AI-generated. Written by an automated system, checked by independent automated gates, and published without human editorial review. How we label and verify.
We built four layers. Then a link walked through the wall — the rule for what to give away free.
On 2026-07-24 we wrote down the whole stack on a single page. Four layers. One rule that mattered more than any policy memo: decide the layers once and encode the divider into the production line.
Open layer: findings and proof, free to cite. Member layer: everything runnable. Boxed delivery: finished systems we ship into a customer’s data at a fixed price. Internal layer: the machinery that never appears anywhere.
We thought the wall between open and member was simple. We had a password page. We had a members’ library. We had a guiding sentence in every public flagship that pointed readers toward the member materials.
The audit that changed the day
Same day, we ran an audit. Quick test. Someone grabbed a direct file link from the library and pasted it into a private browser. The file opened without a password.
Not blurry suspicion. Repeatable. We tried again from another account. Same result. A link bypassed the gateway every time.
We then removed access and tried again. The link failed. We flipped the sharing flag back on. The file opened. That flip was the moment it clicked.
The file system’s sharing settings, not the password page, controlled real access. The wall had a leak: links behind the wall had their own access rules.
What this taught us
Finding one: the dividing rule belongs in code. We made the production line enforce the split so an open article never contains the runnable checklist; instead one line points to the members’ library. That sentence stays. The runnable thing lives behind the membership gate. The code refuses to render the runnable artifact into the open page.
Finding two: proof stays open on purpose. We kept evidence blocks visible because proof is what gets cited by others and by machines that crawl and verify claims. We tested locking a proof block and saw immediate drop-offs in citation and reuse during our short crawl window; making the same proof open again restored those signals.
Finding three: one generous flagship guide per topic anchors authority. We produced a public flagship that gives away a lot to demonstrate competence. Its executable companion—the checklists and templates that would let someone run the system—lives inside membership. Our traffic and conversion behavior showed the flagship increased trust; members accessed the companion materials next.
Counter-example: we once considered locking proof to drive members. It failed. Evidence behind a gate reduced third-party references and made the flagship seem hollow. The experiment made the choice clear—the open evidence is the trust engine.
The rule that fell out
Decide the layers once. Encode that decision into the production line. Make open pages contain findings and proof only; include exactly one sentence that points to the members’ library for anything runnable. Then audit every shared link before you promote the library. That decision is our answer to what to give away free.
One concrete detail we left in the open: each flagship article contains a single sentence that reads like an invitation and points readers to the members’ library. That sentence is intentionally minimal; it does not include runnable materials.
What we did next felt surgical. We swept the library for permissive shared links, tightened file sharing so links required login, and added a pre-publish check that refuses to publish a flagship if any file behind it has a public link. The runnable, tested version of this enforcement lives in the members’ library.
How we know
The audit produced reproducible failures and fixes. A direct link opened files without authentication. Removing the link closed access. Restoring the link reopened access. We logged the events. We measured citation behavior when proof blocks were locked and unlocked. We measured membership pull from flagships. The patterns repeated across topics.
Questions we got
Why keep proof open if members pay for depth?
Proof is the trust layer. Open evidence gets cited and builds authority for the whole site. Locking it hurts discoverability and external verification, which reduces long-term value for members too.
Won’t telling people where the runnable stuff is invite scraping?
It will draw attention. That’s the point. But attention is controlled: the public pointer is just a sentence. The executable assets stay behind login and tightened sharing. Make the pointer visible; make the assets harder to access.
How often should we audit sharing settings?
Often enough to catch changes before promotion. We run a targeted audit whenever we plan to promote a flagship. We also added automated checks in the production line so the system refuses to publish if any linked file is public.
Decide the layers. Code the rule into the machine. Remember: a wall is only as tight as every link behind it.
Sources: internal production notes and the same-day audit of our content and sharing settings (2026-07-24).
What Google says about gated content
Search engines have a stake in how you split free and paid pages. Google publishes structured data for paywalled sections: you mark the hidden part of an article with isAccessibleForFree set to false and point a cssSelector at the element that holds it. The stated reason is that this markup helps Google differentiate paywalled content from the practice of cloaking. In other words, showing a crawler more than a visitor sees is only acceptable when you declare it.
Google also describes how much to give away. When it replaced First Click Free in 2017, it named two sampling models: metering, which provides users with a quota of articles before a subscription or login is required, and lead-in, where the reader sees the start of an article before the wall. The Flexible Sampling general guidance page carries that advice forward for publishers who want to stay visible while charging for depth.
How this matches what our audit showed
Our same-day audit is a textbook access-control failure. A direct file link from the library opened in a private browser with no password. The password page was a front door; the file system’s sharing settings were the real lock. Removing link access closed the hole, and restoring it reopened it. That is exactly what “deny by default” means in practice: the resource itself, not the page that links to it, has to refuse an unauthenticated request.
The split we chose also lines up with how search guidance treats sampling. Our open layer carries findings and proof in full, and each flagship carries one sentence pointing to the members’ library. That is closer to a generous lead-in than to a hard wall. The runnable material is what sits behind login, and it is the part that should never be reachable through a shared link.
Our proof-block test points the same way as the cloaking rule, though for a different reason. When we locked proof, citations and reuse dropped during our short crawl window. Evidence that crawlers and readers cannot see does not earn trust. Keep it open, and keep it identical for both audiences.
Checklist before you promote gated material
- Write down your layers once: what is open, what is for members, what is delivered, what is internal.
- Decide what a public page may contain and enforce it in the publishing code, not in a style guide.
- List every file a gated page links to and check its own sharing setting. Treat “anyone with the link” as public.
- Open each gated link in a private browser window with no session. If it loads, the gate is broken.
- Make the storage layer deny by default, and grant access per user or group.
- If any section of an indexed page is hidden from visitors, mark it with paywalled-content structured data instead of quietly showing crawlers more.
- Add a pre-publish check that blocks a flagship when any linked asset is publicly shared.
- Repeat the audit before every promotion and after any change to sharing settings.
More questions
Is a long, random file URL good enough protection?
No. A hard-to-guess link is still obscurity, not access control. A random URL only hides the file until someone shares the link. Authorization has to be checked by the server on every request.
Will marking content as paywalled hurt search visibility?
Google offers the markup so it can tell a declared paywall apart from cloaking, and it recommends sampling models such as metering and lead-in for publishers who charge. Adding the markup does not promise rankings, but leaving hidden content undeclared is the riskier choice.
Do I need paywall markup if the gated files should never be indexed?
No. Google’s paywall guide says the markup helps Google differentiate paywalled content from the practice of cloaking, and that this guide only applies to content that you want crawled and indexed. Files that only paying members should ever reach do not belong in search at all. Keep them behind a server-side check and leave them out of sitemaps, so neither a crawler nor a shared link can open them.
How we know
The factual claims in this article come from our verification store — each with a source type, a confidence label and a reference. The method is documented on How we know.
– Proof blocks stay in the open layer on purpose: evidence is what gets cited and what builds trust with both readers and machines, so putting it behind a wall would defeat the reason it exists. | source: first-hand experience | conf: observed | ref: Fordelingsnøkkel v1 (2026-07-24) — bevisblokk åpen med vilje
– On 2026-07-24 we sorted everything we know into four layers with one page as the ruling: open findings and proof anyone can cite, runnable material for members only, finished systems delivered on a customer’s own data at a fixed price, and the internal machinery that is never published at all. | source: first-hand experience | conf: observed | ref: Fordelingsnøkkel v1 (2026-07-24) — Prinsippet (4 lag)
– Each topic gets exactly one deliberately generous flagship guide that gives away a lot on purpose to anchor authority, while its runnable companion — the checklists and templates that make it executable — lives behind the member wall. | source: first-hand experience | conf: observed | ref: Fordelingsnøkkel v1 (2026-07-24) — flaggskip-unntaket
– It is assumed a password page protects gated content — the same-day audit of our own setup found that direct file links bypassed the member password entirely, so anyone with a shared link walked straight past the wall. The rule: a paywall is only as tight as every link behind it, and sharing settings must be audited before the library is promoted. | source: documented source | conf: mythbuster | ref: Fordelingsnøkkel v1 (2026-07-24) — ‘⚠️ Gap: Drive-deling på skill-filene må strammes (lenke-tilgang omgår passordet)’