What happens to your data
What we collect and who sees it, kept short where we could manage it.
Four things worth knowing
The short version.
We don’t train models on your data
Your brand profile, questions and results are used to run your workspace and nothing else. We do not use them to train or fine-tune any model, ours or anyone else’s.
One customer never sees another
Every piece of data belongs to one workspace inside your organisation. Our server checks that boundary on every request.
We never ask for your website logins
We read your site the way any visitor or crawler does. You don’t install a plugin or a tracking script, and you never give us a login.
We never see your card
Payments go through a PCI-compliant payment provider. Card details go straight to them and never touch our systems.
The data in your workspace
- Your brand profile, audiences and tone of voice
- The questions in your question banks
- The answers assistants gave, and when
- An inventory of the public pages on your website
- Drafts generated for you
- Keyword rankings, and search and traffic totals if you connect Google
- Who on your team has access
- The Claude connections you approve, and a log of what each one did
Things we deliberately never collect
- Logins or credentials for your website or CMS
- Your customer lists, CRM data or contact databases
- Data about individual visitors to your site
- Payment card details
- Anything behind a login on your site
How access works
Sign-in is handled by an authentication provider, so we never store your password. Access is organised in two levels: an organisation, and the brand clients inside it.
Every request to our API is checked on the server against that structure. Team members see the brands in their organisation. A client guest is scoped to exactly one brand and cannot see any other, which is what makes it safe for an agency to invite a client into a shared workspace.
Connecting Google
Connecting Google Search Console or Google Analytics is optional and read-only: we can see your figures but never change anything. The access Google grants is stored encrypted. We keep totals such as clicks, impressions and sessions, never anything about individual visitors, and you can disconnect at any time.
Connecting Claude
On the Max and Agency plans, an admin can switch on Claude access, and each person then connects their own Claude. You approve the connection in Vizebel: which workspaces it can reach, and whether it may only read, also make changes, or also spend credits. Spending is off unless you choose it, needs a monthly credit limit, and every costed action is quoted in the conversation before it runs.
A connection reaches only the workspace tools it is given. It cannot see billing, API keys, invitations or organisation settings. Removing a keyword, question, topic or focus area is approved one at a time, and a removed topic or focus area is archived, not deleted. Its tokens are stored as one-way hashes and expire: an hour for access, 30 days to stay signed in. Each call is logged under the connection for 90 days. Disconnect it at any time, and switching Claude access off ends every connection at once.
Who processes your data
We use a handful of providers, and each gets only what its job needs. Ask and we’ll send you the current list.
The AI assistants
ChatGPT, Gemini, Claude and Google’s AI Overview receive your questions: Claude directly through Anthropic’s own API, the others through a service that monitors them for us.
AI models that run the product
Help draft your profile, audiences, questions and content, and analyse the answers that come back.
Search data
Keyword rankings and page performance.
Website reading
Reads the public pages of your website to build your content inventory.
Your own Claude
If you connect Claude, it receives the workspace data it asks for, within the workspaces and permissions you approved, under your own agreement with Anthropic.
Running the service
Sign-in, payments, email, hosting and our database.
A note on the assistants
The product asks ChatGPT, Gemini, Claude and Google’s AI Overview about your business, so your questions go to those providers.
ChatGPT, Gemini and Google’s AI Overview are asked through a specialist service that monitors them for us. Claude we ask directly, through Anthropic’s own API: no monitoring service covers Claude, so we go straight to the source.
We ask the way a buyer would and record what comes back. What we send is the question and, where it matters, the market it is asked from. We do not send your customer data, because we never hold any.
Deleting your data
Delete a brand client and it disappears from your workspace immediately. Its questions, results and drafts are erased 30 days later, so a deletion made by mistake can still be undone. To close your whole organisation, email us; the same 30 days apply. Billing records we are required to keep are retained with your name removed, and we keep no shadow copy for analytics.
Reporting something
If you believe you have found a security issue, please email us before disclosing it publicly. We’ll reply to say we have it and keep you posted. If you’d like the credit, it’s yours.
Being straight about our stage
We are a small, young product. We hold ourselves to the practices described on this page, but we don’t yet hold formal certifications like SOC 2 or ISO 27001. If your procurement process needs specific assurances, talk to us and we will answer honestly about what we can and can’t provide today.
For the formal detail, see the privacy policy and terms of service.
Questions about data?
Ask us before you sign up. A person reads every message.