This policy explains what personal data Artixles collects, why, who else handles it, how long we keep it and what you can do about it. In short: we keep what we need to run your account and answer your questions, we don't sell data, we don't show ads, we don't use tracking cookies, and your questions, uploads and essays are not used to train AI models.
1. Who is responsible
Findwell, Van Vollenhovenstraat 3-213, 3016 BE Rotterdam, the Netherlands (Chamber of Commerce (KvK) 92838812), runs Artixles ("Artixles", "we") and is the controller for the personal data described here. For anything about privacy, email hello@artixles.com. We are a small company and have not appointed a data protection officer; this address reaches the people who run Artixles.
When a business uses the Artixles API to process its own users' questions, that business decides what it sends us. For that content we act as its processor (see section 8).
2. The short version per purpose
| What we do | Data | Legal basis (GDPR) |
|---|---|---|
| Run your account and sign you in | Email address, the name you choose, sign-in records; with Google sign-in, your name and profile picture from Google | Contract (art. 6(1)(b)) |
| Answer your questions and run the features you use | Questions, chats, uploads (as text), essays, projects, papers, notes, settings, "About you" | Contract |
| Billing, invoices and tax records | Plan, usage, purchases, Stripe customer number, invoices | Contract; legal obligation (art. 6(1)(c)): Dutch tax law requires us to keep records for 7 years |
| Service emails (sign-in links, purchase confirmations, alerts you asked for, mentions) | Email address, name, the content of the alert | Contract |
| Short welcome and tips emails in your first two weeks | Email address, sign-up date | Legitimate interest (art. 6(1)(f)): telling new members about the features they signed up for; unsubscribe with one click |
| Preventing abuse of free usage, fraud and chargebacks | Normalised email address, a scrambled (hashed) network address, payment disputes | Legitimate interest: keeping free usage available and the service affordable |
| Security, fixing errors and keeping Artixles running | Error records, technical logs (IP address, browser) | Legitimate interest: a secure, working service |
| Improving Artixles | Usage events, ratings and comments, where you came from when you signed up | Legitimate interest: finding out what works and what doesn't |
| Study-association partnerships | Which association's link or code you used, your plan and what you paid | Legitimate interest: paying the association its agreed share and showing it totals (no names) |
| The verified findings database | Checked claims from answers; the question with email and web addresses removed, not linked to you | Legitimate interest: making answers more accurate and building a reusable record of what papers say, which the API passes on (without questions) |
Where we rely on legitimate interest, we have weighed it against your interests and you can object at any time (section 12). We don't currently rely on consent for any processing; where you switch something on yourself (for example "Email me when it's ready" or following a question), you can switch it off at any time.
3. What we collect, feature by feature
Your account
- Email address (you sign in with a link or a 6-digit code sent to it, or with Google; there are no passwords) and the name you choose.
- Sign in with Google (if you choose it): Google tells us your name, email address and profile picture, and Google learns that you signed in to Artixles. We get nothing else from your Google account and never see your Google password. Google's sign-in button loads from Google on the sign-in window and the signed-out home page, and if you're signed in to Google in your browser, Google may offer to sign you in with one tap there, so Google can learn that you visited Artixles even if you don't use it. Google handles its part as a separate controller under its own privacy policy.
- "About you" (what you study or work on, and any notes), your study level, citation style and other settings, and your pinned chats and projects. "About you" is sent to the AI with your questions so answers fit you. Only fill it in if you want to.
- Sign-in records: when you signed up and last signed in. Our sign-in provider (Supabase) also logs the IP address of sign-ins for security.
- Bot check: the sign-in form uses Cloudflare Turnstile, which looks at your browser and IP address to tell people from bots.
- Sign-up checks: to stop people collecting free usage with many accounts, we store a normalised form of your email address (for example without dots or "+tags" in a Gmail address) and a one-way scrambled (hashed, salted) version of the network address you signed up from. We never store that network address itself and can't turn the hash back into it.
- Where you came from: if you arrived through a link with a source tag (for example ?ref=linkedin) or from another website, your browser remembers that tag and sends it once, when you sign up. We use it to see which channels bring people who find Artixles useful.
- Study associations: if you came through a study association's link (?ref=assoc-<name>) or use its code (<NAME>-FREEMONTH) at checkout, we record that association with your account. The association gets a share of what you pay us. On its own dashboard it sees only totals: how many people signed up through it (per month), how many used its code, how many are in their free month, paying now or have ever paid, and its share. It never sees your name, email address or anything you do in Artixles.
- Invites and referrals: if you joined with someone's invite link, we record who invited you (to give both of you the invite credits). If you invite friends, we see how many joined and asked a first question.
Chats, answers and modes
Your questions, the answers, the papers found, the tables and documents made for you, and your follow-up questions are stored as your chat history. This covers normal answers, Deep, Check claims, Literature review, Reference check and the side chat. You need to sign in to ask a question: if you type one first, your browser keeps it until you have signed in, then it is asked and saved to your account. Chats are stored in our database so you can open them on any device. Literature reviews and checks keep running on our server if you close the tab, and the result is saved to your chat.
Uploaded files and essays
When you upload a file (PDF, Word, PowerPoint, Excel, CSV, text or images), we read its text once and keep only that text, inside the chat or project you uploaded it to. Small files are read straight away and never stored; larger files go briefly into private storage and are deleted right after reading. Images are transcribed into text by our AI provider. If a file is a published paper, we look up its DOI in Crossref to cite it properly.
Reference check and Check claims work on the text you paste or upload (an essay, thesis chapter or manuscript). The text is sent to our AI provider to read the references and claims; the references themselves (titles, authors, DOIs) are looked up in public research databases. The result is saved in your chat. Your essay is not added to any database, shared or used to train models.
Please don't upload documents with other people's personal data unless you are allowed to, and avoid putting sensitive information (for example about your health) in questions or files. If you do, we only use it to answer you.
Projects, team papers and notes
A project keeps chats, files, instructions and a paper together. A personal project is visible only to you. In a team project:
- Every member can see the other members' names and email addresses and their role (admin, editor or viewer), and admins see pending invites.
- Members see the project's chats, files and instructions, the shared paper, notes on it, suggested edits, the paper's checks and the team's Artixles thread. Your private Artixles thread in a project is visible only to you.
- While you have the paper open, the others see that you are there, with your name, initials and a colour, and your edits appear live. This presence is not stored.
- When someone @mentions you in a note, we email you the note (unless you have unsubscribed from our emails).
- To invite someone, an admin gives us their email address. We email them the invite on the admin's behalf, with the admin's name and the project's name, and keep the invite until it is accepted or withdrawn. The invite email has a link to never be emailed by us again.
Project admins decide who is in a project. If you leave or are removed, you no longer see it, but what you wrote there (chats, edits, notes with your name on them) stays part of the team's work.
Follow a question
When you follow an answer (the bell), we store a one-line topic, the searches behind it and which papers you have already seen. Once a week we re-run those searches in OpenAlex and PubMed, a small AI model screens new results, and we email you the relevant new studies. Stop following, switch off its emails, or delete the chat at any time.
Voice dictation
Dictation uses your browser's own speech recognition. Artixles never receives or stores audio: only the text that appears in the box, once you send it. Some browsers (for example Chrome) send the audio to the browser maker's speech service (Google) to turn it into text; that is the browser's own processing, under the browser maker's privacy policy, not ours. Firefox doesn't offer it, so the microphone button doesn't appear there.
Browser extension
The Artixles browser extension only acts when you ask it to. It has no access to the pages you visit: when you choose Artixles from the right-click menu, click its icon or press its shortcut, your browser lets it read the text you selected on that one page, and nothing else. If you turn on the optional button beside selected text, your browser asks you first; its script then only watches for a selection so it can show the button, and sends nothing until you click it and pick an action. Only the selected text goes to Artixles, over HTTPS, under your account, and is handled like a question in the app (a claim check you start there is saved to your chats; "Open in Artixles" saves the exchange as a chat). The extension keeps its own sign-in and your preferred citation style in your browser's extension storage. It does not collect your browsing history, show ads or sell data. "Sign out" in the side panel ends the extension's session; removing the extension deletes what it stored. "Cite this" sends the passage only to look up matching papers (Crossref, OpenAlex, Semantic Scholar receive search terms, not who you are).
Word and Google Docs add-ins
The Artixles add-ins for Microsoft Word and Google Docs show the Artixles panel next to your document. While the panel is open, it reads the text you have selected, and the paragraph it is in, so it can show it in the panel (Word: whenever the selection changes; Google Docs: every 3 seconds). That stays on your device: nothing is sent to Artixles until you click an action. Then the add-in sends the text you selected, or the paragraph you chose for "Sources for this paragraph", to Artixles over HTTPS under your account, where it is handled like a question in the app (search terms made from it go to the research databases, never who you are). It doesn't send the rest of your document. To insert a citation, it searches your document for the words before the place it goes and adds the citation there, and it keeps the reference list up to date; it never changes your own words. The papers you cite and your citation style are stored in the document itself (Word's document settings, Google Docs' document properties), so they travel with the file. The add-ins keep their own sign-in in the panel's browser storage on your device; "Sign out" in the panel ends that session. The Google Docs add-on asks Google only for access to the document it is opened in and to show its sidebar; it does not access your Google Drive, email or contacts. Data we receive through Google's APIs is used only to provide the add-on's features, is not sold, not used for advertising and not used to train AI models, and its use follows the Google API Services User Data Policy, including the Limited Use requirements. Neither add-in shows ads or collects your browsing history. To disconnect an add-in, click "Sign out" in the panel before you remove it (Word: My Add-ins; Google Docs: Manage add-ons): removing it alone may leave its sign-in stored on your device. The papers you cited and your citation style stay in the document.
Web search
The side chat can search the web, and an answer may run a web search to find the most-discussed studies on a question. These searches are run by our AI provider, Anthropic, with search terms made from your question; your name and email are not sent.
Research databases
To find and read papers, we send search terms and paper identifiers (never your name or email) to public research databases: OpenAlex, Europe PMC, PubMed (NCBI), Crossref, Semantic Scholar, Unpaywall, arXiv, ERIC, DOAJ, ClinicalTrials.gov, CORE, OpenAIRE, RePEc (through OpenAlex), OSF (through SHARE), OAPEN and DOAB, Zenodo and HAL, and the report pages of the CPB and the PBL; for law and government documents to Rechtspraak, wetten.overheid.nl, EUR-Lex (Publications Office of the EU), officielebekendmakingen.nl, the Tweede Kamer's open data and data.overheid.nl; and for statistics to CBS StatLine, Eurostat, the OECD, the World Bank, Our World in Data, the WHO, the ILO (ILOSTAT), the ECB, De Nederlandsche Bank (DNB) and FRED (Federal Reserve Bank of St. Louis). Search terms are made from your question, so don't put personal details in questions. Some of these services ask for a contact address; we send Artixles' own address, not yours. We also download open-access copies of papers from publishers and repositories; that sends nothing about you. Searches and paper texts are kept in a shared cache that is not linked to any person.
If you set your university library, "via your library" links take you to your library's own sign-in; what happens there is up to your library.
To show whether your university gives you access to a paywalled paper ("Read PDF · your university"), we ask GetFTR, a service run by publishers, with the paper's DOI and your device's IP address: publishers recognise university networks by their IP addresses. Your name and email are not sent. For some paywalled papers we buy the passages relevant to your question from the publisher through Valyu, our licensed-content provider; Valyu receives the paper's title and search terms made from your question, not who you are. Those passages are used for that one answer and are not stored.
The verified findings database
After an answer is written, the figures and findings in it that carry a citation are checked against the cited paper, within a time limit; each answer says how many were checked. What the check confirmed or corrected is kept in a database of verified findings: the claim as it stands in the answer, the paper (only public papers with a DOI or public database id, never your uploads or project files), the paper's own words and the check's verdict. For normal answers and literature reviews the question is kept too, cut to 300 characters and with email and web addresses removed. Nothing in this database is linked to you or your chat: no account, no chat id. In a claim check, a claim can be a sentence from your own text that cites a public paper.
What we pass on: the API's search returns, for each paper, up to five "checked findings" from this database: the claim, the figure, the verdict and the date it was checked. Only findings from Artixles' own answers are passed on, never findings from claim checks (whose claims can be your own sentences), and never the question. We may also publish findings in the same form elsewhere, for example on public evidence pages.
The Artixles API
The API includes the Artixles MCP server (https://www.artixles.com/api/mcp), which lets AI assistants such as Claude, ChatGPT or Cursor call the same API. It uses the same keys and prices, and everything below applies to it in the same way.
- Account: an API account uses your Artixles sign-in, with its own plan, prepaid balance and a separate Stripe customer, so API invoices are separate from your app invoices.
- Keys: we store a key's name, its first 12 characters (to tell keys apart), an optional monthly spending cap, when it was last used and whether it was revoked. The key itself is stored only as a one-way hash (SHA-256): it is shown once, and we can't read it back.
- Call records: for billing and your dashboard we record each call's time, endpoint, key, units, cost, status, duration and any error message (with email addresses and keys removed), but not the content of the call. Per-minute rate counters are deleted within the hour.
- Content: questions and texts sent through the API are processed like questions in the app (by Anthropic and the research databases above) to produce the response, and are not stored as chats. The figures and findings in an API answer are checked as in the app, and what the check confirmed or corrected (the claim, the paper and the verdict) goes into the findings database, but the API question is never kept. These findings are not passed on through the API's search.
- Search results: the API's search returns, per paper, a short summary written by Artixles from the abstract and the checked findings described above. Summaries are stored once per paper and shared by everyone; they contain nothing about the person who searched.
- Emails: purchase confirmations and alerts when you have used 80% or all of your included usage, or your balance is low.
Payments
Payments go through Stripe Managed Payments: a Stripe company is the merchant of record for your purchase, shown at checkout and on your receipts as "Sold through Link". Stripe takes your payment details and billing address (and a VAT number if you give one), works out and collects VAT, and sends you receipts and invoices in its own name from Link. Stripe may ask you to use or create a Link account. Stripe handles that data, and its own billing support, refunds and disputes, under its own privacy policy (link.com/privacy) and shares your order details with us. We never see your full card number. We keep your Stripe customer number, your plan and its status, renewal dates, and each purchase's amount, type and date. If a payment is disputed (a chargeback), the account is paused until we have looked at it.
Emails we send
- Sign-in emails with a link and a code (needed to sign in by email).
- A welcome email and three short tips emails in your first two weeks; then they stop. We don't send newsletters.
- Purchase confirmations (always sent, because they are about a payment).
- Emails you asked for: "your literature review is ready", new studies on a question you follow, @mentions.
- API usage and balance alerts.
Every email except sign-in links and purchase confirmations has a one-click unsubscribe link. Unsubscribing stops all of those emails for your account. Emails are sent through Resend.
Usage events, ratings and feedback
When you use the app, it records simple events, such as a question sent, a mode used, an answer copied, a file uploaded or downloaded, a setting changed or dictation used, with the time and the chat they belong to. For every request we also record its cost in AI usage, how long it took and whether it failed, but not its text. When you rate an answer or a session, we store your rating, the reasons and comment you give, and the question and a short excerpt of the answer, so we can see what went wrong. Only the server can read these records; you can't read other people's and they can't read yours.
Error records and logs
When something breaks, we store an error record: where it happened, the error message (with email addresses and keys removed), and, for errors in your browser, the page and your browser type. It may be linked to your account so we can see whether a problem affects you. It never contains what you wrote. Our hosting provider also keeps short-lived technical logs, which include IP addresses.
4. Who can see your data at Artixles
Only the people who run Artixles, through an admin page limited to their own accounts. It shows, per account: email address, name, the "what you study or work on" line from About you, plan, usage and its cost this month, number of questions, sign-up date and source, last sign-in and who invited whom; plus ratings and comments (with the question), error records and totals. It does not show your chats, files or papers. We only look at those when you ask us to (for example to fix a problem), to investigate abuse or a security issue, or when the law requires it.
5. Who processes data for us
These service providers process data on our behalf, under data processing agreements. They may only use it to provide their service to us.
| Provider | What for | Where |
|---|---|---|
| Supabase | Database, sign-in, file storage, live editing | EU (Ireland) |
| Vercel | Hosting: runs the website and server | US company; global network, server functions may run in the US |
| Anthropic | The AI model (Claude), image transcription and web search: receives your question, the chat so far, "About you", and the text of relevant papers, uploads and project files | US |
| Stripe | Payments, subscriptions and invoices; as merchant of record for your purchase it also acts under its own privacy policy (see Payments) | EU (Ireland) and US |
| Resend | Sending emails, including sign-in links | US |
| Cloudflare | Bot check at sign-in (Turnstile), domain and email routing | US company; global network |
Anthropic does not use data sent through its API to train its models, and keeps it only for a limited time for safety and abuse checks.
Transfers outside the EU: where a provider processes data in the US or elsewhere outside the European Economic Area, the transfer is covered by the EU–US Data Privacy Framework (for providers certified under it) and/or the European Commission's standard contractual clauses. Ask us for a copy of the relevant safeguards.
The research databases (section 3) are independent public services, not our processors: they receive only search terms and paper identifiers. A study association you joined through sees only the totals described in section 3, never who you are. We don't sell or rent personal data, and we only share it with others when the law requires us to (for example a court order) or to protect our rights in a dispute.
6. AI and automated decisions
Answers are written by an AI model from the papers it finds, and automatic checks compare figures and claims with those papers. This helps you research; it does not make decisions about you. Artixles does not make decisions with legal or similarly significant effects based solely on automated processing (art. 22 GDPR). Two small automatic rules do affect accounts: sign-up checks can mean a new account starts without free usage (for example with a throwaway email address or many sign-ups from one network), and a payment dispute pauses the account. Email hello@artixles.com and a person will review it.
7. Cookies and storage in your browser
We don't use advertising or tracking cookies, and no third-party analytics, so there is no cookie banner. Your browser stores a few things that the app needs:
- your sign-in session;
- your settings: theme, Deep, citation style, study level, peer-reviewed only, university library;
- a question you typed before signing in, and an invite code you followed, until you have signed in;
- where you came from (the source tag described in section 3), until you sign up;
- small reminders, such as whether we already asked how a session went, or that you snoozed a low-usage notice.
Dictation stores nothing. Cloudflare's bot check and Stripe's checkout page may set their own strictly necessary cookies for security and fraud prevention. You can clear all of this in your browser settings; you will then be signed out.
8. Business customers and the API
If you send your own users' questions or texts through the API, or use Artixles for your organisation, you are the controller of that personal data and we process it on your behalf. A data processing agreement (art. 28 GDPR) is available on request at hello@artixles.com. Please don't send us more personal data than the question needs.
9. How long we keep data
| Data | How long |
|---|---|
| Account details, settings, About you, pins | Until you delete your account |
| Chats, projects, extracted text of uploads, papers and notes | Until you delete them, the project is deleted, or you delete your account. Notes and edits you made in someone else's team project stay with that project |
| Original uploaded files | Not kept: deleted right after their text is read |
| Followed questions | Until you stop following, delete the chat or delete your account |
| Usage events, ratings, comments and feedback | Until you delete your account |
| Usage and cost records per request | While your account exists; after that kept without any link to you, for our cost accounting |
| Sign-up network hash | With your account (it is only used to count sign-ups from one network over 30 days) |
| Purchase records and invoices | 7 years, as Dutch tax law requires (invoices are kept in Stripe). Our own purchase records lose their link to you when you delete your account |
| API keys (hashes), call records and API balance | Until you delete your account; API purchase records as above |
| Project invites | Until accepted or withdrawn |
| Error records | As long as we need them to find and fix problems; when you delete your account they are no longer linked to you |
| Hosting and sign-in logs | Short periods set by our providers (Vercel, Supabase), typically days to a few weeks |
| Verified findings and the shared search cache | Kept, as they are not linked to anyone |
| After you delete your account: a normalised form of your email address | As long as we give free usage to new accounts, so the same address doesn't get it again. It is not linked to anything else |
| "Don't email me again" list | As long as needed to respect your request |
Deleted data can remain in our database provider's backups for a short time until they are overwritten. Sign-in links and other emails stay in your own mailbox until you delete them.
10. How we protect your data
- All traffic is encrypted in transit (HTTPS), and our database provider encrypts stored data.
- Database rules (row-level security) make sure you can only read your own data and the projects you are a member of. Records such as feedback, events, error logs and the findings database can only be read by our server.
- No passwords: you sign in with one-time links or codes, or with Google, and email sign-in is protected by a bot check.
- API keys and sign-up network addresses are stored only as one-way hashes.
- Uploaded files are deleted right after reading; error records have email addresses and keys removed.
- Only the people who run Artixles have admin access, and our providers are bound by data processing agreements.
11. If something goes wrong
If a data breach puts your personal data at risk, we will report it to the Dutch data protection authority within 72 hours of finding out where the law requires it, and tell you without undue delay if it is likely to put your rights and freedoms at high risk, with what happened and what you can do.
12. Your rights
Under the GDPR you have the right to:
- access your data and get a copy of it;
- portability: receive the data you gave us in a common, machine-readable format (JSON), or have it sent to another service;
- rectification: change your name and About you in Settings yourself, or ask us to correct anything else;
- erasure: delete chats yourself at any time, and your whole account in Settings → Delete my account;
- restriction: ask us to pause using your data while a question about it is being sorted out;
- objection: object to processing based on legitimate interest (such as tips emails, usage events or the findings database). For the tips emails, the unsubscribe link is enough;
- withdraw consent where we ask for it, without affecting what happened before.
For a full copy of your data, use Settings → Account and help → Download my data: it gives you one JSON file with your profile, plan and usage, chats, projects, papers, notes, follows and API account. For anything else, email hello@artixles.com from the address of your account. We may ask you to confirm it is you. We answer within one month (we can extend this by two more months for complex requests, and will tell you if we do). This is free unless a request is clearly unfounded or excessive. Data that is not linked to you, such as the findings database, can't be traced back to you, so we can't find or delete it per person.
You can also complain to the Dutch data protection authority, Autoriteit Persoonsgegevens, or the authority in the EU country where you live or work. We would appreciate the chance to sort it out with you first.
13. Children
Artixles is meant for people aged 16 and over (see the terms). We don't knowingly collect data from younger children. If you believe a child under 16 has an account, email us and we will delete it.
14. Changes to this policy
If we change this policy in a meaningful way, for example to use data for a new purpose or with a new kind of provider, we will tell you by email or in the app before the change takes effect. The date at the top shows when it last changed.