Onbiom

Privacy

What we collect, why, how long we keep it, who else handles it, and how to use your rights.

Found a report page about your website? Remove it now. No account, name or email needed. It disappears at once and is not published again.

Report pages about websites

When a scan of a public website completes and its robots.txt allows it, we publish a report page at /domains/<domain>. This applies both to a scan a visitor started and to a check Onbiom ran because the site is on a public list of business websites. It shows only the domain, the check date, scores and the names of checks that were not met, never page content.

A page is visible for 30 days after the latest scan; its stored results are then deleted. For a site checked from a list, we also keep its domain, the name of the list and the date of our last check, so that we check it at most once every 30 days. We delete this once the site is no longer on a list we use and our last check is more than 30 days old.

Why: we rely on our legitimate interest in showing site owners how search engines and AI assistants read their site.

Remove a page: anyone can remove a page at once with the removal form. We then keep only the domain, the stated reason and the date on a permanent do-not-publish list, so the page is not created again. The form asks for no name or email; your IP address is held in memory for up to one hour only to limit abuse. If you email us about a report, we use your message and address only to answer you and correct the report. To keep your site out of report pages in advance, add the robots.txt rule on our crawler page.

If you scan a website without an account

Why: to show you the result and bring your browser back to it.

We keep the result in encrypted form, shortened and with sensitive parts removed, together with the page your browser should return to or the limit notice you saw. Only the browser that started the scan can open this saved copy, for a fixed 30 minutes; a limit notice lasts a fixed 70 minutes so it survives the next limit reset. Opening it again does not make it last longer, and it stays tied to your browser when you sign in. After that time we no longer show it, and we delete it the next time our storage is used. We do not keep the web pages we downloaded. Refreshing a report does not run the scan again.

A recent result is shared for 3 hours. A scan reads only what a website shows publicly, so we also keep one copy of the latest complete result for each website address, with nothing about you or your browser in it. For 3 hours after the scan, anyone who checks the same address, or opens the link to that result, sees it instead of a new scan; this way one scan answers everyone and the website gets fewer requests. The copy is shortened, with email addresses and other sensitive parts removed, but it can include details the homepage itself shows, such as a business name, address or contact details, which may identify a sole trader. We store it encrypted. After 3 hours we no longer show it, and we delete it the next time our storage is used. If a site owner uses the removal form, we delete that website's copy at once.

If you pick an AI tool server (MCP server) found by the scan, we keep its exact address and which account it belongs to, encrypted and tied to your browser, for 20 minutes. You must review it before saving, and it can be used only once.

If you ask for an email when a site changes

Why: to tell you when the site changes, because you asked.

If you ask to be emailed when a scanned site changes, we send one confirmation message and store only your email address, the site domain and your consent time until you stop. Confirmed domains are re-checked at most once a day, shared across everyone watching the same domain, and we email only when a change appears on two consecutive checks. Every email has a stop link that deletes your address from that watch; an unconfirmed request is deleted after 3 days. We do not use the address for anything else.

If you have an account

Why: to run your account and the checks you set up.

We store your confirmed email address, your login codes in scrambled (hashed) form, the addresses of the servers you monitor, the descriptions of the tools they offer, and your check history.

If you give us keys to reach a private server (access tokens, refresh tokens or app secrets you registered in advance), we encrypt them and use them only for the one monitor and exact server address you saved them for. A key never allows us to run tools or send anything to an AI model; those need their own approval.

If you save a website to check

Why: to check the websites you saved and show you what changed.

When you save a website, yourself or through an AI app you connected, we store its address and the name you give it, for up to 10 websites per account. We store each check result in encrypted form and keep the latest 20 results per website, each for up to 7 days. Removing a website deletes its results. A complete check of a saved website also becomes the shared recent result for that address for 3 hours, as described for scans without an account; it never includes your account, the name you gave the website or checks shown only to you.

If we check a saved website automatically, we email your account address after the first automatic check and later only when a check finds a change. The email holds the website's address, its score and what changed. It is part of the saved-website service, not advertising. Each email has a stop link for that website; stopping does not delete the website or its checks. We keep which checks we emailed about and whether you stopped until you remove the website. Opening the website from an email is counted only in daily totals, with no identifiers.

If you make a fix kit for a website

Why: to prepare the fixes you asked for and check whether they are live.

A fix kit is a set of suggested changes to a website's public files. Without an account, we store your draft in encrypted form, tied to your browser, for 7 days after your last change and never longer than 30 days, together with its latest 3 checks of your live pages. To limit abuse we keep a coded form of your IP address for 1 day. In your account, a draft, its edits and your approval are stored in encrypted form for 24 hours and then deleted; a check of your live pages against it is kept for 24 hours.

If you ask us to re-check an approved kit, we check your live pages once, 4 days later, and email you the result. If you then turn on weekly checks, we check once a week for up to 12 weeks and email you only when something changes. We store the page address, the checks you approved and their results, but not your email address: we read it from your account when we send the email. Every email has a stop link. We delete this 90 days after you set it up or last turned weekly checks on, once weekly checks have stopped, and at once if your account is deleted.

If you ask for AI advice about a saved website

Why: to explain the saved findings and suggest fixes, only when you approve that request.

Before each request, we ask you to confirm sending a shortened summary of your saved website checks to Amazon Web Services through the model gateway we operate. It includes check identifiers, pass/fail results, scores, check times and our standard explanations and fixes from the latest check and, when available, the previous check. It excludes page text, website addresses, account details, credentials and tool inputs. Amazon receives the request in Frankfurt and processes it using its EU model routing. The production and test services use the same AI service, with separate access keys and a new session for each request.

Advice can be wrong and does not change your technical results or publish anything. We store the summary, your one-request approval, the advice and its model and runtime details encrypted and private to your account, until the source checks expire, at most 7 days, or you remove the website. You can cancel a pending request; if it has already been sent, Amazon may already have processed it. Each new request needs a new approval. We keep a separate usage record with coded request identities and counts, without website text, for up to 400 days to account for credits.

If you connect an AI app to your account

Why: to let an AI app you choose, such as an AI assistant, work with your account in the ways you approve.

When an AI app asks to connect, we store the name and return address it gives us. If it is not used within 15 minutes we delete it; otherwise we delete it 90 days after its last use. When you approve a connection, we store which app it is, what it may do and when it ends. A connection lasts up to 30 days. The access codes the app receives are stored only in scrambled (hashed) form, and each works for 1 hour.

If you create an access key for an AI app, we store its name, what it may do, its first and last few characters and a scrambled copy, for the 30, 60 or 90 days you choose. You can disconnect an app or revoke a key at any time on your account's connections page. It stops working at once, and we delete the record when it would have ended.

If you check an AI tool server (MCP server)

Why: to run the checks you chose and show you the results.

Our regular checks only read the list of tools your server offers; they never run a tool. You can separately set up and confirm optional tool runs, which may change things on your side and cost you money with your provider. For these we store the exact fixed tool arguments (the inputs you set) and a shortened copy of the tool's answer with sensitive parts removed.

If you add quality checks, we store what you expect the tool to do, the example data you give us, and a separate history of results. Our rule-based comparisons run on our own servers.

AI reviews are optional and need your approval. Only after you approve it separately, we send the shortened part of the answer you selected and your expectation to "AI review (local)", an AI model we run on our own servers. We never send the tool inputs, and we leave out example data the review does not need. We keep which model and scoring rules were used and the model's opinion; that opinion is advice only and never proves your tool is correct. Approving the local review never lets us send data outside. Only a second, separate approval sends the same selected material and expectation to "AI review (external, sends data to Cloudflare)", a model run by Cloudflare, without the local review's opinion. Both opinions are advice only. You can withdraw an approval at any time on the monitor's page.

To count AI reviews against your credits we keep a small usage record: scrambled attempt codes, counts and how much AI processing was used. It holds no tool inputs and none of the selected material, and we keep it for up to 400 days after the monitor's history ends.

Detailed history follows your plan's limit. Separately, we keep simple connection results (time, success or failure, response time) for up to 366 days when collected on Starter, also after paid access ends, and up to 8 days on Free. If you publish a public status link, it shows only these simple results; you can turn it off, and it is hidden when paid access ends. Deleting a monitor removes its connection results, public link, checks and pending notifications.

Tool lists that a server makes public may contain information from that server, so add only servers you are allowed to monitor. Do not put passwords, keys or personal data in monitor names or addresses.

If you send alerts to Slack or your own web address

Why: to deliver your monitor alerts where you asked.

On Starter you can send alerts to a Slack channel or to your own web address (a webhook), as well as by email. We store the label you choose, the address in encrypted form (we show only a shortened version), a signing secret for your own web address, and the time and result of the last delivery. Each alert sends its subject and text there; after that, Slack or your own service holds the message under your own arrangement with them. A destination stays until you delete it; deleting it also cancels alerts not yet sent to it.

If you send a pitch link

Why: to show our check of a website to the person you send the link to.

When you create a pitch link we store the name and contact you enter, the site address, a copy of the findings from that check and the link's dates. Anyone with the link can see the name and findings, and reach your contact, until it expires after 30 days or you turn it off. We delete it 30 days after it ends, when you remove the site, or when you delete your account. We never contact the people you send it to and keep no contact details for them. Opening a link is counted only in daily totals, with no identifiers.

If you ask for an AI answer sample

Why: we rely on our legitimate interest in showing businesses how AI assistants describe them.

When you ask for a sample answer about a website you saved, we send one fixed question containing only that website's domain to an AI model that Amazon Web Services runs for us in the EU, as our processor. The answer is written by the model, not by us, can be wrong, and may mention people connected to the business, such as a sole trader's name. We store the domain, the answer, its date and token counts for up to 7 days and then delete them. Any account that saved the same domain sees the same answer in that time, and we do not record which account asked. Answers are never published. If an answer mentions you, use the removal form and we delete it and stop creating answers for that domain.

How we measure whether Onbiom works

Why: to learn whether people get value from Onbiom and to improve it.

We note, once each, when a browser first reaches these steps: a scan, saving a monitor, a first successful tool-list check, a later scheduled check, and opening your results. We store only the step and its time, under a coded ID made for this purpose alone from a browser cookie, for up to 90 days from the first record. When you sign in and open your dashboard, run a scan, review a server found by a scan or save a monitor, we can link that coded ID to your account number if no account has it yet; no other account can take it over afterwards.

These records never contain the addresses of servers you check, cookie values, tool inputs, AI prompts, IP addresses or browser headers. We label each record as someone running an AI tool server, someone checking a website, our own staff, test traffic or a bot; unlabelled records are left out of our rates. Our figures use signed-in, labelled records only; anonymous and unlabelled visits are counted separately and are not used to estimate how many visitors become customers.

We may show two versions of the homepage text to learn which explains Onbiom better. The version comes from that same browser cookie, which also protects our forms. We do not store which version you saw; we count only daily totals per version, with no identifiers.

Our legal basis for each use

UseLegal basis
Running your account and sign-in emails; saved websites, monitors and their history; fix kits and re-checks you ask for; AI app connections; alert destinations; scans without an account; pitch links you createNeeded to provide the service you asked for (GDPR article 6(1)(b))
Billing for paid plansNeeded to provide the plan you pay for (article 6(1)(b)); Polar, as merchant of record, keeps billing records because tax law requires it (article 6(1)(c))
Change emailsYour consent, given when you confirm; withdraw it with the stop link in any email (article 6(1)(a))
AI reviewsYour consent, given with each separate approval; withdraw it on the monitor's page (article 6(1)(a))
AI advice about a saved websiteYour consent for each request; cancel it while pending (article 6(1)(a))
Report pages, recent scan results shared with anyone who checks the same address, checks of sites from public lists and AI answer samplesOur legitimate interest in showing businesses how search engines and AI assistants read them (article 6(1)(f))
Our own measurement of whether Onbiom worksOur legitimate interest in learning whether Onbiom helps people, and in improving it (article 6(1)(f))
Abuse limits and the do-not-publish listOur legitimate interest in keeping Onbiom safe and in keeping removed report pages removed (article 6(1)(f))

How long we keep it

WhatHow long
Report page about a website30 days after the latest scan
Record of a site checked from a listUntil the site is off the lists we use and our last check is over 30 days old
Do-not-publish entry (domain, reason, date)Permanently, so the page is not created again
IP address from the removal formUp to 1 hour, in memory only
Scan result without an account30 minutes; limit notices 70 minutes
Shared recent result of a website scan3 hours
Server picked from a scan, before you save it20 minutes
Change email requestUntil you stop; unconfirmed requests 3 days
Saved website check resultsLatest 20 per website, each up to 7 days
Private website AI advice, evidence summary and approvalUntil its source checks expire, at most 7 days, or you remove the website
Website AI advice usage recordUp to 400 days
Fix-kit draft without an account7 days after your last change, at most 30 days
Coded IP address for fix-kit drafts1 day
Fix-kit draft in your account, and checks against it24 hours
Kit re-check and weekly checks90 days after you set them up or last turned weekly checks on; weekly checks run up to 12 weeks
AI app connectionUp to 30 days; access codes 1 hour
AI app registration15 minutes if unused; otherwise 90 days after its last use
Access key for an AI app30, 60 or 90 days, as you choose
Slack or web address alert destinationUntil you delete it
Detailed check historyAs long as your plan allows
Simple connection resultsUp to 366 days on Starter; up to 8 days on Free
AI review usage recordUp to 400 days after the monitor's history ends
Pitch linkLive for 30 days; deleted 30 days after it ends
AI answer sample7 days
Product measurement recordsUp to 90 days
Billing recordsKept by Polar, the merchant of record, as tax law requires

Who else handles your data

Some of these companies are based in the United States and may handle data there. For those transfers we rely on the safeguards each company states in its own data protection terms, listed below.

  • Our own servers: the Onbiom app, its database and the "AI review (local)" model run on servers we operate ourselves.
  • Cloudflare: every visit to our website passes through Cloudflare, which delivers our pages to your browser. Cloudflare also runs "AI review (external)", only after your separate approval. Cloudflare may handle this data outside the EU, including in the United States; its data processing terms say it follows the EU-U.S. Data Privacy Framework and uses the EU standard contractual clauses.
  • Google: confirms your identity when you choose to sign in with Google, under Google's own privacy policy. Google says it is certified under the EU-U.S. Data Privacy Framework and uses the EU standard contractual clauses where required.
  • Resend: sends our emails, such as sign-in links, alerts and change emails. Our emails leave from Resend's sending region in Ireland, but Resend stores its email records and logs in the United States; its data processing terms say it uses the EU standard contractual clauses and follows the EU-U.S. Data Privacy Framework.
  • Polar: handles payments for paid plans as merchant of record (the company that formally sells them). Polar may store data in the United States, Canada or elsewhere outside the EU; its privacy policy says it uses the EU standard contractual clauses.
  • Amazon Web Services: runs the AI model for AI answer samples and separately approved website AI advice, as our processor. We send requests to its Frankfurt region and use only its EU model routing.

Your rights and how to use them

Under the GDPR (articles 15 to 22) you can ask us to:

  • show you the data we hold about you and give you a copy;
  • correct data that is wrong;
  • delete your data;
  • limit how we use it;
  • give you your data in a common format, to keep or move elsewhere;
  • stop using it where we rely on our legitimate interest, such as for report pages, shared recent scan results, checks of sites from public lists, AI answer samples and our own measurement;
  • withdraw your consent at any time, for change emails and AI reviews; this does not affect what we did before;
  • not make decisions about you by computer alone that significantly affect you.

When we correct, delete or limit data, we also tell anyone we shared it with, unless that is impossible.

How: email support@onbiom.com and say what you want. We may ask you to confirm it is you, for example by writing from your account's email address. It is free. We answer within one month, or tell you within that month if a complex request needs longer. To delete your account, email us the same way; billing records required for tax remain with Polar as merchant of record.

Faster, without emailing us: remove a report page with the removal form; stop change emails with the stop link in any of them; delete a monitor or revoke its public status link in your account; withdraw an AI review approval on the monitor's page; disconnect an AI app or revoke its key on your account's connections page; remove a saved website or alert destination in your account.

If you think we handle your data wrongly, you can complain to the Dutch Data Protection Authority (Autoriteit Persoonsgegevens) or to the authority where you live.

Contact us

Questions about your data or this page? Email support@onbiom.com.

Contact: support@onbiom.com.

Contact: support@onbiom.com.