outlives the file.

The fourth is the file itself. This is the rare case and the one nobody talks about until a story breaks. Some operators retain processed files longer than the policy says. Some train on them. Some hand them to third parties under conditions the user wouldn’t guess from the homepage. The reader doesn’t usually find out until a breach, a regulator action, or an investigative reporter documents it. By that point the file has been wherever it’s been.

This is the failure mode browser-side architecture was designed around. If the operator doesn’t have your file, they can’t sell it, train on it, lose it, or be subpoenaed for it. The architecture removes the opportunity rather than asking the user to trust the operator’s restraint.

I should be clear about Loft’s own ledger. The site costs us money to operate: R2 storage and bandwidth for code chunks, edge infrastructure for analytics and utilities, domain and certificates, Stripe fees on sponsor checkout, and the time I spend writing and shipping. The site is funded today by sponsor donations, with planned ad inventory on listing pages to come. The total is below break-even. It may stay below break-even. That’s fine; “free” doesn’t mean “obligation-free for the operator”, it means “not paid by the user in the four currencies above.”

What I hope this post does is shift the question. The next time you see “free online” on a tool’s marketing copy, the useful thing to ask is not “is this safe” but “what is the operator getting out of this, and how.” Cloud-side tools have real costs; that cost is being paid somehow, by somebody. If the tool doesn’t take a subscription, doesn’t show ads, doesn’t ask for an email, and provably doesn’t retain the file, that’s a sponsorship model or an adjacent revenue model funding the operation — which is rarer than the marketing suggests, and worth checking when you see it claimed.

That check is the thirty-second DevTools verification. Open Network tab, filter on POST, see if your file size shows up. Watch for ads. Notice the email gate or its absence. Skim the privacy policy section on data collection. Five minutes, total. The work pays for itself the first time you avoid a tool that wanted a currency you didn’t realise you were handing over.

I went looking for an honest “free PDF tool” with no ads and no email gate a few years ago and gave up; that’s part of why I ended up building this one. The reader doesn’t need to do anything that dramatic. Just notice which currency you’re trading the next time you click an upload button.

---

Pillar: /docs/how-it-works/ — section 9 covers Loft’s own ledger in full.