AI terms

01What these terms cover

These AI terms explain how Comma uses artificial intelligence to deliver its services, and the rules that apply to AI features in your workspace. They form part of your agreement with us, alongside our terms of service and data processing agreement. If they conflict with the terms of service on a point about AI, these AI terms win.

02Where we use AI

AI helps our team work faster and more thoroughly. We use it to:

• Research: read your website and public sources to build an accurate picture of your brand, products and market.
• Analyse AI answers: understand how AI engines answer your buyers' questions, which brands they name and which sources they cite.
• Analyse community signals: summarise public community threads, keyword and brand mentions and sentiment, so our team can act on what we find.
• Draft: produce first drafts of content, briefs, recommendations and reports.
• Review pages: score pages for how easily AI engines can read and cite them.
• Answer questions: power the assistant in your workspace, which answers questions about your data and work.

03Our AI providers

We use a small number of providers, each for a specific job:

• AI model provider: our current model provider is Anthropic. We use its Claude models through its commercial API, including its web search tool.
• Web search provider: finds and reads public web pages for research.
• AI answer monitoring provider: collects the public answers AI engines give to the questions we track for you.

We may change or add providers. The named list of all our service providers is available on request at privacy@getcomma.ai, and changes are notified under our DPA.

04What data goes to AI providers

We send an AI provider only what a task needs. Depending on the task, that can include:

• your brand information, such as your website content, brand facts, products and positioning;
• drafts, briefs and comments relevant to the content being worked on;
• public content we're analysing, such as AI answers, web pages and community posts;
• performance data relevant to the question, such as search queries from your connected Search Console;
• your questions to the workspace assistant, and the workspace data needed to answer them.

Data is sent only to deliver the services to you, never to produce work for another client.

05No training, short retention

Under our provider's commercial terms, our AI model provider may not train its models on the data we send it. We don't opt in to any programme that would allow it, and we don't send it feedback on outputs. We choose providers on that basis and won't move to one that trains on client data. We don't use your data to train AI models of our own either.

Our model provider deletes the data we send it, and what it sends back, within 30 days. It keeps data longer only where the law requires, or for up to 2 years if its automated safety systems flag a request as a possible breach of its usage policy. It stores this data in the United States, and the processing itself may run in other countries where it operates.

06Human review

AI outputs are drafts. Our team reviews and edits them before anything is published or sent on your behalf, and content that goes to your CMS follows the approval workflow agreed with you. Analysis and reports are checked by our team before we rely on them for recommendations. Answers from the workspace assistant are shown to you directly, so treat them as a starting point and check anything important.

07Ownership and accuracy of outputs

• Ownership: you own the outputs we create for you from the moment they exist. We assign our rights in them to you on creation and keep none, as set out in our terms of service.
• Accuracy: AI can produce statements that sound right but are wrong, out of date or incomplete. Our review reduces this risk but can't remove it, so we don't warrant that outputs are free of errors.
• Similarity: AI models can produce similar wording for different users. Outputs for you may resemble content created elsewhere, and other people may receive similar outputs.
• Third-party rights: we work to avoid outputs that copy protected material, and we cover intellectual property claims about deliverables as set out in the terms of service.

08Your responsibilities

• Give us accurate brand facts and tell us when they change. AI outputs are only as good as the facts behind them.
• Review content before you approve it, especially claims about your products, results, pricing, regulated topics and competitors.
• You're responsible for the content you approve or ask us to publish, and for making sure it complies with the laws and platform rules that apply to you.
• Don't put special categories of personal data, or other people's confidential information, into prompts or the workspace assistant unless you're allowed to and we've agreed it.

09Prohibited uses

Neither we nor you will use AI in the services to:

• impersonate real people or organisations, or create content that pretends to be from someone it isn't;
• create or run fake, purchased or automated accounts, or manipulate communities, votes, reviews or rankings;
• produce deceptive, defamatory, hateful or unlawful content, or misleading claims about competitors;
• infringe intellectual property or privacy rights;
• break the usage policies of our AI providers or the rules of any platform or community.

We may decline or stop work that would break these rules.

10Community data

We monitor public community threads, keyword and brand mentions and sentiment through a public community data source and our own analysis, and act on what we find as part of the services. We use AI to summarise and classify this content to understand brands and topics, not to profile individuals. Our privacy policy explains how we handle it.

11No guaranteed results

AI engines are run by other companies and change often. The same question can get different answers from one day to the next. We work to improve how your brand appears, but we can't guarantee that any AI engine will mention, rank or cite your brand, and monitoring results reflect what we observed at the time.

12AI assistants you connect

You can connect your own AI assistant to your workspace through MCP. When you do, the data the assistant pulls from your workspace is handled by that assistant's provider under your agreement with them, not ours. MCP tokens act with the permissions of the person who created them and can be revoked at any time in your workspace settings.

13Changes to these terms

AI moves quickly, and we'll update these terms when our use of it changes. We'll change the date at the top and, for material changes, tell clients by email at least 30 days in advance. Questions go to legal@getcomma.ai.