---
title: "Privacy Policy – Probeli AI"
description: "What personal data Probeli AI processes on its website, in the app, in emails and payments, and when its crawler reads public pages, and the rights you have."
url: "https://probeli.ai/legal/privacy-policy"
updated: "2026-11-19"
---

# Privacy Policy

Last updated: November 19, 2026

**On this page**

1. 1. Who is responsible and how to contact us
2. 2. Summary
3. 3. When you visit our website
4. 4. When you create an account and sign in
5. 5. When you use Probeli AI
6. 6. Public web pages our crawler reads (ProbeliAIBot)
7. 7. Emails we send
8. 8. Payments
9. 9. Error monitoring and logs
10. 10. Preventing fraud and abuse
11. 11. Cookies and similar technologies
12. 12. Recipients and sub-processors
13. 13. International transfers
14. 14. How long we keep data
15. 15. Deleting your data
16. 16. Your rights under the GDPR and UK GDPR
17. 17. US state privacy rights
18. 18. Security
19. 19. Children
20. 20. Automated decision-making
21. 21. Customer content and our role
22. 22. Changes to this policy
23. 23. Contact

Probeli AI is a tool for businesses that shows how AI engines (ChatGPT, Gemini, Google AI Mode and Perplexity) talk about a brand and its competitors. This policy explains what personal data we process when you visit our website, create an account, use the app, receive our emails or pay for a plan, and when our crawler reads public web pages.

The short version: we collect what we need to run Probeli AI, we don't sell personal data, and we measure how our website and app are used with DataFast analytics, and the Reddit Pixel and Meta Pixel for ad attribution (section 3.2).

## 1. Who is responsible and how to contact us

> **In brief:** Probeli AI is responsible for your personal data. Write to [hi@probeli.ai](mailto:hi@probeli.ai) with any privacy question or request.

The controller responsible for the processing described in this policy is Probeli AI, operating as Probeli AI ("we", "us"), which operates our website at probeli.ai and the Probeli AI app at app.probeli.ai. Probeli AI is established outside the European Union. For any privacy question or request, write to [hi@probeli.ai](mailto:hi@probeli.ai).

When our customers track prompts and brands in Probeli AI, they decide what goes into their workspaces. For personal data inside that content, we act on the customer's behalf (see section 21).

## 2. Summary

| Situation | Personal data | Why | Legal basis (GDPR) | How long |
| --- | --- | --- | --- | --- |
| You visit our website | IP address, browser details, pages requested | Deliver and protect the website | Legitimate interests, Art. 6(1)(f) | Up to 30 days |
| You contact us | Name, email, topic, message, and company and website if you add them | Answer you | Legitimate interests, Art. 6(1)(f); steps before a contract, Art. 6(1)(b) | As long as needed to handle your request |
| You run the free site audit | Email address, website domain, article language, the audit's results, when you agreed to our emails; a one-way hash of your network | Run the audit, email it, one sample article and one follow-up to you, and see whether you become a customer | Consent, Art. 6(1)(a), for the emails; legitimate interests, Art. 6(1)(f) | One year unless you create an account; unsubscribed addresses are kept as a hash |
| You create an account and sign in | Name, email, password hash, passkeys, two-factor data, Google profile, session IP address and browser, bot-check signals | Run your account and keep it secure | Contract, Art. 6(1)(b); legitimate interests, Art. 6(1)(f) | Until you delete your account; sessions 30 days after last use |
| You use Probeli AI | Workspace content, team members, invitations, activity | Provide the service | Contract, Art. 6(1)(b) | Until the workspace is deleted; full answer text per plan |
| You use the support chat | What you type in the chat, the page you opened it from, IP address and browser details, a random chat ID | Answer your questions about Probeli AI | Legitimate interests, Art. 6(1)(f) | Until we delete the conversation or close our account with the chat provider |
| You connect Google Search Console or Analytics | Your Google properties' names; search clicks, impressions, positions, pages and queries; visits from AI assistants per page; indexing status of your articles; an encrypted access token | Show your own Google numbers in your workspace and flag your articles that lose traffic | Contract, Art. 6(1)(b); your consent in Google's screen | Synced data 16 months; deleted at once when you disconnect or delete the workspace |
| Our crawler reads public pages | Anything a public page contains, such as names | Find indirect mentions, run site audits, write business briefs, research articles, check published articles and suggest writing settings from them | Legitimate interests, Art. 6(1)(f) | Page text 90 days, page records 365 days after the last citation; what a brief, an article or a writing suggestion keeps stays with it (section 14) |
| We send you emails | Email address, name, email content; hashes of your address and of the requesting network | Sign-in, invitations, reports and alerts | Contract, Art. 6(1)(b); legitimate interests, Art. 6(1)(f) | Our email delivery provider keeps sent emails for 30 days; our sending records 2 days |
| You pay for a plan | Paddle customer and subscription IDs, plan status | Manage plans | Contract, Art. 6(1)(b); legal obligations, Art. 6(1)(c) | Until the workspace is deleted; Paddle keeps invoices as tax law requires |
| Something goes wrong in the app or on our website | User ID and workspace ID (app only), technical error details | Find and fix errors | Legitimate interests, Art. 6(1)(f) | Up to 30 days |
| You create a workspace | User ID, one-way hash of your email, hash of your IP address, website domain | Prevent abuse of new workspaces and offers | Legitimate interests, Art. 6(1)(f) | Kept after deletion (see section 10) |

The sections below explain each situation in detail.

## 3. When you visit our website

> **In brief:** We measure visits to our website and use of our app with DataFast analytics, and use the Reddit Pixel and the Meta Pixel to attribute eligible sign-ups, trial starts and subscriptions to Reddit, Facebook and Instagram ads. These services may set cookies or similar identifiers. Our hosting provider sees the technical data every website visit sends.

### 3.1 Hosting and server logs

When you open a page of our website, your browser sends technical data: your IP address, browser and device details, the page requested, the referring page and the time. Our hosting provider processes this data to deliver the page and may keep it in server logs to keep the website secure and fix problems.

- **Legal basis:** our legitimate interest in running a secure, working website (Art. 6(1)(f) GDPR).
- **Retention:** up to 30 days, after which the logs expire automatically.

### 3.2 Analytics (DataFast)

We use DataFast to understand how visitors find and use our website and app: pages viewed, the referring website and campaign, country, browser, device type, and actions such as clicking "Sign up", submitting the contact form, creating a workspace or starting a checkout. DataFast sets cookies with a random visitor ID and session ID. After you sign in, we link that analytics visitor to your internal Probeli user ID so the website visit and app activity remain one journey. We also use the Reddit Pixel on the marketing site and app to send `PageVisit`, `SignUp` when you create your first workspace, `Lead` when a free trial starts with a card, and `Purchase` after a confirmed paid subscription. Each `SignUp`, `Lead` and `Purchase` event carries an internal ID of that sign-up, trial or payment so it is counted only once; the ID contains no name, email address or website. Reddit uses its own identifiers and attribution signals to match those events to Reddit ad views or clicks. We also use the Meta Pixel on the marketing site and app to send Meta (the company behind Facebook and Instagram) `PageView` when a page loads, `CompleteRegistration` when you create your first workspace, `StartTrial` when a free trial starts with a card, and `Purchase` after a confirmed paid subscription, each conversion with the same internal ID. The Meta Pixel sets its own cookies (section 11) and Meta uses them and its own signals to match those events to Facebook and Instagram ad views or clicks; we turn off its automatic collection of page details and its automatic matching of form fields, so it never reads what you type. When you start a paid subscription, directly or when a trial ends, we send Reddit that `Purchase` from our server with the Reddit ad click ID you arrived with, a hashed IP address and your browser's user agent, which we keep for at most 28 days. When your subscription changes, our server also tells DataFast about it: when a trial starts, a payment goes through or fails, you move to a bigger plan, or the subscription ends. Each of these events goes to the analytics visitor ID stored with your checkout. It carries only the plan, the billing period, the amount and currency, the stage (for example, the end of a trial or a renewal), and internal IDs of your workspace and subscription. We keep the campaign parameters of the link you arrived with (its source, medium and campaign names) with your account for as long as it exists. We do not send DataFast, Reddit or Meta your name, email address, website addresses you type, prompts or answers. When you first open our website, a short notice tells you about this.

- **Legal basis:** our legitimate interest in understanding and improving our website and product (Art. 6(1)(f) GDPR).
- **Retention:** as long as the DataFast account keeps it; the cookies expire as listed in section 11.

We don't use social media plugins. Fonts and images are served from our own servers, so opening our website doesn't send your data to font or image services.

### 3.3 Contact form

When you use our contact form, we receive your name, email address, the topic and your message, and, if you add them, your company and website. The form sends your message by email, through our email delivery provider, to our team's inbox, [hi@probeli.ai](mailto:hi@probeli.ai), which our email hosting provider hosts and which we reply from. We don't store your message in our database; it stays in that inbox. The same applies when you write to [hi@probeli.ai](mailto:hi@probeli.ai) directly. To prevent spam, we keep your IP address (for IPv6, only your network's part of it) for at most 24 hours after the last message you sent, to limit how many messages can be sent, and then delete it.

- **Purpose:** to answer your message.
- **Legal basis:** our legitimate interest in answering inquiries (Art. 6(1)(f) GDPR) or, if you ask about a plan or your account, steps before or under a contract (Art. 6(1)(b) GDPR).
- **Retention:** as long as needed to handle your request and any follow-up, unless the law requires us to keep it longer.

### 3.4 Error reports

When an error happens on our website, on our servers or in your browser, we send a technical report to our error monitoring provider, which stores it in the EU: the error message and where it happened, the page address with personal values such as email addresses removed, and your browser and operating system. It contains no IP address, no cookies and nothing you typed into the contact form. Error reporting on our website sets no cookies and stores nothing in your browser, and we don't record sessions or screens.

- **Legal basis:** our legitimate interest in running a working website (Art. 6(1)(f) GDPR).
- **Retention:** up to 30 days.

### 3.5 Free site audit

When you run our [free site audit](https://probeli.ai/free-ai-site-audit), you give us a website address, an article language and your email address. A bot check (section 12) confirms you are a person. Our crawler then reads that website's homepage and sitemap (section 6), and our data and AI providers (section 12) help us analyse it.

We store your email address, the website's domain, the language, the results and when you agreed to our emails. We email you the audit, a free sample article for that website and one short follow-up, unless you sign up first. We don't add you to a newsletter, and every email has a one-click unsubscribe link. If you unsubscribe, or an email bounces or is marked as spam, we keep a one-way hash of your address so we don't email you again. If you later create an account with the same email address, we link the audit to it, so we can see which audits lead to customers. To prevent abuse, we keep short-lived hashed records of your network and email address.

- **Legal basis:** your consent (Art. 6(1)(a) GDPR) for these emails, which you can withdraw at any time with the unsubscribe link; our legitimate interest in running the audit, preventing abuse and seeing which audits lead to customers (Art. 6(1)(f) GDPR) for the rest.
- **Retention:** an audit that isn't linked to an account is deleted one year after it was run, with its sample article. Abuse records are deleted after 3 days. Hashes of unsubscribed addresses have no deletion date.

## 4. When you create an account and sign in

> **In brief:** We store what you need to sign in and keep your account safe, including the IP address and browser of each session, which you can see in your security settings.

### 4.1 What we process

- **Account details:** your name, email address, whether your email is verified, your language and, if you add one, a profile picture.
- **How you heard about us:** when you set up your first workspace, we ask where you heard about Probeli AI and keep your answer (one choice from a list and, if you add one, a short note such as a newsletter's name) with your account to learn which channels bring new customers, under our legitimate interest (Art. 6(1)(f) GDPR); our analytics (section 3.2) receive the choice but never your note.
- **Password:** if you use one, we store only a secure hash, never the password itself.
- **Sign-in and verification links:** one-time sign-in (magic) links, email verification and password-reset links, which expire after a short time.
- **Passkeys:** the public key, credential ID and device type your device creates. Your fingerprint or face data never leaves your device.
- **Two-factor authentication:** the secret for your authenticator app and your backup codes, if you turn it on.
- **Sign-in with Google** (where offered): your name, email address, profile picture and account ID from Google, and the access tokens it issues, which we store to keep the connection working. Your browser, and the browsers of your team members, load these profile pictures directly from Google, which then receives the viewer's IP address and browser details.
- **Sessions:** for each sign-in, the IP address, browser (user agent), time and expiry. You can see and end your active sessions in your account's security settings.
- **Bot protection:** our sign-up, sign-in and email-link forms use a bot-protection service, which receives your IP address and browser and device details to check that you are not a bot.
- **Security limits:** we count sign-in attempts per IP address for a short time to slow down attacks. We also keep short-lived records of the account emails we send, to limit email abuse (section 10.2).

### 4.2 Why and on what legal basis

We use this data to create your account, let you sign in and keep your account secure. The legal basis is our contract with you or your organization (Art. 6(1)(b) GDPR) and our legitimate interest in preventing unauthorized access and abuse (Art. 6(1)(f) GDPR).

When you choose to sign in with Google, Google processes your data as an independent controller under its own privacy policy. Our use of information received from Google APIs follows the [Google API Services User Data Policy](https://developers.google.com/terms/api-services-user-data-policy), including the Limited Use requirements (section 5.8).

### 4.3 Who helps us

Our hosting and database providers, our email delivery provider (sign-in and verification emails), our file storage provider (profile pictures) and our bot-protection service. See section 12.

### 4.4 How long we keep it

Until you delete your account. Sessions end 30 days after you last use them, or when you sign out. Sign-in, verification and reset links expire after a short time. Profile pictures are not yet deleted from our file storage automatically when you delete your account: ask us at [hi@probeli.ai](mailto:hi@probeli.ai) and we will delete them.

## 5. When you use Probeli AI

> **In brief:** Your workspaces hold business data: your website, brands, competitors, prompts and the AI answers to them, and the articles you generate. Your prompts go to our data provider, which asks the AI engines. Don't put personal data into prompts.

### 5.1 What we store in a workspace

- your website address, its homepage title and description, and the workspace's market (its language, country and, if you set one, city);
- your website's business brief (what the business does, its audience, offers and prices, what sets it apart, its tone and what not to say), with the addresses and titles of the pages it was written from, the addresses of those it couldn't read, and which team member generated or last edited it;
- your brand's name, domains and aliases, and those of the competitors you track;
- your prompts, including archived ones and the wording each answer was asked with;
- the AI answers: their full text, the sources they cite, the searches the engine ran and the brands they name;
- the results of our analysis: mentions, position, sentiment, citations and Insights (with their status);
- AI site audit results and which team member started each audit;
- your articles: their keyword and the topic they are written for, the brief and your edits to it, the text, its images (with the credit of a stock photo or of the page an image comes from; sections 5.2 and 6.1), the research they were written from (section 6.1), and which team member started each one and approved its brief;
- refresh suggestions for one of your own pages (section 6.1): the target keyword, each suggestion's text, the short excerpt of your page it would replace (a title, a meta description, a section's text up to 1,500 characters or a phrase up to 200 characters), whether you marked it done or dismissed and which team member did, and which team member asked for them;
- for an article you mark as published: the page's address, and the result, HTTP status code and time of the latest check that the article is on that page (section 6.1);
- writing suggestions: words and short phrases (at most six words) that you removed or changed in your published articles, in Probeli AI's editor or on your page, and style habits measured from those changes (such as sentence length or punctuation), which articles they came from, and whether you added or dismissed each (section 6.1);
- the Google data you connect (section 5.8);
- the workspace's name and logo, its members, their roles and invitations;
- when a member last opened the workspace, which we use to pause inactive trials;
- your notifications and notification settings.

Most of this is business information. It contains personal data where you enter some (for example a prompt about a named person) or where an AI answer mentions a person.

**Legal basis:** our contract with you or your organization (Art. 6(1)(b) GDPR).

### 5.2 How your prompts and content are processed

To collect answers, we send your prompts, together with the workspace's location and language settings, to our AI answer data provider, which is based in the EU. It submits them to the AI engines we measure, which today are ChatGPT, Gemini, Google AI Mode and Perplexity, and returns their answers to us. These engines are operated by OpenAI, Google and Perplexity. We don't send your name, email address or other account details with your prompts.

To suggest competitors and prompts and to write your website's business brief, we send AI model providers your brand name and domain, your homepage title and description, competitor names, existing prompts, your business brief, the workspace's market and, for the brief, the language to write it in and excerpts of the pages ProbeliAIBot read on your website (section 6.1).

To write an article, we get the search results for its keyword, language and location from a search-results data provider and keep them with the article's research (section 6.1). We then send AI model providers what the article needs: its keyword and market, your brand name and domain, the pages of your website it may link to, competitor names, your business brief, the related keywords and questions of its topic, those search results (which are text from other websites), the text of the pages ProbeliAIBot read for the article (section 6.1), the article's brief and its draft, so that their models write and check the article. When YouTube videos are on, or when you check a video or search YouTube in the article editor, we send the article's keyword or your search words (or the video's link), with the article's language and location, to the search-results data provider, and keep the chosen video's public details (such as its title, channel, length and thumbnail) with the article.

To keep a published article's facts current, we send an AI model provider our own sentences that may be out of date and, for each, a short passage of the page it cites as that page reads today (section 6.1), so that its model checks and corrects them. We keep our corrected sentences and the new values in the article's history, not the passages.

When you ask for refresh suggestions for one of your own pages, we send that page's article text, title and description, the target keyword and what its report found weak to an AI model provider, whose model suggests changes (section 6.1).

To illustrate an article, we send an AI image generation provider a short image description (or the description you type when you generate an image yourself) and your workspace's image style. When you choose a sample picture for your covers, we also send a small copy of it, so the cover is drawn in its look. The images it makes carry an invisible watermark that marks them as made by AI. For an image inside the article, we may search a stock photo library with a few words about its section, and we keep the photographer's name and the links to their page and the photo with the image. Images you upload to your image library are kept as a copy without their location and camera details. We store the images in our file storage (section 12).

The AI model providers that write and check our articles, write refresh suggestions and make images don't use what we send them to train or improve their models. We don't send them your name, email address or other account details.

We have no contract with the operators of the AI engines. They process prompts under their own terms, and we cannot control whether they store or use them. This is why you should not put personal data or confidential information into prompts, brand names or aliases.

### 5.3 How answers are analyzed

Our analysis is rule-based. We match brand names, aliases and domains, record the order in which tracked brands appear, and calculate a word-based sentiment score. It concerns brands, not people, and we make no automated decisions about people (see section 20). We do not use your data to train AI models.

### 5.4 Website icons from Google

To show the icon of each website next to brands, competitors and sources, your browser loads these icons from Google's favicon service. Google then receives your IP address, browser details and the domain whose icon is shown, and may use its own cookies under Google's privacy policy.

**Legal basis:** our legitimate interest in an interface where websites are easy to recognize (Art. 6(1)(f) GDPR).

### 5.5 Team invitations

When a workspace owner or admin invites someone, we store the invitee's email address, the role, who sent the invitation and its status, and we send the invitation by email through our email delivery provider. This also applies to people who don't have a Probeli AI account yet. The invitation record stays with the workspace until the workspace, or the account of the person who sent it, is deleted. An invitation that couldn't be emailed is deleted at once, and an expired one is deleted once the same person is emailed a new invitation to that workspace.

**Legal basis:** the legitimate interest of the inviting customer and our own in letting teams work together (Art. 6(1)(f) GDPR).

### 5.6 Access by our staff

Authorized Probeli AI staff can view accounts and workspaces and can sign in as a user to give support, investigate problems, and prevent abuse or security incidents. When staff sign in as a user, this is recorded on the session. Staff can also block accounts that break our [Terms of Service](https://probeli.ai/legal/terms).

**Legal basis:** our contract (support) and our legitimate interest in running a secure service (Art. 6(1)(b) and (f) GDPR).

### 5.7 How long we keep workspace data

- **Full text of AI answers:** kept for the period that applies to the workspace's plan, shown on our [pricing page](https://probeli.ai/pricing) as "Full answer text kept". During a free trial, the period of our lowest-priced plan applies. The text is then deleted, oldest first. While a payment has failed, we keep it for our longest period until the payment issue is resolved.
- **Everything else in a workspace** (including mentions, citations and metrics): kept until the workspace is deleted, so that your history stays complete. Changing a workspace's website deletes the AI answers, competitors, prompts, Insights, audit results and business brief of the previous website.
- **Workspaces whose trial or plan has ended:** we don't currently delete these automatically. Their data stays until someone deletes the workspace.

### 5.8 Google data you connect (Search Console and Google Analytics)

An owner or admin of a workspace can connect a Google account to it, read-only, in Settings › Website. Google then asks you to allow two kinds of access: Search Console (read-only) and Google Analytics (read-only). We never ask for access to anything else in your Google account, and you can allow just one of the two.

That connection stays read-only unless you choose to add one thing. An owner or admin can separately allow Probeli AI to submit a sitemap of your own site to your Search Console property. That is a second Google permission, asked for on its own screen, and nothing happens until you allow it and pick the sitemap. Google's screen will say the permission covers managing your Search Console sites and sitemaps; we use it for exactly one thing: telling Search Console that the sitemap address you picked has changed, after we publish a page for you and when you press "Submit now". We never submit a sitemap on another host, never change or delete anything else in Search Console, and never ask Google to index a page — Google has no API for that. You can stop it at any time by turning it off in Settings › Website, and you can remove the permission itself in your Google account. We keep, for each attempt, which property and sitemap address it was for, what asked for it, and what Google answered.

- **What we read and keep:** the names of the Search Console and Google Analytics 4 properties you pick; each day's search clicks, impressions and average positions for your site, its pages and its search queries; the sessions that AI assistants such as ChatGPT, Perplexity, Gemini, Copilot and Claude sent to your pages, per day and landing page; and, for each article you publish, whether Google has indexed it and when Google last crawled it. We store the access token Google gives us encrypted, and never the password of your Google account. Google also tells us your Google account's ID (no email address or profile); we keep only a one-way hash of it.
- **Why:** to show your own Google numbers in your workspace (the Traffic page, an article's Performance tab), to flag your own articles whose Google clicks or rank decline, and — only if you allowed it — to submit the sitemap you picked. Nothing else.
- **Limited Use:** Probeli's use and transfer of information received from Google APIs adheres to the [Google API Services User Data Policy](https://developers.google.com/terms/api-services-user-data-policy), including the Limited Use requirements. We don't sell this data, use it for advertising, use it to train or improve any AI model, or share it with anyone except the service providers that host our database (section 12). Our staff don't read it, except with your consent for support, for security, or to comply with the law.
- **Who it comes from:** Google provides the data under its own terms and privacy policy, as an independent controller for your Google account.
- **How long:** synced numbers are kept for 16 months, Search Console's own history. When you disconnect, we revoke our access at Google and delete every synced number of the workspace at once. Deleting the workspace does the same. If Google stops accepting the connection (for example, you removed it in your Google account), we stop syncing and delete the synced numbers after 30 days unless someone connects again. If the person who connected Google leaves the workspace or stops being an owner or admin, we revoke their connection and delete its data. Changing the workspace's website deletes the synced numbers of the previous website.
- **Your control:** you can disconnect at any time in Settings › Website, or remove Probeli's access in your Google account's security settings.

### 5.9 Support chat (Botpress)

The app has a "Chat with us" button. Nothing from our chat provider, Botpress, loads until you click it. When you click it, your browser loads Botpress's chat, and an AI assistant answers your questions about Probeli AI.

- **What Botpress receives:** what you type in the chat, the address of the page you opened it from, your IP address and browser details, and a random chat ID. The ID is kept in your browser's storage (section 11), so a conversation can continue when you come back.
- **What we don't send:** your name, email address, workspace or plan. If you type your email address so we can reply, Botpress processes it as part of the conversation. Don't send passwords or payment details in the chat.
- **Legal basis:** our legitimate interest in answering questions about our product (Art. 6(1)(f) GDPR).
- **Where:** Botpress processes the chat on our behalf in the United States (section 13).
- **How long:** until we delete the conversation or close our account with Botpress.

## 6. Public web pages our crawler reads (ProbeliAIBot)

> **In brief:** Our crawler, ProbeliAIBot, reads public web pages that AI answers cite, your own website during audits and for its business brief, the public pages an article is researched from, and the page where you published an article. Those pages can contain personal data. It follows robots.txt, except for the requests listed on our [ProbeliAIBot page](https://probeli.ai/bot).

### 6.1 What we fetch and store

Probeli AI fetches public web pages with its own crawler, which identifies itself as ProbeliAIBot:

- **Homepages:** yours, when you create a workspace, change its website or ask to read it again, and a competitor's, when you add one, to read their title and description. From yours, we also use its language tags with that title and description to propose the workspace's language and country, and check whether its text shows without JavaScript. We keep the outcome of the last read of yours (when, whether it could be read and why not), not its text.
- **Pages that AI answers cite,** to find indirect mentions of tracked brands. We store the page text (up to a size limit), its title and the domains it links to, in one cache shared by all workspaces.
- **A website someone checks with the free site audit:** its homepage and one sitemap file (section 3.5). We keep the homepage's title and description, a few page checks and up to 10 addresses from the sitemap with the audit, not the page text.
- **Your workspace's website during an AI site audit.** The audit reads a limited number of pages and files such as robots.txt and sitemaps. It keeps only the results and the score, not the page text.
- **Key pages of your workspace's website, for its business brief:** its homepage and a few pages it links to or its sitemap lists (such as about, products or pricing). Excerpts of their text go to an AI model provider, which writes the brief (section 5.2). We keep the brief, which quotes short passages of these pages, the address and title of each page read and the address of each page it couldn't read, not the full page text.
- **Public pages an article is researched from, when you generate an article:** a limited number of pages, such as the top search results for the article's keyword and a few related searches, the official pages, studies and product pages those results point to, and a few pages of your own website on the article's subject. Their text is used to write the article (section 5.2) and is not kept, except for the few sentences the article's facts come from (at most 8 of each page, kept with the article). We keep each page's address, title, headings, word count and dates, the address of each result it didn't read, short quotes from the pages with their link and a one-sentence summary of each, as the article's sources. The article's research also keeps the search results for the keyword (each result's address, title and snippet, and the questions, answers and related searches shown with them), which can contain personal data like any public page. We keep all of this with the article until the article or its workspace is deleted. When an attempt to write an article fails or is cancelled, or its brief isn't approved within 14 days of being written or last edited, we delete its research 7 days after the attempt ended.
- **Your workspace's sitemaps and its RSS or Atom feed, when you generate an article,** to see whether one of your pages already covers the article's keyword. We keep the address and title of each matching page with the article's brief and the addresses of a few of your pages on the article's subject, not the list. A matching post found in your sitemap is also listed in your workspace's pages (its address and the sitemap that lists it) so you can update it instead, until a scan of your sitemap no longer finds it.
- **Your own posts and pages, when you scan them or ask for refresh suggestions.** A scan reads the pages your sitemap or your Search Console lists, and we keep each page's address, title, headings and counts for its report, not its text. When you ask for refresh suggestions for one page, we read it once more and send the article part of its text, its title and description, the target keyword and what its report found weak to an AI model provider, whose model suggests changes (section 5.2). We keep the suggestions and short excerpts of the parts they would change (section 5.1), never the whole page. When you read the page again, we compare it with the open suggestions to mark the ones you applied, and keep only those marks.
- **Images for an article.** When an article lists products or tools, we read the start of each listed item's own page (an address the article already cites or links, up to 10 per article) to find its preview image, copy that image into the article and credit and link the page under it. A rewrite draft of one of your own pages made before refresh suggestions replaced it may also hold copies of the images that page showed. We keep each copy (resized, without the file's metadata) with the article until the article or its workspace is deleted, and not the page's text.
- **The pages a published article cites, about once a month,** to check that its prices, versions and dates still hold (a workspace owner or admin can turn this off for the workspace or an article): up to 8 of them per article. We keep, for each fact, the page's address and a fingerprint (a hash) of its text to see whether it changed, never the page text. When a fact may have changed, a short passage of the page (about 300 characters) goes to an AI model provider with our sentence (section 5.2).
- **The page where you published an article,** when you mark the article as published or ask to check it again, to confirm the article is there. Each check loads that one page and at most 5 redirects within the website the article was written for, and reads that website's robots.txt when it has no recent copy. When the check finds the article, we also compare the page's text with the article we gave you, without storing the page, to suggest changes to your writing settings. We keep the result of the check, the page's HTTP status code, the time and those suggestions (words or short phrases removed from our text, style habits such as sentence length and, for a changed wording, at most three words of the page's replacement), never the page text. A workspace owner or admin can turn this off in the workspace's writing settings.

### 6.2 Personal data on public pages

Public pages can contain personal data, such as an author's name or a team page. We don't look for or build profiles of individuals: we use these pages only for what section 6.1 describes. For a business brief, excerpts of the key pages of a workspace's website, and for an article, the text of the pages it is researched from, go to AI model providers, which write the brief or the article (sections 5.2 and 12).

**Legal basis:** our legitimate interest, and our customers', in understanding which public sources AI engines rely on and how they mention brands (Art. 6(1)(f) GDPR). We only process information that was published openly.

### 6.3 How long we keep it

The stored text of a page is deleted after 90 days in which no AI answer cited it. The rest of the page record (address, title and linked domains) is deleted after 365 days without a citation. What an article's research keeps stays with the article; the research of an attempt that failed, was cancelled or expired is deleted 7 days after the attempt ended (section 6.1). The few words of a published page kept in a writing suggestion stay with it (section 14).

### 6.4 How to block ProbeliAIBot or object

ProbeliAIBot follows the rules in robots.txt, except for the two requests listed on our [ProbeliAIBot page](https://probeli.ai/bot): robots.txt itself, and the homepage read of a workspace's own website when its customer enters it or asks to read it again. Sitemaps and every redirect it follows are subject to robots.txt. It limits how often it requests pages from each site. The check of a published article runs only when a customer asks for it, loads that one page and a few redirects within the same website, and runs within a limit per workspace; comparing the page for writing suggestions sends no extra request, and blocking ProbeliAIBot stops it too. To block it on your website, add this to your robots.txt:

```
User-agent: probeliaibot
Disallow: /
```

Our [ProbeliAIBot page](https://probeli.ai/bot) explains what it reads, how often, and which requests robots.txt doesn't cover. You can also object to this processing or ask us to delete a page from our cache by writing to [hi@probeli.ai](mailto:hi@probeli.ai).

## 7. Emails we send

> **In brief:** We only send emails about your account and your workspaces, and announcements about Probeli AI. There is no newsletter. Reports and alerts can be switched off in your settings.

### 7.1 Service emails

These emails are part of the service, so you can't switch them off while you have an account:

- email verification and confirmation of a change of email address;
- password reset links and sign-in (magic) links;
- workspace invitations;
- a welcome email when you sign up.

### 7.2 Workspace reports and notices

These are on by default. You can switch each type off, for email and in the app, in your account's notification settings:

- the weekly visibility report, sent to workspace owners and admins;
- visibility-drop alerts, sent to workspace owners and admins (at most one every 7 days per workspace);
- a notice when daily tracking pauses, sent to workspace owners and admins;
- announcements about changes to Probeli AI;
- a short note from our founder that asks one question: a few days into a free trial (why you signed up), a week after a trial ends without a plan or a day after you cancel a plan (what got in the way), and a month into a paid plan and every few months after (what is missing). It contains no offer, your reply goes to the founder's own inbox, and you can switch it off with the link in it or in your notification settings.

### 7.3 Free site audit emails, and no newsletter

When you run the free site audit and agree to our emails, we send you the audit, one free sample article when it's ready and one short follow-up a few hours later (section 3.5). Each has a link to unsubscribe in one click. We don't send newsletters or other marketing emails. If we start, we will update this policy and ask for your consent where the law requires it.

- **Legal basis:** our contract (Art. 6(1)(b) GDPR) for service emails, reports and alerts; our legitimate interest in telling customers about changes to the service (Art. 6(1)(f) GDPR) for announcements, and in learning why people sign up, why a trial or a plan didn't work out and what is missing (Art. 6(1)(f) GDPR) for the founder's note.
- **Who helps us:** our email delivery provider sends our emails and keeps them for 30 days, in the United States (see sections 12 and 13).

## 8. Payments

> **In brief:** Paddle sells our plans as Merchant of Record and handles your payment details. We only receive references to your subscription.

When you buy a plan, you pay through the checkout of Paddle (Paddle.com Market Limited), our reseller and Merchant of Record. The checkout opens on our checkout page, which loads Paddle's checkout script from Paddle, and runs in Paddle's own window. Paddle collects your name, email address, billing address, payment details and, if you give one, your VAT ID. Paddle processes this data as an independent controller under its own privacy policy, together with the payment processors it works with. Paddle also runs the billing portal that opens when you click "Manage billing".

We send Paddle only the plan chosen, the workspace's ID and, if the workspace has bought before, its Paddle customer ID. From Paddle we receive the customer ID, subscription ID, the plan and the subscription status, which we store to manage the workspace's plan. We never receive your full card details.

- **Legal basis:** our contract (Art. 6(1)(b) GDPR) and legal obligations such as tax and accounting rules (Art. 6(1)(c) GDPR).
- **Retention:** our plan records are deleted together with the workspace. Paddle keeps invoices and transaction records for as long as tax law requires.

## 9. Error monitoring and logs

> **In brief:** When something breaks in the app, we receive a technical report with IDs, not names or email addresses.

### 9.1 Error reports

When an error happens in the Probeli AI app or on our website, on our servers or in your browser, we send a technical report to our error monitoring provider, which stores it in the EU. A report contains the error message and where it happened, the page address, browser and operating system, and your user ID and workspace ID. Before a report is sent, we remove names, email addresses, IP addresses, cookies, sign-in tokens, request contents, and the text of prompts and answers. We don't record sessions or screens.

### 9.2 Server logs

Our servers write technical logs, which contain IDs and error messages and occasionally an email address, for example in a warning about a failed email.

- **Legal basis:** our legitimate interest in running a reliable, secure service (Art. 6(1)(f) GDPR).
- **Retention:** error reports and server logs are kept for up to 30 days, after which they expire automatically.

## 10. Preventing fraud and abuse

> **In brief:** To stop abuse of new workspaces, free trials and our money-back guarantee, we keep a small record of each new workspace, even after an account is deleted.

### 10.1 Sign-up records

When a workspace is created, or an unpaid workspace changes its website, we record your user ID, one-way hashes of your email address and IP address, the website's domain, the workspace and the time. We use these records to limit abuse of new workspaces, free trials where they are offered, and refunds. So that deleting an account doesn't reset these limits, the records are **not** deleted when you delete your account or workspace. They currently have no fixed deletion date.

- **Legal basis:** our legitimate interest in preventing abuse of our service and offers (Art. 6(1)(f) GDPR). You can object (see section 16).

### 10.2 Other measures

- We count requests per IP address for a short time to limit sign-in attempts and contact-form messages (section 3.3).
- For each sign-in, verification, invitation or welcome email we send, we record a one-way hash of the recipient's address and a one-way hash of the requesting network, to limit email abuse. These records are deleted after 2 days.
- We keep a record of the cost of each run per AI engine. It contains no personal data and is kept after a workspace is deleted.
- We may block accounts that break our [Terms of Service](https://probeli.ai/legal/terms).

## 11. Cookies and similar technologies

> **In brief:** We use cookies and browser storage that the website and app need to work, that remember your settings, DataFast's analytics cookies and the Meta Pixel's advertising cookies. When you arrive from a Reddit ad, one short-lived app cookie keeps the ad click ID until you create your account (section 3.2).

Most cookies and browser storage we use are strictly necessary to provide the service you ask for, or remember choices you make in the interface. The DataFast cookies measure use of the website and app, and the Meta Pixel's cookies attribute sign-ups, trials and subscriptions to Meta ads (section 3.2).

### 11.1 Our website

| Name | Purpose | Lifetime | Type |
| --- | --- | --- | --- |
| `datafast_visitor_id` | Recognizes a returning visitor (random ID) | 365 days | Analytics |
| `datafast_session_id` | Groups the pages of one visit (random ID) | 30 minutes | Analytics |
| `_fbp` | Meta Pixel: recognizes a returning browser | 90 days | Advertising measurement |
| `_fbc` | Meta Pixel: the Meta ad click you arrived from, if any | 90 days | Advertising measurement |
| `probeli-privacy-notice` (local storage) | Shows the privacy notice only once | Until you clear it | Functional |

### 11.2 The Probeli AI app

| Name | Purpose | Lifetime | Type |
| --- | --- | --- | --- |
| `better-auth.session_token` | Keeps you signed in | 30 days | Necessary |
| `better-auth.session_data` | Short-lived, encrypted copy of your session | 60 seconds | Necessary |
| `better-auth.dont_remember` | Ends your session when you close the browser, if you chose not to be remembered | Browser session | Necessary |
| `better-auth.two_factor` | Holds a pending two-factor check | A few minutes | Necessary |
| `better-auth.trust_device` | Skips two-factor checks on a device you trust | 30 days | Necessary |
| `better-auth.state`, `better-auth.oauth_state` | Protects sign-in with Google | A few minutes | Necessary |
| `better-auth-passkey` | Holds a pending passkey check | A few minutes | Necessary |
| `probeli_website` | Carries the website you typed on our website into setup | 1 hour | Necessary |
| `probeli_acquisition` | Keeps the Reddit ad click ID and campaign names you arrived with until you create your account | 1 day | Advertising measurement |
| `better-auth.admin_session` | Keeps a staff member's own session while they help a user | Browser session | Necessary (staff only) |
| `better-auth.last_used_login_method` | Highlights the sign-in method you used last | 30 days | Functional |
| `NEXT_TIME_ZONE` | Your time zone, to show dates correctly | 365 days | Functional |
| `sidebar_state` | Whether the sidebar is open | 7 days | Functional |
| `sidebar_groups` | Which navigation groups are collapsed | 365 days | Functional |
| `mode` | Normal or Expert: how much the pages explain | 365 days | Functional |
| `cards_closed` | Which cards on the Overview you collapsed | 365 days | Functional |
| `insights-view.<workspace ID>` | List or board view on the Insights page | 365 days | Functional |
| `theme` (local storage) | Light, dark or system theme | Until you clear it | Functional |
| `probeli.publishPlatform.<workspace ID>` (local storage) | Where you publish a workspace's articles, to show its tips | Until you clear it | Functional |
| `bp-webchat-<chat ID>-client` (local storage) | Support chat (section 5.9): a random ID that continues your conversation. Set only after you open the chat | Until you clear it | Functional |
| `probeli:chunk-reload-at` (session storage) | Reloads the page once after an update | Browser tab session | Necessary |
| `datafast_visitor_id`, `datafast_session_id` | Analytics (section 3.2), random IDs | 365 days, 30 minutes | Analytics |
| `_fbp`, `_fbc` | Meta Pixel (section 3.2): a returning browser and the Meta ad click you arrived from | 90 days | Advertising measurement |

On secure connections, some names start with `__Secure-`. Loading website icons from Google (section 5.4) is a request to Google, not a cookie we set, but Google may use its own cookies there. The same goes for the checkout (section 8): Paddle's checkout window may use Paddle's own cookies and storage to process the payment and prevent fraud.

### 11.3 Managing cookies

You can delete cookies and browser storage in your browser settings at any time. If you block the necessary ones, you won't be able to sign in.

## 12. Recipients and sub-processors

> **In brief:** A few kinds of service providers help us run Probeli AI. They only get the data they need for their task.

We use these kinds of service providers:

| Kind of provider | What they do for us | Data involved | Where |
| --- | --- | --- | --- |
| Hosting and database | Runs our app, its background jobs and our database; server logs | All account and workspace data, server logs | EU |
| Website hosting, content delivery and file storage | Delivers our website and stores profile pictures, workspace logos and article images | Website requests (IP address, browser details, pages requested), stored files | Global network of a US company (Data Privacy Framework) |
| Bot protection | Checks sign-in and sign-up forms, the free SEO tools and the free site audit for bots | IP address, browser and device signals | Global network of a US company (Data Privacy Framework) |
| Website and app analytics (DataFast) | Measures use of the website and app (section 3.2) | Pages viewed, referrer, country, browser and device, random visitor ID, internal user ID after sign-in, product actions, subscription events sent from our server | Can be outside the EU (section 13) |
| Advertising measurement (Reddit) | Attributes sign-ups, trials and subscriptions to Reddit ads (section 3.2) | PageVisit, first-workspace SignUp, card-trial Lead and paid-subscription Purchase events, each conversion with an internal ID of the sign-up, trial or payment; attribution identifiers; for a server-sent Purchase, the ad click ID, a hashed IP address and the browser's user agent (kept at most 28 days) | United States |
| Advertising measurement (Meta) | Attributes sign-ups, trials and subscriptions to Facebook and Instagram ads (section 3.2) | PageView, first-workspace CompleteRegistration, card-trial StartTrial and paid-subscription Purchase events, each conversion with an internal ID of the sign-up, trial or payment; the page address; the Meta Pixel's cookies and attribution identifiers | Ireland and the United States |
| AI answer and search-results data | Asks the AI engines your prompts and returns their answers; returns search and YouTube results for an article's keyword, and a website's search numbers for the free site audit | Prompts, location and language settings, article keywords, YouTube search words and video links, the domain checked with the free site audit | EU |
| AI model providers | Write competitor and prompt suggestions, business briefs, articles and refresh suggestions, check articles (section 5.2), describe a homepage and write a sample article for the free site audit (section 3.5) | Brand and competitor names, domain, homepage title and description, prompts, the business brief, market, article keywords and search results, the text of pages read for a brief or an article, article briefs and drafts, the text of your own page when you ask for its refresh suggestions, the homepage text and domain of a website checked with the free site audit | United States |
| AI image generation | Generates article images | The image description we write from the article (topic, brand name), the workspace's image style and, for a cover drawn after a sample picture, a copy of it; no account details | United States (Data Privacy Framework) |
| Stock photo library | Finds stock photos for articles | A few words about an article's section | EU |
| Email delivery | Sends our emails, including contact-form messages to our team | Email addresses, names, email content | United States (Data Privacy Framework) |
| Email hosting | Hosts our team's inbox, [hi@probeli.ai](mailto:hi@probeli.ai), which receives contact-form messages and your emails | Email addresses, names, email content | United States |
| Support chat (Botpress) | Answers questions in the app's chat (section 5.9) | What you type in the chat, the page you opened it from, IP address and browser details, a random chat ID | United States |
| Error monitoring | Error monitoring for the app and the website | User and workspace IDs, technical error details | EU |

The current list of our service providers is available on request at [hi@probeli.ai](mailto:hi@probeli.ai).

These recipients are independent, not our processors:

- **Paddle.com Market Limited** (United Kingdom) and the payment processors it works with, for purchases (section 8).
- **Google,** when you sign in with Google or your browser shows a profile picture from it (section 4), when you connect Search Console or Google Analytics (section 5.8), and for website icons (section 5.4).
- **The operators of the AI engines** (today OpenAI, Google and Perplexity), which receive prompts through our AI answer data provider (section 5.2).

We may also disclose personal data to authorities when the law requires it, to our professional advisers, and to a buyer or successor if our business is sold or merged, in each case only as far as necessary.

## 13. International transfers

> **In brief:** Our core data is stored in the European Union. We are established outside the EU, and some providers are in the United States; we use the legal safeguards the GDPR requires for them.

Our database and app servers are in the European Union, error reports are stored in the EU, and our AI answer and search-results data provider is based in the EU. Probeli AI itself is established outside the European Union (section 1), and we access this data from there to run the service. Some providers are based in the United States, even where they store data elsewhere: our email delivery and email hosting providers, the provider that delivers our website, stores files and runs our bot check on its global network, our AI model and AI image generation providers, our support chat provider, and Reddit and Meta for ad measurement. Our analytics provider may also process data outside the EU. The operators of the AI engines, which receive prompts through our data provider, may process them in the United States and other countries.

Where personal data goes to a country without an adequacy decision of the European Commission, we rely on the EU-US Data Privacy Framework for certified companies, or on the European Commission's Standard Contractual Clauses. Several of our providers are certified under the Data Privacy Framework, and their data processing agreements include the Standard Contractual Clauses. You can ask for a copy of the safeguards at [hi@probeli.ai](mailto:hi@probeli.ai).

## 14. How long we keep data

| Data | How long |
| --- | --- |
| Account details and sign-in methods | Until you delete your account |
| Sessions | 30 days after last use, or until you sign out |
| Sign-in, verification and reset links | A short time, until used or expired |
| Workspace data | Until the workspace is deleted |
| Full text of AI answers | The period of the workspace's plan, shown on our [pricing page](https://probeli.ai/pricing) |
| Google data you connect (section 5.8) | 16 months; deleted at once when you disconnect or delete the workspace |
| Cached public page text | 90 days without a new citation |
| Cached public page records | 365 days without a new citation |
| Article research from an attempt that failed, was cancelled or expired (section 6.1) | 7 days after the attempt ended |
| Refresh suggestions and their page excerpts (section 5.1) | Until the page's next refresh suggestions, until the page leaves your workspace's pages, or until the workspace is deleted |
| Writing suggestions you haven't added or dismissed (section 5.1) | Until the articles they came from are deleted, unpublished or moved to another address, or learning is turned off |
| Writing suggestions you added or dismissed (section 5.1) | Until the workspace is deleted |
| Contact-form messages | As long as needed to handle your request |
| Sign-up records (section 10) | Kept after deletion; no fixed deletion date yet |
| Email-sending records (section 10) | 2 days |
| Free site audits not linked to an account (section 3.5) | One year after the audit, with its sample article |
| Free site audit abuse records and email records (section 3.5) | 3 days |
| Unsubscribed, bounced or complained addresses, as a hash (section 3.5) | No deletion date |
| Profile pictures and workspace logos | Until you ask us to delete them |
| Server logs and error reports | Up to 30 days |
| Database backups | Up to 7 days |
| Plan records | Until the workspace is deleted (Paddle keeps invoices as tax law requires) |

## 15. Deleting your data

> **In brief:** You can delete a workspace or your account yourself. Deleting your account also deletes the workspaces you are the only member of.

- **Delete a workspace** (owners, in the workspace's general settings): locks the workspace at once and, 3 days later, deletes its AI answers, prompts, competitors, metrics, Insights, audit results, business brief, articles, members and invitations and its plan records, and ends its plan. Until then an owner can restore it; after that this can't be undone. It needs a sign-in within the last day: if yours is older, sign out and sign in again first. Its article images are deleted from our file storage shortly after; the workspace logo stays there until you ask us to delete it ([hi@probeli.ai](mailto:hi@probeli.ai)).
- **Change a workspace's website:** deletes the AI answers, competitors, prompts, Insights, audit results and business brief of the previous website.
- **Leave a workspace, or be removed from one:** deletes your membership only.
- **Delete your account** (in your account settings): deletes your profile, your sign-in methods (password, passkeys, two-factor data and a linked Google account), your sessions, memberships, the invitations you sent, your notifications and your notification settings.
  - It also **deletes every workspace you are the only member of** right away (without the 3-day wait), which ends their plans immediately. It cancels any subscription billed to your account rather than to a workspace.
  - A workspace that has other members stays, without you. While you are its only owner, you can't delete your account: make another member an owner, or remove the other members, first.
  - To delete your account, you need to have signed in within the last day: if yours is older, sign out and sign in again first.
  - The sign-up records with the hashes of your email address and IP address remain (section 10).
  - Your profile picture remains in our file storage until you ask us to delete it.
  - Paddle keeps its own records (section 8).
- **Backups:** data you delete stays in our backups for up to 7 days.
- **Get a copy of your data:** Probeli AI doesn't have a self-serve export. Write to [hi@probeli.ai](mailto:hi@probeli.ai) and we will send you a copy.

For anything else, write to [hi@probeli.ai](mailto:hi@probeli.ai).

## 16. Your rights under the GDPR and UK GDPR

> **In brief:** You can ask us what data we hold about you, correct it, delete it, limit or object to its use, and take it with you. You can also complain to a data protection authority.

You have the right to:

- **access** the personal data we hold about you (Art. 15 GDPR);
- have inaccurate data **corrected** (Art. 16). You can change your name and email address yourself in your account settings;
- have your data **deleted** (Art. 17);
- **restrict** how we use your data (Art. 18);
- receive your data in a structured, machine-readable format (**portability**, Art. 20);
- **object** at any time to processing based on our legitimate interests, on grounds relating to your particular situation (Art. 21). This includes our crawler's page cache (section 6) and sign-up records (section 10);
- **withdraw consent** at any time, where we rely on consent;
- **complain** to a data protection authority, in particular in the country where you live or work, or where you believe your rights were infringed (Art. 77).

To use these rights, write to [hi@probeli.ai](mailto:hi@probeli.ai). We may need to confirm your identity first. We answer within one month; for complex requests, this can be extended by two further months, and we will tell you if it is.

If your data is part of a customer's workspace, for example because a prompt or an AI answer mentions you, we may pass your request to that customer, who decides on it, and help them answer it.

## 17. US state privacy rights

> **In brief:** We don't sell your personal information. Our ad pixels may count as sharing for targeted advertising, and your browser's Global Privacy Control signal turns them off.

Depending on where you live in the United States (for example California, Colorado, Connecticut, Virginia, Utah, Texas or Oregon), and as far as these laws apply to us, you may have additional rights:

- to know what personal information we collect, use and disclose, and to get a copy;
- to have it corrected or deleted;
- to opt out of the sale or sharing of personal information, or its use for targeted advertising. We don't sell personal information. The Reddit Pixel and the Meta Pixel (section 3.2) send Reddit and Meta events about visits, sign-ups, trials and subscriptions to measure our ads, which some of these laws treat as sharing for cross-context behavioral advertising. To opt out, turn on Global Privacy Control in your browser: with it on, neither pixel loads on our website or in our app. You can also write to us;
- to limit the use of sensitive personal information. We only use your sign-in details to let you sign in;
- not to be treated differently for using these rights.

**What we collect:** identifiers (name, email address, IP address, account IDs), commercial information (your plan and subscription status), internet activity (sessions, browser details, when you last opened a workspace) and professional information (your workspaces, brands and team). The sources, purposes and recipients are described in sections 3 to 12.

**How to make a request:** write to [hi@probeli.ai](mailto:hi@probeli.ai). You can use an authorized agent, and we may ask them for proof that you gave them permission. If we decline your request, you can appeal by replying to our decision. If your appeal is declined, you can contact your state's attorney general.

## 18. Security

> **In brief:** We protect your data with standard technical measures, but no system is completely secure.

Our measures include:

- encrypted connections (HTTPS) everywhere, and session cookies that scripts can't read and that are only sent over secure connections;
- passwords stored only as secure hashes, verified email addresses, and optional two-factor authentication and passkeys;
- a membership and role check on every request to a workspace;
- access controls on our database;
- checks that stop requests to private and internal network addresses through web addresses that users can enter;
- error reports with personal data and secrets removed;
- access to customer accounts only for authorized staff, with sign-ins as a user recorded;
- encryption of stored data by our hosting providers.

We don't hold security certifications. If you find a security problem, please tell us at [hi@probeli.ai](mailto:hi@probeli.ai).

## 19. Children

Probeli AI is a tool for businesses and is not directed at children. You must be old enough to enter into a contract to use it. We don't knowingly collect personal data from children under 16. If you believe a child has given us personal data, write to [hi@probeli.ai](mailto:hi@probeli.ai) and we will delete it.

## 20. Automated decision-making

We don't make decisions about people based solely on automated processing that have legal or similarly significant effects on them (Art. 22 GDPR), and we don't profile individuals. Our analysis concerns brands, not people.

The only automated checks about you as a person are the sign-up limits in section 10, the limits on sign-in attempts and account emails (section 10.2), and the bot check on our sign-in forms (section 4), free SEO tools and free site audit (section 3.5). The sign-up limits can stop you from creating more unpaid workspaces or starting another free trial, for a while or, for the same website, for good; workspaces with a paid plan don't count towards them. If you think a check stopped you by mistake, contact us.

## 21. Customer content and our role

> **In brief:** For personal data inside a customer's workspace content, the customer decides and we act on its behalf.

Customers decide which websites, brands, competitors and prompts they track. For personal data in that content, and in the AI answers and results we store for it, the customer is the controller and we process the data on its behalf as a processor. Customers who need a data processing agreement under Article 28 GDPR can contact us at [hi@probeli.ai](mailto:hi@probeli.ai).

For everything else in this policy, such as accounts, sign-in, emails, payments, security, our website and our crawler's page cache, we are the controller.

## 22. Changes to this policy

We may update this policy when our service or the law changes. The "Last updated" date at the top shows the current version. If we make material changes, we will tell you by email or in the app before they take effect.

## 23. Contact

Probeli AI operates the website probeli.ai and the app at app.probeli.ai.

- **Privacy contact:** [hi@probeli.ai](mailto:hi@probeli.ai)

For other questions about Probeli AI, you can use our [contact page](https://probeli.ai/contact).
