Legal

Doom Scroller Privacy Policy

Publisher, operator and responsible party

Doom Scroller is published by Vo Creations LLC, a Virginia limited liability company (Virginia SCC entity 11815104). Viewd is a service operated by Vo Creations LLC. The extension uses the Viewd creator account you sign in with, and Vo Creations LLC is responsible for the data the extension sends. The operator will change to a Viewd company once it is formed, and this policy will be updated then.

Who this is for

Doom Scroller is for Viewd creators. It cannot start a warm-up until you sign in to Viewd, open your dashboard, pair this extension install, and select a linked social account.

What the store build does

When you press Start on Instagram, TikTok, YouTube or Facebook, Doom Scroller watches Reels, TikToks, Shorts or Facebook Reels for the session length you chose. It varies its pace, adjusts how long it watches based on how each item compares with the recent feed, and can use verified search, profile and grid paths to find content matching the selected niche. It counts both automatic and manual verified viewing. Engagement stays off until it is enabled for that platform account. Within small daily limits, this release may Like, Save and Follow on Instagram; Like, Favorite and Follow on TikTok; Like or Subscribe on YouTube; and Like, Save, Follow or Share to feed on Facebook; and, on Instagram, TikTok, Facebook and YouTube, like one comment under a post it is watching when "Like comments" is enabled for that account. While watching it sometimes opens a post's comments, scrolls them and closes them again. Sharing a post to your own profile happens on Facebook only, and only when you have enabled it for that account. It does not comment anywhere, and it does not repost on Instagram, TikTok or YouTube. It pauses automatic control when you scroll, leave a supported surface, or type, and resumes only through the extension's controls.

Research tasks

When you ask Viewd to research a creator or a post on Instagram or TikTok, Viewd hands this browser a task for your paired account and, on its next minute, the extension opens that creator's videos page or that post in a tab on your signed-in account. It scrolls the creator's grid one screenful at a time at a reading pace, up to the task's item limit or the end of the grid, never likes, saves, follows or comments, and closes the tab when it is done. Each of that creator's videos the page shows is sent to Viewd with the same post details listed below, including the video URL, so Viewd can transcribe and analyze it; for a TikTok post it opens, the video URL is the link of the video playing on the page. It runs only while public-detail sharing is on and only under the same exclusive session lease as a warm-up; turning sharing off discards any research capture not yet sent. If you scroll, type, or the platform asks you to confirm your account, the task stops and the tab is left open for you.

On-device analysis of research videos

For those videos, this browser may also download each video from the platform's own servers, using your signed-in session for TikTok, transcribe its speech and read its on-screen text on this computer, and send Viewd only that transcript and text, up to 48 small still frames of it (one at each scene change and a few between, about 320 pixels wide, so your AI assistant can see the video), the video's fingerprint, how long each step took, and which speech model, number format and compute backend (WebGPU or WebAssembly) did the work, or, when a video cannot be processed, a short code saying why (such as a failed download or an unsupported computer); the video itself is never sent. The first time, the speech and text models are downloaded from viewd.net, which, like any website, sees your IP address. Only videos a research task you asked for captured are processed, one at a time, and never while a session is scrolling a tab for that account. A TikTok video is fetched with the cookies your browser already holds for TikTok and a TikTok Referer that the extension sets on its own requests to TikTok's video servers only (a session rule; your own TikTok tabs are untouched); the extension never reads those cookies. An Instagram video is fetched from the signed link the page was given, with no cookies. The speech and text models run in the extension; the first time, about 60 to 340 MB of them is downloaded once from viewd.net, checked against a fingerprint pinned in the extension, and kept in the extension's storage. Like any request to a website, that download shows viewd.net your IP address and browser; it carries no account token and nothing about the videos. The still frames (at most 48 per video, JPEG about 320 pixels wide: the first frame of each scene and a few spread between) are found on this computer and uploaded to viewd.net only after Viewd stored that video's transcript, under your account; only you (through your own AI assistant's connection to Viewd) can see them, never another user. What is sent about how a video was processed: the extension version, the compute backend (webgpu or wasm), the speech model's name and number format (dtype), and the text reader's name, frame rate and frame height. A video that could not be processed sends only its post's id and link and one code: media_fetch_failed:<why> (the download failed, for example the link expired, the server refused it or the file was not a video; no_host_access when on-device video analysis is not turned on), decode_failed, asr_failed, ocr_failed, unsupported_device (this computer cannot run the model), too_large (the result was too big to send) or cancelled. Viewd stores what this computer sends as the device's own reading, separate from Viewd's own results. Turning sharing off discards every video waiting to be processed and every result not yet sent.

Information sent to Viewd

The extension sends four groups of information to the Viewd service under your Viewd creator account:

1. Setup and account information. The Viewd issuer, extension version, installation device id and public key, pairing status, selected platform account key, selected preset, and the public profile-picture URL and identity the platform page shows for the signed-in account, with the short code of the rule that read it and whether it matched the selected account (also sent for a run that never started, so Viewd staff can see why). Pairing uses a one-time ticket. Runtime and public-report requests use short-lived Viewd bearer tokens. The extension does not send your social-platform cookie or password.

2. Session and activity records. These are sent whenever automatic or manual warm-up tracking runs, whether or not public-detail sharing is on. They include run ids; start, pause, resume and stop times and reasons; automatic and manual verified-viewing intervals; each credited public content id, when it was first credited, its author, caption or public context text, page type, relevance result and reason; the selected account and preset; planned and completed actions with their target content or creator (for a comment like, the post, the comment's author handle and the comment's id when the page shows one; the activity record never carries the comment's text), action type, outcome and machine reason; navigation, profile-dig, search, Explore/category and hunt outcomes; account readings used to check warm-up, including public following counts, up to 200 public handles shown on the account's Following page, and up to 50 public repost tiles; platform restrictions; restart outcomes; and the diagnostic records described below. The Viewd service keeps the session evidence and activity ledger.

3. Public text previews for scoring. During a session, the extension may send a public content id and a bounded block of public text for posts the page already loaded next in the feed and for visible grid tiles. These previews can include a title or caption, author or channel name, hashtags, sound, place, category, public creator keywords and a bounded description or bio line, depending on the platform. A preview is sent so Viewd can score niche relevance before the item is reached. It is not recorded as watched and is not stored as a viewed item. Viewd may send this text to its configured text-embedding provider solely to calculate the relevance score.

4. Public post, comment and creator reports. The “Share what you scroll past” switch is on by default and can be turned off under Advanced settings in the popup. On public Instagram, TikTok, Facebook and YouTube feeds, Reels, Explore, hashtag, search, post, video and profile pages supported by the reader, it can send: the platform; post id and link; caption and hashtags; thumbnail and video URL; video length; upload time; media type; credited co-authors; paid-partnership label; view, like, comment, share and save counts; creator handle and public id; display name; verified badge; profile-picture URL; follower and following counts; lifetime likes; post count; the page or feed where it appeared; and when it was observed. Video URLs are the short-lived links the page already received; the extension does not fetch the video. Private-account reports are dropped before they are queued.

Public following and followers lists (from the first build after 5.4.21). When you, or an automated session, open the Following or Followers list of a public Instagram or TikTok creator, the extension reads the list the platform already sent the page and reports each public account on it as a creator: its handle, public id, display name, verified badge, profile-picture URL and, where the list states them (TikTok does, Instagram does not), its follower, following, video and lifetime-like counts; plus whose list it was on (that creator's handle and, on Instagram, public id), which list (following or followers), the page it appeared on and when it was observed. Private accounts on a list are dropped before they are queued. A list whose owner hid it, or whose owner the page does not name, is not read. The extension never sends your own relationship to anyone on a list (whether you follow them or they follow you), never sends a listed account's bio, and does not open or scroll a list in order to report it: only a list already on the page is read. Your own account's list is read the same way when it is on the page (because you opened it, or because a session opened your Following list to check warm-up), since it is the same public list anyone can open. Reading other parts of the platform's relationship data (who you follow through follow buttons, follow requests, suggested accounts, mutual followers) stays refused.

Public comments. Under the same “Share what you scroll past” switch, on by default, the extension also sends the public comments it sees under public Instagram, TikTok, YouTube and Facebook posts, whether you or an automated session opened the comments. For each comment it can send: the platform; the post's id; the comment's id and, for a reply, the id of the comment it answers; the commenter's public handle, public id, display name and profile picture link; the comment text as shown; its like and reply counts; when it was posted; whether it is pinned and whether the post's creator liked it; the page or feed where it appeared; whether it was read from the platform's own response or from the page as displayed; and when it was observed. Comments are never sent from a private or non-public post, and never from your own direct messages or notifications. Viewd keeps public comments indefinitely.

Ads (from 5.4.28). Under the same switch, an ad the platform serves in a feed (an Instagram ad, a TikTok For You ad, a YouTube Shorts ad, a Facebook sponsored story) is recorded as an ad and never as a post: it never reaches the post, creator, view-count or format records. For each ad sighting the extension can send: the platform; the platform's ad, creative and advertiser ids where it states them; the advertiser's name and handle; the ad's own post or video id; whether it is a creator's post the advertiser boosted (TikTok Spark Ads); the ad label ("Ad", "Sponsored"); its caption; its button text ("Shop now"); the landing page it points to; its cover image link and video length; the reasons the platform gives for showing it to this account (TikTok's "About this ad" lines, such as an age range); its view, like, comment and share counts; which feed it appeared in; when it was seen; and the handle of the platform account the extension has bound to that tab (the signed-in account its runs already report). The extension never clicks, taps or opens an ad, its button or its link, and never requests a click-tracking link (fetching one would count as a click the advertiser pays for): the landing page is read from a plain field (TikTok) or decoded on this computer out of the ad's own link (YouTube's Google ad link, Instagram's and Facebook's link redirectors), and the platform's per-click ids (ttclid, gclid, fbclid and similar) are removed from it. How ads are found: TikTok and Instagram ads come out of the same responses their posts do (the For You, Reels, feed and Explore answers), marked by the platform's own ad fields; YouTube Shorts ads come out of the two Shorts responses that carry them (reel_item_watch and reel_watch_sequence), and Facebook ads out of its GraphQL feed and Reels answers, where a story's sponsored_data is set. On YouTube and Facebook those responses are read for ads ONLY: a response with no ad marker is not parsed, nothing else in it is read, and requests for messages, notifications, accounts or saved items are never read. Facebook ads are marked unconfirmed until a captured Facebook ad pins their fields; until then only their ad id is read. Ad scrolling does not change: a session still moves past an ad without lingering, and nothing is pressed on it.

YouTube public reporting (5.4.4) reads the visible Shorts or watch player’s already-loaded metadata, only when its video id matches the page and it explicitly says the video is public, crawlable and not unlisted. Reports include the title/description, public channel handle and id, thumbnail, duration, upload date and any supplied view/like counts. Missing counters and channel statistics stay null. Unlisted/private videos, unknown visibility, ads, hidden tabs and playlist contexts are withheld. Channels without a supported public @handle are withheld rather than guessed from a display name. No API request, stream URL collection or MP4 download occurs, and no response is read for these post reports (from 5.4.28 the two Shorts responses above are read for ads only, and a Short served as an ad is never reported as a post). The same sharing switch applies. Observation uploads include the extension build version for provenance. An upload Viewd refuses or cannot answer keeps its reports queued and is retried only after Viewd's Retry-After, or otherwise one minute doubling to thirty; the first delivered upload clears that wait.

Facebook, released in 5.4.0 on 2026-09-20, differs in how it is read and in what it can say. Instagram and TikTok reports are parsed out of the JSON those sites already sent the page. Facebook sends no such payload this extension can read, so a Facebook report is read from the page as it is already displayed: the reel the player is showing, the view count printed on each tile of a profile's Reels grid, and the follower and following counts printed on a profile or Page header. No Facebook network response is read for these reports (from 5.4.28 GraphQL feed answers are read for ads only, as above), and no Facebook page is clicked, expanded or fetched to produce a report. Three consequences are worth stating plainly:

- Facebook counts are sent as Facebook prints them. Facebook publishes no exact figure beside an abbreviation, so 40.8K is reported as 40,800. - From 5.4.8 a Facebook post is reported only when the page shows its video id and exactly one owner account: the reel in the Reels player (its like, comment and share counts and its length), a tile in a profile's Reels grid (its view count), or a feed, search, hashtag or profile row that links a reel or video (the id alone, with no counts, because the row prints bare unlabelled numbers). Any other Facebook post, or a row naming two accounts, has no readable id or owner, so no post report is sent for it. Its author is still reported as a creator. - Facebook reports never carry a thumbnail, a video URL, an upload time, co-authors, a paid-partnership label, a verified badge, a profile picture, a save count, a lifetime-likes total or a post count: Facebook does not display any of them where the reader can see them.

The public-report switch controls group 4 only. Turning it off stops those reports and discards any queued public reports that have not been sent. It does not turn off pairing, account selection, session records, activity records, preview scoring or the account-avatar update. Stopping all of those requires stopping tracking and removing or signing out of the extension.

What Viewd sends back

Viewd sends the extension the linked accounts and presets it may use, warm-up status, limits, the kill-switch status for this build, and a 0-to-1 niche score for content it can assess. A closed code may say that Viewd already processed, sorted or tracked the content. Viewd can also lend the public view count it already has for a known post.

What is not read or sent

Doom Scroller does not read or send your social-platform password, cookie or session token; direct messages; notifications; comments on private or non-public posts; saved or liked lists; manually entered search history or search suggestions; payment, health or precise-location information; or content from a private account through the public-reporting arm. The public-reporting arm does not send the viewer-state flags that say whether you liked, saved, reposted or followed a public item, nor your own relationship to any account on a public following or followers list. Session records do record actions Doom Scroller itself requested and their outcomes, and account verification reads the public following handles and repost tiles described above.

Diagnostics

When a supported platform changes, the extension reports the build version, page type, machine reason, the URL path with variable segments replaced by <segment>, and counts of page structures such as dialogs, main regions, headers, articles, videos, tiles, search inputs, toolbars and navigation markers. It also reports performed navigation names and whether the expected destination appeared. Unknown-page diagnostics contain no caption, handle, heading or other page text.

Remote code

None. Every script the extension runs ships in the package. The package contains an integrity file with a hash for every shipped source file.

Local storage

Chrome extension storage holds the paired device record, short-lived Viewd tokens, selected accounts and presets, current and recent run state, unsent public reports, account avatars and scheduled warm-up state. The extension keeps this under Chrome's 10 MB extension-storage quota and compacts acknowledged run state when it needs room. Web pages cannot read this extension storage.

Retention

Vo Creations LLC keeps session records, activity records and public reports through the Viewd service against the Viewd creator account for as long as the account exists, unless a shorter removal request or legal requirement applies. Public comments reported under group 4 are kept indefinitely, and so are ads reported under group 4, with each sighting's signed-in account handle and the platform's stated reasons the ad was shown to that account. Public scoring previews are not stored as viewed items. Contact the operator to request earlier removal.

Automation and platform rules

Doom Scroller automates an Instagram, TikTok, YouTube or Facebook account at the user's instruction. Using it may conflict with a social platform's terms. The user runs it at their own risk. The extension uses limits, starts automation only after the user asks, and stops automatic control when the user takes over.

Contact

For privacy and data questions, contact Vo Creations LLC at viewdnet@gmail.com. You may also use the support contact in your Viewd dashboard.