Data and analytics
Updated on 24 October 2026.
Your texts
Texts pasted into the reader or imported from TXT, their titles, settings and reading positions are stored in this browser’s localStorage. Adding a text this way does not upload it to our server. Before saving, the current text and title are also kept in the tab’s sessionStorage so the form can be restored after a reload if saving fails. The draft is removed after a successful save. It may be lost when the tab closes and is never sent to the server. The original TXT file is unchanged. Clearing site data deletes the library. There is no automatic synchronisation between devices.
You can save a JSON backup and restore your library from it. The file includes full texts, titles, positions, reading modes, settings, the last session’s results and the language of saved explanations. Keep it somewhere appropriate for its contents. Creating and restoring a backup takes place on your device without uploading it. Interface language and analytics consent are stored separately and are not included. Restoring does not count as adding new personal text.
@read_agent and Telegram
Sending a link or forwarding a post to the bot submits it to our server for processing. Your inbox stores the text, title, source link, time added and a technical Telegram user number used to control access. Your language preference for the bot and mini app is also stored on the server. Access to the inbox requires verified Telegram data. You can delete a text from the inbox; a copy already saved on a device is deleted separately in “My texts”. The local Telegram library is separate from the regular browser and from other Telegram accounts. Reading positions do not currently sync across devices. Yandex Metrica is not loaded inside the mini app.
“What is this text about?” summaries
A summary is created only at your request in the bot or the reader inside Telegram. “What is this text about?” opens a saved result or starts a new summary using AI. The chosen text is sent to an external AI service; our code does not add your Telegram user number, the title or the source URL. A machine-generated summary of three to five ideas is stored on our server with the material and reused without another model request. Russian and English versions are stored separately. Creating another language version uses another attempt and requires a separate confirmation. Changing the interface language alone does not call the model. The original text is unchanged. Deleting an inbox item also deletes its saved summaries.
To control costs, we store technical records of attempts: user number, time, status and available token counts. These records contain neither the original text nor the summary. They remain after deletion so that deleting texts cannot bypass the limit. Failed requests also count. The model request is not retried automatically. Daily and monthly limits apply; texts are limited to 32 KiB. Machine-generated summaries may contain errors.
Passage explanations
“Explain passage” sends the selected paragraph, up to 8 KiB, and up to 2 KiB of context on either side to our server and an external AI service. The full text, its title and your user number are not added to the model request. The completed explanation is saved with the text in your local library and backup. Changing the paragraph, context or answer language requires a new explanation. Reading normally does not call the model.
No account is required on the website. Explanations use a random technical identifier in a secure cookie lasting up to 180 days. There is a limit of five new attempts per day per browser and an additional limit per network address. The server stores an HMAC of that address, without the original IP, to control costs. Telegram has a separate limit of five explanations per user per day; its summary limit remains separate. Both features share a budget of 50 requests per day and 500 per month, with day boundaries in Moscow time. Failed and interrupted requests count.
Closing the window does not cancel preparation. On the server, the passage and context are cleared after processing finishes or is interrupted. The temporary operation and answer are deleted 24 hours after creation. Technical attempt records remain without the content. Saved local explanations can be restored from a backup after clearing browser data. Cookies and pending request numbers are not included in backups. Refusing the functional cookie disables only new explanations on the website.
Browser extensions
Fastreader extensions extract an article or selection only when you ask them to. The title, plain text and source URL are transferred locally to the fastreader.ru/read tab. The server does not receive the text during this import. Source-page markup and scripts are not executed in the reader. The extension needs access to fastreader.ru to deliver the text. Access to other pages lasts only for your action.
Until saving is confirmed, the temporary text is held in the extension’s storage.session. It is not synchronised and is removed after successful saving, closure of the receiving tab, browser shutdown or 30 minutes. Expired entries are cleaned up the next time the extension is used. The library and progress stay on your device in the same browser profile as the website. The source URL and import fingerprint are included in backups; temporary transfer numbers are not.
The extension contains no analytics, does not inspect browsing history and does not call the model by itself. Website analytics remains optional. A selected passage is sent to the model only through the separate explanation feature. Data is used only to add and read the chosen material, is not shared for advertising and is not made available to people for review. Support: hichnick@gmail.com.
Optional analytics
Yandex Metrica counter 113701852 loads only after you press “Allow”. It is not loaded before a choice or when you opt out. The reader works regardless of your choice.
After consent, Metrica receives information about page visits, referral source, browser and device, plus events for opening the reader, successfully saving a new personal text and starting word-by-word playback for the first time. On the home page, clicks on links to the bot, mini app and published Chrome and Firefox extensions are counted. These events record following a link, rather than installing an extension or beginning to read. Events do not include texts, titles, filenames, document identifiers, link labels or target URLs. Our counter code does not send the page URL’s query parameters or fragment.
Session replay, click maps, outbound-link tracking and page-title reporting are disabled. Forms and reader content are additionally excluded through markup. Enhanced matching and automatic form collection must remain disabled in the counter settings.
Metrica may use its own cookies and technical identifiers for visit statistics. Its processing is described in the Yandex privacy policy.
Changing your choice
Use “Analytics settings” below the top navigation on any page. Opting out after consent reloads the page without the counter. This does not delete statistics already collected. Your choice is stored separately from the library. A technical flag in sessionStorage prevents pause and reload actions from duplicating the first-play event. These flags contain no texts.
The interface language preference is stored separately in this browser. In automatic mode it follows your browser settings. Public-page language is determined by its URL; either version can always be opened directly. Telegram keeps its language preference on the server for the bot and mini app. Changing the language does not replace user texts, titles or reading positions.
Server requests
Your browser requests pages and supporting files to open the website. The web server keeps technical request logs which may include the requested address, IP address, time, referrer and browser information. Do not put private text in a page URL: use the text form instead.