TitleTrackr AI Assistant — Privacy Policy

Overview

The TitleTrackr AI Assistant is an internal, unlisted Chrome extension provided to TitleTrackr's title-search customers. It is a side-panel assistant that helps title examiners and abstractors work on county-records and title-plant websites: it answers questions about the page you are viewing, collects PDFs you choose into a document stack, and hands those documents to the TitleTrackr web application to create an abstract.

You must sign in with a TitleTrackr account to use the extension. Access is restricted to authenticated TitleTrackr users and their associated tenant (organization). This policy describes exactly what the extension collects, what it does not collect, where data goes, and why each browser permission is requested.

Information We Collect

The extension collects only the following, and only in the circumstances described:

• Account and authentication data. When you sign in via TitleTrackr's OAuth flow, the extension receives and stores a bearer token along with your user ID, your email, and (where applicable) your tenant ID and tenant name. The token is read from the OAuth redirect hash (not query parameters) so it stays out of server logs.
• Chat messages and conversation history. When you use the assistant, we send the text you type and the prior conversation history to the TitleTrackr backend.
• Current page content (at question time). When — and only when — you send a chat question, the extension reads the current tab's visible text (capped at 12,000 characters; password/hidden fields and script/style content are excluded) and its URL and title, and captures a single compressed screenshot of the visible area, sending these with your question so the assistant can answer about the page you are viewing (including scanned documents whose text can't be selected). This happens per question, not continuously or in the background.
• PDF text content. When a PDF you have manually added to your document stack contains a readable text layer, its extracted text (no OCR; capped at 50,000 characters) may be sent to the TitleTrackr backend for analysis.
• Usage telemetry. We record lightweight product-usage events (for example when the panel is opened, a document is added, or an abstract is created) along with optional non-identifying metadata, a timestamp, and your tenant ID and user ID. Telemetry does not include the hostname or URL of the site you are on, page content, or your chat messages.
• Abstract templates. We fetch your tenant's list of abstract templates so the correct template can be pre-selected when you create an abstract.

How We Use It

• Account data authenticates your API requests and shows which account and tenant you are connected as.
• Chat messages, conversation history, page content, screenshots, and PDF text are used solely to generate the assistant's responses and analyze the documents you are working on, on your behalf.
• Telemetry is used to understand feature usage and reliability for the internal product. It is not used for advertising or profiling.
• Templates are used only to populate the abstract-creation workflow.

We do not use collected data for purposes unrelated to the extension's single core function, and we do not sell or share it with third parties for their own purposes.

AI Processing

To generate responses, your content is processed by large language model (LLM) systems operated by or on behalf of TitleTrackr. Visual questions that include a screenshot are processed by a vision-capable model. Content sent for AI processing includes your chat messages and conversation history, the current page's visible text and a screenshot captured when you ask a question, and (for document analysis) extracted PDF text.

This content is processed solely to return a response to you. It is not used to target advertising, and it is not used to train AI models unless you explicitly opt in.

Data We Do Not Collect

The extension is deliberately limited. It does not:

• Continuously scrape, log, or transmit pages in the background. The current page's text and a screenshot are read only at the moment you send a chat question — never passively as you browse.
• Track or transmit your browsing history.
• Record the hostname or URL of the sites you visit in its usage telemetry.
• Upload raw PDF bytes to a separate document store. PDFs are processed locally (thumbnail, page count, and text-layer detection happen on your device), and the merged document is handed to the TitleTrackr web app's create-abstract page, which owns upload and OCR.
• Send search queries you type into a page to TitleTrackr. Search-fill is a purely local browser action.

Storage & Retention

• Authentication: Your bearer token and identity details are stored locally in the browser and persist across sessions until you disconnect or the token becomes invalid; an expired token automatically clears your stored credentials.
• Document handoff: When you create an abstract, your selected PDFs are merged and temporarily stored locally only long enough to hand them to the create page, then cleared.
• In-memory only: Chat history, the document stack, and PDF thumbnails are held in memory and cleared when the panel closes.
• Backend telemetry: Usage events are stored server-side in TitleTrackr's systems. Screenshots and chat content are sent for processing and are not persisted in the telemetry store.

Permissions & Why

• sidePanel — to open and display the assistant's side-panel interface.
• identity — to complete TitleTrackr sign-in via Chrome's web auth flow.
• storage — to persist your auth token/tenant data and temporarily stash merged PDFs for handoff to the create-abstract page.
• scripting — to read the active tab's visible text for a chat question, fill a search field, and hand documents to the web app's create page.
• tabs — to read the active tab's ID and URL needed for screenshot capture, search-fill, and detecting the create-abstract handoff page.
• downloads — to detect when you download a PDF so it can be offered for capture into your document stack.
• Host access (all sites) — when you ask the assistant a question, the panel reads the current tab's visible text and captures a screenshot so it can answer about the page you are viewing. Because there is no fixed list of county-records or title sites a searcher uses, this requires access to all sites, which is why Chrome shows the "read your data on all websites" notice at install. The page is read only at question time, never passively.

Security

All extension API requests are authenticated with a bearer token and a tenant identifier; the backend endpoints require a valid token and reject unauthenticated requests. Communication with the TitleTrackr backend occurs over HTTPS, and local data is cleared on disconnect or token expiry.

Contact

Questions or requests regarding this policy or your data can be directed to support@titletrackr.com.