Documents stay local.
The desktop app processes PDFs on your computer.
Recovery stays with you.
Working copies and preferences are local files you manage.
Website visits are different.
Hosting providers receive the requests needed to serve these pages.
What this notice covers
This notice covers the public website, the Windows desktop app, and the optional connection-sharing feature (off by default). Two things are still being resolved and this notice will be updated when they are: how the older pre-migration database copies described under Optional connection sharing are handled, and the applicable hosting and email transfer arrangements. See Operator and contact.
1. About this notice.
This notice covers the Pagekind Windows desktop release and this Pagekind website. In this notice, “Pagekind”, “we” and “our” refer to the business operating the Pagekind product. Business and contact details are listed below. The website operator is responsible for personal information processed through the site; hosting does not transfer that responsibility to the hosting provider.
The desktop app and website have different roles. The app edits documents on your computer. This website explains the product and distributes the installer; it does not open, edit or accept uploads of your PDFs.
2. Your documents and local data.
Pagekind processes PDFs locally. The current desktop release has no Pagekind account system, document-upload service, advertising SDK, automatic crash-reporting service or enabled automatic updater.
Add text and Annotate keep unplaced words and settings in memory while their PDF stays open. They are separate from PDF content, saved copies and recovery drafts. Closing asks before discarding pending values, and they do not survive app exit. Find also keeps its current query in memory for an explicit refresh. Read how to place and save new text.
Editing, Undo and recovery require local working files. The following information can remain on your computer:
| Information | Why it is used | What to know |
|---|---|---|
| PDFs, edits and history | Display, edit, undo and recover your work. | Working copies can contain the document and previous edits. Interrupted work may remain for Recover. |
| Preferences, reading places and saved groups | Remember favorite tools and PDF references you explicitly save. | Groups can include document paths, names, order, hashes and page counts. Reading places also record file size, page, zoom, position, view settings and update time. Remove or Clear removes those references, not your PDFs. Preferences survive uninstall. |
| Comments, imports and reports | Keep drafts, review feedback and record batch results. | Some companion records contain plain-text words, author labels, filenames or paths. A PDF password does not protect every related file. |
| OCR review records | Let you review recognized words and resume a kept review. | Current word-review records use Windows-account protection. Older preview OCR transcripts may still exist in plain-text receipts. |
| PDF passwords | Open a protected document or apply authorized changes. | Passwords remain in application memory for the session and are not written into the workspace state records. You may need to enter them again after restarting. |
We cannot retrieve documents held only on your computer through the Pagekind website. Local-only processing does not prevent another program, a shared Windows account, an administrator or backup software from accessing local files.
2b. Basic setup and app reports.
Pagekind sends basic setup and app-step reports to us to find installation problems. No PDFs or filenames. For a fresh interactive installation, reporting starts enabled after a notice in the existing Setup flow. You can turn it off in Setup or under About in the app. There is no extra first-open decision popup or manual report form. Reporting does not affect PDF tools or enable premium.
Existing disabled, declined or withdrawn choices stay off. An unreadable or ambiguous old choice also stays off. Silent installation does not enable reporting. The app starts automatic reporting only when a recent protected completion record belongs to the same Windows account, identifies a fresh installation with reporting enabled, and no earlier disabled or ambiguous app state exists. Other profiles stay off unless reporting is explicitly enabled in About. A website cookie choice does not authorize native reports.
While enabled, Setup reports its measured start after the notice and successful completion. The app reports its first measured opening; the premium offer and full terms actually being displayed; explicit premium acceptance saved on the profile or explicit decline; an observed enable failure; and locally observed enrollment and a ready state. Each stage is stored at most once for a reported journey and release while its record is retained. Reports for different releases remain distinguishable; updating the app does not prove a new person or installation. Closing a screen is not a decline. Premium still needs its own explicit consent and is off by default.
Reports contain a new random journey ID, release and measurement-policy versions, a fixed shared download-source label and the event name. An enable failure may include a coarse category: not permitted, unavailable, failed or unknown. No PDF content, filenames, paths, Windows account IDs, connection keys, ad click IDs, IP address, browser details, precise event time or raw error text is included in the event body.
The random ID lets us match retained stages for that measured journey; these reports are pseudonymous, not anonymous. Setup and app share it only through the protected same-account receipt. The Windows account identifier stays on the computer. Shared Reddit, Google, Microsoft and direct/unknown package labels identify download routes, not unique people or verified paid-ad attribution. Copying an installer keeps its label.
Turning reporting off in Setup affects that setup and prevents it from enabling app reporting automatically for a fresh profile. It does not change a previously saved app preference; use About to turn that profile's app reporting off. Turning app reporting off stops future app reports and drops unsent app reports. It does not erase records already received. The random ID and preferences can remain locally across reinstalls. Enabling reporting later does not reconstruct earlier unmeasured actions. Earlier optional measurement records remain separate and are not relabelled as these basic reports.
Reports go to getpagekind.com/funnel-v3-events. The server replaces the submitted ID with a secret-keyed deduplication value and stores the fixed source/release labels, event, UTC receipt date and any allowed failure category. It does not store the submitted ID or request IP/browser headers in these records. Cloudflare receives ordinary HTTPS requests, including IP addresses, to deliver and protect the service. Native reports are not sent to Google, Microsoft or Reddit. Website ad measurement remains a separate optional choice in Cookie settings.
These events and the separate timestamp-only source-labelled installer-link requests are scheduled to expire after 30 days; provider deletion may follow asynchronously. New collection ends at 00:00 UTC on 7 November 2026. Aggregate reports may remain longer. New request records also identify the routed release; older request records without a version remain unversioned. Request records have no join to app journeys and include repeats and crawlers. The reports count retained client observations and same-journey co-occurrences, not event order, complete installation coverage or current active peers. Offline use, opting out, delays, blocked requests, resets, separate profiles, copies, expiry and fabricated events affect counts. Missing events cannot reveal browser, SmartScreen or UAC warning abandonment or why someone stopped.
Earlier 1.1.6 campaign measurement.
This section describes the separate October 2026 release 1.1.6 Reddit installers. It does not authorize the 1.1.7, 1.1.8, 1.1.9, 1.1.10, 1.1.11, 1.1.12, 1.1.13, 1.1.14, 1.1.15, 1.1.16, 1.1.17, 1.1.18 and 1.1.19 basic reports above. Downloads through our October 2026 Reddit campaign offer Basic campaign measurement, selected in Setup. You can clear that checkbox before installing, or turn it off in the app under About. Pagekind also shows a notice with a Turn off button on the first measured launch. This choice does not enable connection sharing and does not affect your PDF tools.
When enabled, Pagekind sends a randomly generated measurement ID, the shared campaign and ad label, app version and event name to this website. It reports each of three milestones at most once per Windows profile: the first measured app launch; a connection-sharing consent acceptance saved on that profile; and a successful enrollment observed locally after that acceptance. Accepting the notice and successful enrollment are separate events. We do not send PDF content, filenames, names, browsing history, connection keys, Reddit click IDs or your precise event time. The app does not report opening or declining a consent screen.
The app keeps its random ID, preference and delivery state in your local Pagekind preferences. These survive uninstall, so reinstalling does not create a new measurement identity. Turning measurement off drops unsent events and stops future reports; it does not erase reports already received. Normal downloads, silent installs and upgrades from an older unmarked installation start with measurement disabled. Copying a campaign installer can carry its shared label to another person, and separate Windows profiles can count separately.
Our server replaces the random ID with a secret-keyed deduplication value and stores only that value with the fixed campaign/ad/version, event name and UTC receipt date. It does not store the submitted random ID, IP address or browser details in these measurement records. Cloudflare necessarily receives your IP address when delivering the HTTPS request and may process it under its own service and security practices. No measurement events are sent to Reddit.
These are client-reported milestones, not proof of every installation, a count of unique people, or proof that sharing is currently running. Opt-outs, offline use, blocked requests, copied packages or fabricated events can affect the totals. Collection stops on 5 November 2026. These event records are scheduled to expire after 30 days; deletion by the storage provider may follow asynchronously. Aggregate campaign reports may be retained longer to assess this test. Other website request records have the separate retention described below.
2a. Optional connection sharing.
Off by default. Most PDF tools never use it.
Connection sharing is a separate feature you can choose to turn on. Editing, combining, reading and saving PDFs never use it. Premium tools and extras are the exception: PDF to Word, table export, CSV form fill, watched-folder automation, and the extras in Compress, Merge and Split (a target file size, merging a whole folder, and splitting by bookmarks, page count or file size) unlock for 180 days when you turn sharing on, and lock again the moment you turn it off. If you never enable it, nothing else in this section applies to you and Pagekind behaves as a local PDF editor. Enabling it shows you a consent screen first, it needs an administrator account and a declaration that you are 18 or older and in the United States, and you can turn it off at any time.
What you should know before enabling it. Once on, it runs in the background and keeps forwarding traffic even while Pagekind is closed, and it is granted Windows firewall access automatically, so no Windows prompt appears. Turning it on is what unlocks the premium tools and extras above for 180 days; turning it off locks them again immediately.
What it does. While connection sharing is enabled and running, your computer forwards other people’s internet traffic out through your internet connection, as a continuous stream. Any website, service or game that traffic reaches sees your IP address, not theirs. Any kind of internet traffic is carried, not only web browsing. The exceptions are outgoing email, which is always blocked, and your own home network and other private addresses, which the traffic being carried cannot reach. There is one narrow exception to that: name lookups are passed to your own resolver, which is usually your router and so is itself on your home network; that path is used only to resolve names and is checked to stop it being used to reach anything else there.
What this means for you. Because destinations see your address:
- Your IP address may see extra CAPTCHAs, be rate-limited, or appear on third-party blocklists.
- Your internet provider’s terms may prohibit this, and any complaint about that traffic would come to you.
- Turning sharing off locally is immediate, stops your computer sharing within moments, and needs no server contact.
What leaves your computer when you enable it.
Enabling sharing enrols your computer with Pagekind’s connection coordinator (operator-run servers). The following then leaves your machine and is stored by the coordinator:
- Your public IP address, which is also sent to an external IP address-check service (api.checkmyip.bot) that returns the country, region, city and network operator (ASN) it maps to, plus a reputation score; those results are stored. Your IP is looked up by a third-party service, not derived only on our own servers.
- A device and connection fingerprint, used to make shared traffic look like an ordinary home machine: timezone, locale, your resolver (DNS) addresses, connection MTU and whether you have IPv6. Your computer’s CPU cores, memory, screen and GPU are reported separately by the sharing component while it is running — shortly after it starts and about every half hour after that — rather than when you first enable it.
- The public halves of your Pagekind connection keys, your platform and app version, and your consent record (its version, a hash of the accepted text, and the time).
- While sharing runs, a visit ledger: for each job your connection carries, the coordinator records your IP, the destination’s domain (not the full web address or page), start and finish times, and whether it succeeded.
What does not leave your computer.
- The content of the traffic you forward is never written to disk or sent anywhere as content; it passes through memory only long enough to forward it. There is no packet capture and no payload logging.
- Your private keys never leave your computer.
- The names people look up (DNS) are relayed to your own resolver and checked only to protect your home network; those names are never logged or stored.
Who can access this data.
The coordinator record described above is not public. Retrieving it over the internet requires two things together: the request must reach the coordinator’s panel host by that host’s own name (a request that does not is refused before it is answered), and it must present an operator access token. Both controls apply; neither alone is enough. Operators use this access to run the service and issue connection configurations. The record is guarded by those controls, but it is not confined to the server’s disk either.
Operator alerts (Telegram).
The coordinator also sends the operators an alert through Telegram each time something changes for a household — when one joins, is promoted or demoted, withdraws, or goes offline and later comes back. Each alert identifies the household and its country and region, and says what changed and when. Because these events recur, this is an ongoing feed rather than a one-time lookup. The alerts do not include your IP address. Telegram receives and stores these alerts as our messaging provider; it is listed under Providers and optional actions.
How long it is kept.
In the live database, server-side data is kept for a bounded time rather than indefinitely, and it is not deleted the instant you turn sharing off:
- Visit-ledger rows age out after about 90 days.
- If you withdraw your household, its record is reduced rather than removed, after about 45 days. At that point the coordinator deletes your visit-ledger rows outright — earlier than the 90-day window above — and clears your IP address, location, network operator and device details. It keeps a small permanent marker: your household’s public key, the fact that you withdrew, and the date. These are kept indefinitely, with no expiry. That key is what lets the coordinator recognise a withdrawn household and refuse to let it re-enrol, so that access that has been revoked cannot be restored by signing up again. Because retirement is designed to be permanent for that reason, a withdrawn record cannot be fully deleted on request; what can be removed is the location, network and visit-ledger data described above, which withdrawal already clears.
- Turning sharing off locally is immediate; the windows above are for the server-side copy, which is reduced on that schedule rather than erased on the spot.
Older copies exist outside that schedule. Copies of household records — including database backups and saved diagnostic captures (for example exported panel data in .json, .html or .txt files) — are retained on operator-controlled systems, including the coordinator itself, and are not covered by the removal schedule above. These can contain household records, public keys, IP addresses and visit-ledger entries that are no longer in the live database, so a household that has been reduced in the live database may still appear in them. We are reviewing how these copies are handled and will update this notice. There is no automated backup system beyond this.
Logs. The coordinator’s application log records household public IP addresses. It is truncated when the service restarts rather than kept to a defined retention period, and there is no separate web access log.
The coordinator runs on servers in the United States (Miami, Florida). The providers involved when sharing is enabled — Pagekind’s connection coordinator, the external IP address-check service (api.checkmyip.bot), and Telegram (operator alerts) — are also listed under Providers and optional actions.
3. Website visits, downloads and email.
When your browser requests a page, image or installer, hosting infrastructure receives technical information such as your IP address, requested URL, request time and browser information. That information is used to deliver the response, operate the service and protect it from abuse.
This public website is hosted on Cloudflare Pages, with the Windows installer served from Cloudflare R2. Visitors do not need a Pagekind account to read these pages or download the preview. The desktop app also does not require a Pagekind account. The Pagekind page code does not collect an account profile or provide a sign-up, newsletter, contact or document-upload form.
Downloading Pagekind sends a request for the installer. It does not send a PDF from your computer.
Counting installer-link requests. When you select a download button, the request passes through a short redirect on getpagekind.com before the installer starts. Each time, we store one record containing only the time of the request, so we can count how many download links are selected per day. It contains no IP address, browser details, cookie or identifier of any kind, and it cannot be traced to a visitor. Daily totals are shown on an operator dashboard. Requests that identify themselves as known automated crawlers, link-preview fetchers or our own health checks are left out of these totals; robots are counted separately as a bare time. Robots that do not identify themselves are still counted, so these totals are an upper bound and not a count of people. If the link carries the exact labels for our first Reddit campaign, we also store its fixed campaign and ad label with the request time, so we can compare the two ads. These labels are shared by everyone using that ad link; we do not store a Reddit click identifier, IP address, browser details or visitor identifier in this record. Only daily totals by ad appear in the campaign report. The download buttons on our tool pages also send the name of the page they are on, for example pdf-to-word; we store that name and the fixed source label in the record's storage name together with the request time, so we can see which page people download from. It names a page, not a person. These are installer-link requests, not completed downloads or installations. Repeated requests, automated crawlers and copied or fabricated campaign links can affect the totals. We do not send these request records to Reddit or run a Reddit pixel or conversion API. If you arrived from a Pagekind advertisement on Google, a separate record is stored only if you have accepted optional ad measurement; section 6 describes it and what is disclosed to Google. These older request records are kept until they are removed manually — no automatic expiry is configured yet, and this notice will state a period once one is. The current downloaded preview does not automatically check for later versions. If that changes, this notice will describe the update connection and its controls before the change is released.
Source-labelled requests for releases 1.1.7, 1.1.8, 1.1.9, 1.1.10, 1.1.11, 1.1.12, 1.1.13, 1.1.14, 1.1.15, 1.1.16, 1.1.17, 1.1.18 and 1.1.19
The new shared installer routes also record a fixed source label (Reddit text or OCR, Google Performance Max or Search, Microsoft text or OCR, or direct/unknown) and the request time. The record name contains a random storage nonce to keep requests separate; it is not a browser or visitor ID. These records contain no click ID, cookie, IP address, browser details or app journey ID. They are counted separately from the optional setup and app reports and cannot be joined to them. Repeated requests, crawlers and copied source labels are included, so a link request does not prove that the file arrived or Setup completed. These new records have the 30-day expiry and 7 November cutoff described above; the older request records keep their separately stated retention.
Counting page visits. When a page on getpagekind.com is loaded in a browser, we store one record containing only the time of the request. The record's storage name also holds four coarse facts: the UTC day, our fixed campaign label if the address carried one, whether the visitor looks like an automated crawler, and whether a well-formed Google click identifier was present. The identifier itself is never read into storage. The record contains no IP address, browser details, cookie or identifier of any kind and cannot be traced to a visitor. We use the daily totals to see whether people who select an advertisement reach the page, and where they stop. Reloads and automated visits that are not recognised as crawlers can affect the totals, so they are not a count of people. This counting does not depend on the optional ad measurement choice below because it stores nothing about you. Records are deleted automatically after 30 days.
Counting download answers. After you select Download, some guide pages offer an optional question about how the download went. If you tap an answer, we store one record containing only the time of the answer. Its storage name also holds the UTC day, the answer you chose and our fixed campaign label if the address carried one. It contains no IP address, browser details, cookie or identifier of any kind and cannot be traced to you. We use the daily totals to find the step where people get stuck. Anyone can answer more than once, so the totals are not a count of people. Records are deleted automatically after 30 days.
Emailing Pagekind
When you email us, we receive your email address, any name you provide, your message and attachments, and technical email details such as sending time and routing headers. We use this information to answer general enquiries, help with Pagekind problems and handle privacy requests.
Sending an attachment is optional. An attachment you email leaves your computer and is received by us and the email providers involved. Pagekind’s local PDF processing does not apply to files you choose to email. Describe the issue first and use a synthetic or redacted example when possible.
5. Why information is used.
Desktop data is used on your device to carry out the PDF task you choose and support local history, preferences and recovery. Website technical information is used to provide pages and downloads, maintain availability, enforce access controls and investigate abuse.
For processing covered by the GDPR, the proposed basis for necessary website delivery and security is legitimate interests in operating a reliable, secure product website (Article 6(1)(f)). Confirmation and documentation of the applicable basis and balancing assessment remain outstanding. Any processing required by a specific legal obligation would use that obligation as its basis.
For correspondence covered by the GDPR, the proposed basis for general enquiries and support is our legitimate interest in answering requests and helping people use Pagekind (Article 6(1)(f)). Handling statutory privacy requests relies on the applicable legal obligation (Article 6(1)(c)). Confirmation and documentation of the applicable bases remain outstanding.
The current Pagekind implementation does not use your document content for advertising, and uses no visitor information for audience profiling or for automated decisions that produce legal or similarly significant effects. Acquisition measurement includes the basic app-step reports and earlier optional milestones described above, installer-link request counts and fixed shared source labels. The separate Google click identifier described in sections 3 and 6 is recorded only when it is present in the link and you have accepted optional website ad measurement. Optional Google and Microsoft website measurement relies on your consent, which you can withdraw in Cookie settings.
7. Keeping and deleting data.
Your saved files: saved PDFs, exports and reports remain in the folders you choose until you or other software remove them. Pagekind does not remotely expire or delete them.
Working files and recovery: ordinary cleanup removes eligible temporary files when an operation finishes or a document closes normally. Crashes, failed cleanup and explicitly kept work can leave records available for recovery. Local preferences, saved groups, queues and recoverable work are preserved on uninstall. Kept OCR reviews may also survive closing and uninstall, but operating-system temporary-file cleanup can remove them.
Use the relevant Recover, Discard or Remove action to manage kept work, and keep a saved PDF for a durable copy. Removing a saved group removes its list entry, not the PDFs. Avoid deleting a workspace while Pagekind is using it.
Website and hosting records: this static website does not maintain a separate Pagekind visitor database. Hosting-log retention follows the provider configuration and applicable agreement. A site-specific retention period has not yet been confirmed; we do not promise an unverified number of days.
Email messages and replies can remain in our mailbox, and separate copies may remain with you and your email provider. A Pagekind-specific review and deletion schedule has not yet been finalized. The operator still needs to document retention periods or criteria that consider request completion, necessary follow-up and applicable legal obligations. Fastmail may retain separate recovery copies and operational records under its policies.
Ordinary file deletion is not secure erasure. Backups, earlier PDF revisions, Windows print data and other software may retain separate copies.
8. Protecting your information.
The website is served over HTTPS. Locally, Pagekind keeps edited output separate from the original and checks supported saved results. Neither behavior makes every local record encrypted: text reports, comment drafts, imports and some recovery metadata may be plain text.
PDF encryption protects the PDF where supported; it does not automatically protect exported images, print/spool data or companion files. Current OCR word-review records have Windows-account protection, while older preview records are not retroactively erased or protected.
Use a secured Windows account, control access to shared folders and backups, and avoid posting confidential documents in a public issue tracker or review. A synthetic sample is usually a better way to describe a problem.
9. Your rights and choices.
Depending on the law that applies, you may have rights to access, correct or delete personal information, receive a portable copy, restrict or object to processing, and withdraw consent where processing depends on it. These rights have conditions and exceptions. You may also complain to your local data protection authority; people in the EEA can find their authority through the European Data Protection Board.
Your right to object: where the GDPR applies, you can object to processing based on legitimate interests for reasons relating to your particular situation. That processing must stop unless the operator demonstrates overriding legitimate grounds or needs it for legal claims. To exercise your right to object, email privacy@getpagekind.com.
For documents and records that exist only on your device, use Pagekind and Windows to manage them. The website operator cannot remotely retrieve or erase those local files. Requests about information processed by a hosting provider for its own purposes can be directed through that provider’s privacy policy.
For a privacy request, email privacy@getpagekind.com. Start with a description of your request; do not include passwords or identity documents in an initial message.
10. Operator and contact.
- Product name
- Pagekind
- Country
- United States
- Business address
- 116 Russell Farm Road, Winchendon, MA 01475, United States
- Privacy contact
- privacy@getpagekind.com
- Product support
- support@getpagekind.com
- General enquiries
- info@getpagekind.com
These email addresses reach the same Pagekind operator mailbox. The business address is listed above. The operator’s legal identity and any other applicable business disclosures still need to be confirmed and completed.
11. Changes to this notice.
We will update the date at the top when this notice changes. New hosting, an update service, contact channels or optional measurement tools must be described before they are introduced. Where required, a material change will also be explained when you use the affected feature.
This notice covers the current release. It does not claim legal certification or guarantee the behavior of other software on your computer.
Detailed release notes.
The original engineering record contains additional storage locations and feature-specific recovery details. Its version references describe the preview in which each behavior was introduced; the sections above describe the current notice.
Local storage and feature details
Pagekind privacy note — release 1.1.19, October 10, 2026
THE DESKTOP APP PDF operations run on your computer. The app has no account system, document-upload endpoint, analytics SDK, advertising SDK or automatic crash reporting service. Automatic updates are disabled. The interface exchanges requests with a local worker through anonymous pipes. It rejects network file paths and URLs. No document content is sent by our code.
PDFs, temporary working files and saved copies are local files. Applied changes are stored as local working snapshots until the document is closed normally. After a crash or interruption, unsaved workspaces remain available through Recover. The state file does not store PDF passwords. Text still being checked may not have reached a stored snapshot. Your chosen folder may be managed by third-party backup or sync software; Pagekind does not control Windows, OneDrive, other programs or an administrator's policies. Saving into a synced folder is not a Pagekind document-upload service, but that other software may transmit the file. Use an ordinary local folder when that matters.
Automated core tests reject Python network calls. Native-library traffic has not been exhaustively instrumented. Physical network-disconnection validation and full clean-machine installation coverage remain release checks. We do not claim certification.
Unplaced Add text and Annotate values: each open PDF keeps separate words and settings for these tools in application memory. Changing tools or pages, saving and PDF Undo/Redo retain them. They are not PDF content until placed, are not included in a saved copy, and are not stored as recovery drafts. Closing asks before discarding pending values; they do not survive application exit. Find also keeps its current query in memory for explicit refresh, clearing it when the document session changes. Neither feature adds a persistent history file.
EARLIER 1.1.6 CAMPAIGN MEASUREMENT The following scope applies only to the earlier 1.1.6 Reddit installers. It does not authorize the 1.1.7, 1.1.8, 1.1.9, 1.1.10, 1.1.11, 1.1.12, 1.1.13, 1.1.14, 1.1.15, 1.1.16, 1.1.17, 1.1.18 and 1.1.19 setup/app diagnostics described below. October 2026 Reddit campaign installers offer Basic campaign measurement, selected in Setup. Clear it before installing, or turn it off under About. The first measured launch also shows a notice with Turn off. This is separate from connection sharing and does not affect PDF tools. Normal downloads, silent installs and older unmarked installations start with measurement disabled. When enabled, the app sends a random per-Windows-profile measurement ID, shared campaign/ad label, app version and event name to https://getpagekind.com/campaign-events. It reports each milestone once: first measured launch, saved sharing-consent acceptance, and a successful enrollment observed locally after that acceptance. It does not report consent screens being opened or declined. No PDFs, filenames, names, browsing history, connection keys, Reddit click IDs or precise event timestamps are sent. The local preference, ID and delivery state survive uninstall. Turning this off drops unsent events and stops future reports, without deleting records already received. The server stores a secret-keyed deduplication value instead of the submitted random ID, with fixed campaign/ad/version, event name and UTC receipt date. It does not store IP addresses or browser details in these records. Cloudflare receives the ordinary HTTPS request including IP and may process it for service delivery and security. No measurement events are sent to Reddit. Event records are scheduled to expire after 30 days; provider deletion may follow asynchronously. Aggregate reports may remain longer. Collection stops on 5 November 2026. These client-reported milestones are not proof of every installation or unique people, or that sharing is currently running. Opt-outs, offline use, copies, separate profiles and fabricated events affect counts. Full details: https://getpagekind.com/privacy/#campaign-measurement
BASIC SETUP AND APP REPORTS — RELEASES 1.1.7, 1.1.8, 1.1.9, 1.1.10, 1.1.11, 1.1.12, 1.1.13, 1.1.14, 1.1.15, 1.1.16, 1.1.17, 1.1.18 AND 1.1.19 Pagekind sends basic setup and app-step reports to us to find installation problems. No PDFs or filenames. Reporting starts enabled for eligible fresh interactive installs after a concise notice in the existing Setup flow. Turn it off in Setup or About. There is no extra first-open decision popup or manual report form. PDF tools work with reporting off. Connection sharing is separate, explicit and off by default. Existing disabled, declined or withdrawn choices remain off. Unreadable or ambiguous old choices remain off. Silent installation does not enable reports. App automatic reporting requires a protected completed fresh-install receipt, no older disabled or ambiguous state, the same authenticated Windows account and a receipt no older than 30 days. Other profiles stay off unless explicitly enabled in About. Website cookies do not authorize native reporting. Old measurement records are not relabelled as v3. While enabled, Setup reports measured start after notice and successful completion. The app reports first measured opening, actual display of the sharing offer and full terms, saved explicit sharing acceptance or explicit decline, observed enable failure, and locally observed enrollment and ready state. Closing or inactivity is not decline. Each event is kept once per retained journey and release. Reports from different releases do not prove another person or installation. Failure categories are not_permitted, unavailable, failed or unknown; raw error text is never sent. The body contains a new random journey ID, fixed shared source label, app and measurement-policy versions and event. No PDFs, filenames, paths, Windows account IDs, peer keys, ad click IDs, IP/browser details or precise event time appear in the body. The ID joins measured stages and is pseudonymous, not anonymous. Setup/app ID sharing requires the protected same-account receipt; the Windows account identifier stays local. Fixed Reddit text/OCR, Google PMax/Search, Microsoft text/OCR and direct/unknown labels are shared and copyable, not unique people or verified paid-ad attribution. Turning reporting off in Setup affects that setup and prevents fresh automatic app reporting from that setup. It does not change an already saved app preference: use About to turn that Windows profile's app reporting off. Turning app reporting off stops future app reports and drops unsent app reports; it does not erase records already received. Preferences and IDs can survive reinstall. Later enablement does not reconstruct actions that happened while reporting was off. Reports go to https://getpagekind.com/funnel-v3-events. The server stores a secret-keyed deduplication value instead of the submitted ID, fixed source/release labels, event, UTC receipt date and any coarse failure category. It does not store submitted IDs or request IP/browser headers in these event records. Cloudflare receives ordinary HTTPS connection metadata, including IP addresses, for service delivery and security. Native reports are not sent to Google, Microsoft or Reddit. Website ad measurement remains optional through Cookie settings; native controls do not inherit browser GPC. V3 event and timestamp-only source-request records are scheduled to expire after 30 days; provider deletion may follow asynchronously. New collection stops at 00:00 UTC on November 7, 2026. Aggregate reports may remain longer. Source requests include repeats/crawlers, use a storage nonce rather than a visitor identifier, and have no join to app journeys. Older request counters retain their separate policy. Reports count retained client observations and stage co-occurrences, not event order, every installation, unique people or current active peers. Offline use, opting out, loss, blocking, different profiles, resets, copies, expiry and fabricated reports affect counts. Missing events cannot reveal browser/SmartScreen/UAC warning exposure or why someone stopped. No browser-warning abandonment count is claimed. Full details: https://getpagekind.com/privacy/#setup-diagnostics
OPTIONAL CONNECTION SHARING Off by default, and separate from every PDF tool. If you turn it on, other people's internet traffic is routed through your connection and destination websites see your public IP address, including while Pagekind is closed. It runs in the background and is granted Windows firewall access without a prompt. Turning it on unlocks table export, CSV form fill and watched-folder automation for 180 days; they lock again the moment it is turned off. It needs an administrator account and a declaration that you are 18 or older and in the United States. The full notice at https://getpagekind.com/privacy/ describes what the coordinator stores.
THE WEBSITE The website distributes the installer and explains the app. It does not process PDFs. The website offers optional Google Ads and Microsoft Advertising measurement. Both tags load only after you choose Accept measurement. The providers can then receive page and referring-page addresses, ad-click information, IP address and browser/device details, and may use advertising measurement cookies. Google and Microsoft also receive an event when you select an installer link. This is not a completed download, Windows installation or active-peer measurement. Google ad personalization and Google Analytics storage stay disabled. Microsoft Clarity, UET Insights and enhanced conversions are off; we do not add remarketing audience lists or a Reddit pixel. Reject optional leaves downloads and PDF tools available. Cookie settings at the bottom of any page lets you change your choice. A first-party pk_ad_measurement preference cookie stores the decision and its date for up to 180 days, without a random visitor ID. An earlier Google-only choice does not authorize this setup; choose again. Global Privacy Control keeps optional measurement off. A successfully saved rejection stops future tag activity and Google click-ID recording, removes accessible first-party Google and Microsoft measurement cookies and reloads the page. It does not erase earlier reports or cookies held on provider domains. If your browser prevents saving a rejection, follow the error guidance to close the page and clear site cookies or retry the choice. Google policy: https://policies.google.com/privacy Microsoft policy: https://privacy.microsoft.com/privacystatement
Separately, the site counts installer-link requests: one record holding only the request time, with no IP address, browser details or visitor identifier. A link with the exact labels for our first Reddit campaign also records the fixed campaign and ad label with the request time; these labels are shared by everyone using that ad link. The campaign report contains only daily totals by ad. No Reddit click identifier is stored, and no request records are sent to Reddit. No Reddit pixel or conversion API runs on this site. These counts include repeated requests and automated crawlers; copied or fabricated campaign labels can affect totals. They do not measure unique people, completed downloads or installations. Only with accepted optional ad measurement, a Google advertisement click identifier present in a download link is separately stored for disclosure to Google Ads for installer-link request measurement. These older separate request records are kept until manually removed; no automatic expiry is configured. The full notice at https://getpagekind.com/privacy/ describes this measurement. The hosting provider necessarily receives ordinary web requests, including IP address and requested URL. Any hosting access controls are separate from the desktop app.
Do not send confidential documents to a public issue tracker or review site. Use a synthetic sample when reporting a problem.
Printing: Pagekind prepares page images locally. Only choosing Print in the Windows dialog submits them to your chosen destination. A network printer can receive the document over the network. Print/spool data and temporary page images are not protected by the PDF password. Pagekind removes its temporary image directory after preparation/submission finishes or is cancelled, when Windows permits. Printer spool storage is managed by Windows and the selected printer, not by Pagekind.
Favorites: the app saves only tool identifiers and their order in LocalAppData/Pagekind/Preferences/tools.json. No document filenames, extracted text or passwords are saved there. The file is not synced or transmitted. Application uninstallation preserves these personal preferences and unsaved PDF workspaces.
Batch tools: local TEMP/Pagekind-Batches-v1 stores selected local paths, hashes, settings and result messages, with no passwords. Each running job stages a copy inside a marked .pk-... folder in the chosen output directory. Normal completion, verified cancellation and explicit New queue cleanup remove these owned copies. A crash may leave them until Recover an interrupted queue verifies and cleans them. Completed outputs are kept. Text/image exports and result reports have no PDF password protection; reports contain local filenames and paths. Selected files on disk, not unsaved edits in another workspace, are batch inputs. There are no batch uploads, recursive folder scans, scripts or print jobs. Uninstall preserves queues and unsaved work for recovery.
Document tabs: page/zoom settings, tab records and remembered passwords remain in application memory. Applied PDF edits and undo history use the existing owned local workspaces; no new persistent tab file or password store is added. Closing a tab clears its remembered snapshot keys. After a window exits, unsaved workspaces can be recovered into separate tabs; passwords must be supplied again. A workspace that cannot be checked can be explicitly retained while closing the application.
Redacted image copies: the engine reads the source into bounded local memory and builds a new PDF using only masked RGB page pixels. Original PDF objects and hidden payloads are not copied. The exported file is unencrypted, and visible content outside your marks remains. Source files and editing history are unchanged; this does not sanitize or erase them from disk. Unredacted raster intermediates are not written to disk. A hard-killed worker can leave an unpublished temporary file of partial redacted output. Marks and their local undo history stay in memory only while the document tab remains open; they are not recovered after the app exits. Lost export replies are checked against the actual selected output before success is claimed. Existing files are never overwritten by this feature.
Interrupted saves: local workspace records contain the chosen output path, snapshot identifier, revision, byte count and hash, without passwords. Exact byte checks verify a completed copy before its saved state is recovered. Saved PDFs retain their snapshot encryption. An abrupt stop can leave a temporary PDF copy containing unredacted content until cleanup. Recovery preserves working copies after an unverified result. Uninstall does not erase recoverable workspaces.
New OCR workspace operations keep page counts and completion metadata in their recovery receipt, without a recognized-text transcript. The searchable PDF keeps its existing password encryption. Older previews could retain an OCR transcript in their plaintext workspace receipt; this change does not erase older records.
Reply drafts: TEMP/Pagekind-Workspaces-v2 keeps reply words, the local display name, an exact parent target and a short parent preview in the owned workspace state file. These drafts are local plain text without PDF password protection; reply drafting for encrypted PDFs is not offered in this preview. A pause in typing or a successful draft keep stores the current valid draft. Invalid text or an unconfirmed keep may remain only in memory, with feedback in the app.
Choosing to keep a reply while closing retains its workspace, including the working PDF and history, for Recover. This also applies to otherwise unchanged PDFs. Uninstall preserves that recoverable work. Posting clears the matching draft in the same state commit that records its PDF snapshot and result. The PDF then contains the reply words, display name and dates as unverified metadata. Undoing a post does not restore its old draft; an existing separate draft stays kept. No reply content is uploaded or sent to another reviewer by Pagekind.
Review decisions: The reviewer label, choice and conversation context stay in memory while navigating an open document. Unrecorded choices are not retained for restart recovery. Record decision writes the exact label, standard decision, date and previous-record reference into the working PDF, which is retained with its local undo/recovery history. Save a copy writes those records to your chosen PDF file. Names and dates are unverified metadata, not authenticated identities. No decision content is uploaded or sent to another reviewer by Pagekind.
Posted-reply corrections: A kept correction stores its new words, original reply text, original author label and short parent context in the local workspace for recovery. It is separate from a new-reply draft. Apply reply correction changes the working PDF and consumes its matching draft; discarding removes the draft from the current workspace state. Save a copy writes applied changes to a PDF. Reply correction and leaf removal preserve earlier PDF bytes and Undo history; removal from the current comment list is not confidential-content erasure. No reply text is uploaded or sent to other reviewers by Pagekind.
XFDF comment exchange: Inspection reads a chosen local XFDF file as data and keeps its inspection in that document tab while it is open. It does not open referenced files or URLs, execute actions or import comments into the PDF. Export writes supported plain-text comments, author labels, dates, relationships and document filename/ID hints to a new local XFDF file. The PDF and its Undo history are unchanged. Pagekind does not upload or send the exported file; its comment text and metadata are included for the person you choose to share it with.
Deliberate XFDF import: Choosing Review an XFDF import keeps a local, unencrypted copy of the chosen packet and its analysis with the receiving document workspace. The stored records include comment words, author labels, dates, relationships and position metadata. Explicit receiving-page review is required before Apply. Kept reviews can survive closing the document and recovery; they are not applied on recovery. Discard removes the task-owned kept review, leaving the external XFDF and original PDF unchanged. Pagekind does not send any of this data elsewhere. Applied comments remain in working PDF revisions and Undo history; Discard is not a secure erasure operation. Referenced URLs and files are not opened.
Saved reading places: only explicitly kept references are written to LocalAppData/Pagekind/Preferences/places.json. These contain the document path, file hash and byte count, page, zoom, position, fit and continuous-view settings, and update time. They do not contain PDF text, thumbnails or passwords. Remove and Clear saved places remove references without deleting the PDF. These local preferences survive uninstall.
Saved PDF groups: only explicitly saved groups are written to LocalAppData/Pagekind/Preferences/groups.json. Each contains a chosen name, member order, local PDF paths, saved-file hashes and byte/page counts, with no PDF contents or passwords. Renaming, reordering or removing a group changes this list only. Checking a member reads its saved bytes without opening a PDF workspace. Opening selected files copies verified bytes into owned local workspaces. Per-window opening records under TEMP/Pagekind-Workspaces-v2/owners contain request and workspace identifiers, hashes, byte counts and completion states; they omit passwords and plaintext source paths. They help reconcile interrupted openings without duplication. Finished dead windows' records can be removed; uncertain records and workspace files are kept. Group lists are preserved on uninstall. Ordinary PDF opening and closing do not add members automatically.
0.50 and earlier OCR receipts: applying OCR can keep a recognized-text transcript in the local plaintext workspace receipt, even when the PDF remains password protected. Avoid OCR on sensitive password-protected documents in these previews. A later correction removes transcripts from new receipts; it does not erase older records.
OCR word review (0.54): TEMP/Pagekind-OCR-Reviews-v1 stores an independent copy of the scan with its original PDF encryption and a protected recognition/correction record for the current Windows account. It is separate from the ordinary document workspace and remains after closing a document so Keep for later can be recovered. PDF passwords remain transient and are required again after restarting. The recognition image and review preview are processed in memory without a temporary raster file. Protected word records are authenticated; unavailable protection does not fall back to plaintext. This does not promise secure memory erasure.
Opening the kept scan creates an ordinary owned PDF workspace. Its opening and application receipts contain local source metadata, hashes and counts, without the recognized transcript, corrections or password. Completed reviews remain available to manage or discard. Discard first records a protected terminal result, then removes only its own scan and protected temporary writes; a failed removal can be retried from Recover. It is normal file deletion, not secure erasure. Uninstall preserves these local reviews. Operating-system temporary-file cleanup can remove them; keep a saved PDF for a durable copy. A hard stop before review preparation finishes can leave an uncommitted local scan, and no automatic global storage eviction is performed. Earlier preview OCR records are not retroactively erased by this version.
PDF access and password settings (0.54): PDF reading and owner passwords remain in memory for the session. They are not written into workspace records. Supported edited copies retain the original PDF encryption credentials and permissions; saved settings are checked before output is published. Explicit replacement or removal requires owner access. PDF owner access is separate from Windows account access. Print and export restrictions are checked; low-resolution printing is limited to 150 DPI. Existing weak PDF encryption is retained unless the owner explicitly changes protection. Unusual security filters that cannot be safely retained are refused.
Password choices and access guidance (0.54) Reading and owner passwords for new protection remain in memory. Workspace records do not contain these passwords or the submitted password-setting options. Permission-only reading can use an empty reading password; reopening such a PDF does not recover owner access. Selected-text copying follows the PDF permissions. Opening settings and selecting protected text do not remove protection. Creating or changing form fields requires both content and comment permissions; filling existing fields is checked separately. Signed structures are not redesigned.