Archive the Terms You Agreed To, Before They Quietly Change

TL;DR

Terms of service and privacy policies change constantly, and the version you agreed to usually vanishes. Most companies publish only the current text. Some list an "effective date" and nothing else. A minority keep a real version history, and even those often quietly edit within a version.

This matters more in 2026 than it did two years ago. Twenty states now have comprehensive privacy laws in effect, California's Delete Request and Opt-Out Platform (DROP) went live for data brokers on August 1, 2026, and companies have been rewriting privacy disclosures at pace to keep up. Every one of those rewrites replaces the text somebody agreed to.

The fix takes about fifteen seconds: convert the page to PDF the moment you accept anything that matters, and keep it. Convert: Web to PDF does it in one click, entirely on your machine, producing a real PDF with selectable text and working links — so the archived document is searchable and its cross-references still resolve.

One honest caveat up front, stated properly below: a self-made PDF is documentation, not proof. It's excellent for your own records and for reconstructing a timeline. It is not a notarised instrument, and this post is not legal advice.

Why the old version disappears

There's rarely anything sinister here. It's how content management works.

A legal page is a page. Editing it replaces it. Unless someone deliberately built a versioning system — and legal pages are usually the last thing to get one — the previous text is simply overwritten. The Wayback Machine may have caught it, or may not: coverage of legal pages is patchy, and pages behind a login are never captured at all.

Three patterns you'll see in the wild:

Effective-date-only. The page says "Last updated: 14 March 2026" and shows current text. What changed, and what it said before, is unrecoverable from the page itself.

Real version history. Some companies — more common in developer tools and infrastructure than in consumer apps — publish dated archives of every version. Genuinely good practice, and worth noticing when a vendor does it.

Silent edit. The text changes without the effective date moving. Typo fixes, mostly. Occasionally not.

Where it gets awkward

The scenarios where people wish they had the old text are pretty consistent:

• A dispute about what a service promised. Uptime, data handling, refund conditions, ownership of your content.
• A pricing or plan change where the earlier terms described something different.
• A vendor security review, where you need to show what a supplier's DPA said at the time you onboarded them.
• A data rights request — you exercised a deletion or opt-out right under the policy as it stood, and the policy has since been rewritten.
• An AI training clause appearing in a product you've been using for years, where the question of what you agreed to and when is the whole question.

That last one has driven a lot of archiving behaviour over the past two years, and it's the clearest example of why a snapshot beats a memory.

What's worth archiving

Not everything. A rule that produces 400 PDFs a year produces nothing you'll ever look at.

Archive when you accept, pay, or entrust:

Don't bother with: free sites you're browsing, one-off purchases with no account, anything where you'd never realistically contest the outcome.

The sub-processor list is the underrated one

For anyone doing vendor due diligence: the sub-processor list is often a separate page, updated more often than the main policy, and it's the page that actually tells you where data goes. It's also the one that changes without announcement most often.

Doing it well

Capture the whole page, not the visible part

Legal pages are long and often use accordions or tabbed sections. Before converting:

• Expand every collapsed section. A collapsed accordion is genuinely absent from the capture — you'll archive the headings and none of the content.
• Scroll to the bottom if the page loads progressively.
• Check for "see also" links to separate documents. A privacy policy that incorporates a cookie policy by reference means you need both.

Get the URL and date into the document

This is the difference between a useful archive and a folder of anonymous PDFs. Chrome's print engine can stamp headers and footers with the page URL and the capture date. Turn it on for legal archiving specifically — a PDF showing the source URL and the date it was captured, printed by the engine rather than typed by you, is far more useful eighteen months later.

The customization section of the FAQ covers enabling headers and footers.

Keep the links live

A real PDF preserves clickable links. That matters here more than usual, because legal documents are a web of cross-references — "as described in Section 8", "see our Cookie Policy", "our sub-processors are listed here". A screenshot-based capture turns all of those into dead pixels.

Don't strip too much

Remove Elements is useful for cookie banners and chat widgets that would otherwise cover text. Resist the urge to tidy further. For an archival document, more context is better than a cleaner page — the nav, the footer, the effective-date line, and the surrounding page furniture all help establish what this document was.

For the same reason, be cautious with Article Mode here. It produces a beautiful clean read and it discards the page's structure, including sometimes the effective-date line sitting in a sidebar. For legal archiving, default mode is the safer choice.

Name the file so you'll find it

vendor-name-privacy-policy-2026-08-04.pdf beats Privacy Policy.pdf by a distance when you have thirty of them. Date-suffixed filenames sort chronologically, which means you can see at a glance when a policy last changed.

Pages behind a login

This is where local conversion stops being a preference and becomes the only option.

A lot of the terms that actually bind you aren't public:

• Enterprise agreements shown only inside your admin console
• The specific plan terms attached to your account
• In-app consent screens and permission grants
• Data processing addendums accessible only when signed in
• Beta programme terms behind a feature flag

URL-based converters cannot reach any of these. PDFCrowd, Webtopdf-style services, and anything where you paste a link fetch the URL from their own servers with no session. They get a login page. That's not a limitation you can configure around; it's what those tools are.

Hosted converters that accept an upload have the opposite problem: you'd be sending your account's contractual documents to a third party to process. For a public marketing page, whatever. For your company's negotiated enterprise agreement, that's a decision someone should be making deliberately.

Convert: Web to PDF converts the page already rendered in your tab, locally. Nothing is uploaded, the page never leaves your machine, and your session isn't involved in any network request. That's the only reason converting an admin console page is a sensible thing to do.

What a self-made PDF is and isn't

Being straight about this is more useful than overselling it.

What it is:

• Contemporaneous documentation of what a page said when you looked at it
• Searchable, quotable text you can point at in an email thread
• A reliable memory aid for reconstructing a timeline
• Perfectly adequate for internal records, vendor reviews, and most commercial disagreements

What it isn't:

• Cryptographically timestamped
• Notarised or witnessed
• Tamper-evident — a PDF you made is a PDF you could have edited
• Proof of anything on its own, in a formal dispute

If the stakes are genuinely legal, the tools that exist for that purpose are different ones: a qualified timestamping authority (RFC 3161), a notary, a litigation-hold service, or your counsel's own preservation process. Convert: Web to PDF makes a document; it does not make an attestation, and we'd rather say so than let you find out at a bad moment.

Where the self-made PDF genuinely shines is the enormous middle ground: the vendor review, the "wait, didn't this used to say something else?" moment, the support conversation where you can paste the exact sentence, the compliance file where you need to show you read the disclosure. That's most situations.

A small routine that works

1. A folder per vendor. ~/Documents/Terms/acme-corp/
2. Capture at the moment of consent, not later. Later is when the text has already changed.
3. Date-suffix everything.
4. Headers and footers on, so URL and date are in the document.
5. Re-capture when you get a "we're updating our terms" email — capture the new version and keep the old one. The pair is what's useful; either alone isn't.
6. Twice a year, re-capture the policies for the five vendors that hold the most of your data.

Step 5 is the one that carries the value. A single policy tells you the current state. Two dated versions tell you what changed, which is almost always the actual question.

Frequently asked questions

Why can't I just use the Wayback Machine?

Sometimes you can, and it's worth checking first. But coverage of legal pages is inconsistent — some are captured monthly, others never — and pages behind a login are never archived at all. It also won't have the version specific to your account or plan. Use it as a supplement, not a plan.

Is a PDF I made myself legally valid evidence?

Treat it as documentation rather than proof. It's contemporaneous, searchable, and genuinely useful for records, vendor reviews, and commercial disputes. It is not notarised or cryptographically timestamped, and you made it yourself, so it isn't tamper-evident. For matters where formal evidentiary weight matters, use a qualified timestamping service or your counsel's preservation process. This isn't legal advice.

How do I capture a terms page that uses collapsible sections?

Expand every section before converting. A collapsed accordion isn't rendered, so it won't appear in the PDF — you'd archive the headings with none of the text underneath. Expand all, scroll to the bottom, then convert.

Can I archive terms shown only inside my account?

Yes, and this is where a local browser extension is the only workable option. Convert: Web to PDF converts the page rendered in your tab, so anything you can see while signed in converts normally. URL-based converters fetch the page from their own servers without your session and get a login screen instead.

Should I use Article Mode for legal pages?

Generally no. Article Mode produces a clean single-column read by discarding page structure, and on a legal page that structure — the effective-date line, the navigation between related documents, the footer — is part of what makes the archive useful. Use default mode and Remove Elements for the cookie banner if it's in the way.

How do I get the URL and capture date into the PDF?

Enable Chrome's print headers and footers in the extension's settings. The engine stamps the source URL and the date onto each page. For archival captures this is worth doing every time — it's what distinguishes a useful record from an unlabelled PDF.

Does converting a terms page send it anywhere?

No. Conversion happens entirely in your browser. Nothing is uploaded and no copy is stored anywhere but your own machine — which is what makes it appropriate for account-specific contractual documents.

Bottom line

The version of the terms you agreed to is usually the version that no longer exists. Companies don't hide it; content management just overwrites it, and legal pages are the last thing to get proper version control.

Fifteen seconds at the moment of consent solves it. Expand the sections, turn on headers and footers, convert, date the filename. Do it again when they email you about an update — because the useful artefact isn't a policy, it's a pair of policies with a diff between them.

Be realistic about what you've made: a good record, not a notarised instrument. For almost everything short of litigation, a good record is what the situation calls for.

Convert: Web to PDF is free, watermark-free, and never uploads the page. The full FAQ covers the privacy and security details.

Unrelated but also ours: CineMan AI, which picks something to watch so you don't spend the evening reading catalogue descriptions.