Privacy

Privacy details

What Readback does with your reading, what leaves your device, and how to remove it.

What stays on your device

Reading activity. When tracking is enabled, active reading time and scroll depth for eligible articles on supported sites are kept on your device. An article is saved only after you have read it for the minimum reading time in one visit (30 seconds by default, 10 seconds to 30 minutes in settings) or you keep it; a page you leave earlier leaves no record. The supported-site snapshot includes entries from the engineering blogs and personal blogs lists, plus selected sites. Some listed sites use unencrypted HTTP.
Active reading minutes per day. When tracking is enabled, Readback also keeps the total active reading time for each day on your device, for the private Streak view. It has no addresses, titles or text, it is not synced, and "Delete local data" removes it. Nothing about your streak leaves your device, and the sync list below is unchanged.
Article text. Kept article text stays on the device that kept it. Another device can show questions and results but cannot write new questions from text it does not have.
Excluded pages. Readback applies local checks to skip home, search, feed and shopping pages. General social, shopping and video sites are excluded; selected AWS and LinkedIn engineering-blog paths are supported. On shared sites, reading detection is limited to listed blog paths, but Chrome grants the extension access to the whole host.

Only kept articles. The text of articles you keep is sent to write questions, and only with your consent. Consent is a separate setting.
Grading. If you ask for a written answer to be graded, the question, the reference passage and your answer are sent for grading, with your consent.
Not stored. Readback's server does not store the article text, its title or address, or the prompts and replies. It does keep the generated questions in a shared cache (see below).
More questions. On a paid plan, the More button asks for another round of questions for an article. Along with the article text, Readback then sends the prompts of the questions you already have for that article (up to 30, each cut to 300 characters) so the AI does not repeat them. They go with the same AI consent and are not stored by the server.
No AI account. Readback provides the AI. You never bring a key or create an AI account.

What is stored. When the AI writes questions for an article, the server keeps that set of questions so the next person who reads the same article does not trigger the AI again. The set quotes short passages copied from the article (a few sentences each), never the whole article.
How it is found. It is filed under a one way hash of the article text, so it can only be looked up by someone who already has the exact same text. The title, address and your account are not in the file name or the file.
Shared. Everyone who reads identical text gets the same stored questions. Using a stored set still needs you to be signed in, to have agreed to AI, and to have kept the article, and it does not use up your free question sets.
Deleting it. Because these files contain no name, email or other identity, Readback cannot tell which ones are yours, so they are not removed when you delete your cloud data or your account. They hold only short passages of articles and no way to link them to you.

No account is needed for tracking, Keep and the local dashboard. A Readback account is needed for AI question writing, AI grading and sync. You create it with Continue with Google.
What Google shares. Your email address and a stable Google account id. Readback never receives your Google password or any other Google data, and never stores a password of its own.
Country. When you buy a plan, Readback's server uses the country of your connection only to choose rupees or dollars, and does not store it.
Sessions. After you sign in, Readback issues its own sign in token, valid for 30 days. Only a one way hash of it is stored on the server.

Off by default. Sync is a separate, explicit choice and needs an account.
What syncs. Article metadata (address, title, domain, active seconds, scroll percentage, word count, last read, kept, never quiz), question sets, cards and their schedule, answer results (correct or incorrect, time), and settings flags (tracking, deny list).
Never synced. Full article text, your typed short answers, raw events.

Anonymous product counts are a separate opt in, off by default. They never include addresses, titles or text.

On this device. Delete local data from Readback settings.
In the cloud. Delete all synced rows. If Sync is on, deleting your history asks whether to delete cloud data too.
Your account. Delete the account, its sessions and all of its data.