Is my data used to train AI?
No. Silktide does not use your data to train AI models, and every AI provider that processes your data for Silktide is barred by contract from training on it.
There is one exception, and it is deliberate: the questions Presence asks public AI assistants such as ChatGPT. This page explains both, so you can decide with the full picture.
What Silktide promises
Silktide does not use your data, or anything AI writes for you, to train or fine-tune AI models.
This is in our AI terms. They also require every AI provider that processes your personal data to be under a written contract that prohibits it from training on your data unless you expressly agree otherwise.
What our AI providers promise
When a Silktide feature needs an AI model, Silktide calls that provider's business API. That covers Silktide Agent, AI reports, AI optimizers, Inspector explanations, AI spell check, and the AI that reads and analyzes answers in Presence. Business APIs are covered by commercial terms, not consumer terms, and each provider's terms rule out training on what Silktide sends and what comes back:
- OpenAI. "Data sent to the OpenAI API is not used to train or improve OpenAI models (unless you explicitly opt in to share data with us)." Data controls in the OpenAI platform
- Anthropic. "Anthropic may not train models on Customer Content from Services." Commercial Terms of Service, section B
- Google, United States region. "When you use Paid Services, including, for example, the paid quota of the Gemini API, Google doesn't use your prompts ... or responses to improve our products." Silktide uses paid access. Gemini API Additional Terms of Service
- Google, European Union region. "Google will not use Customer Data to train or fine-tune any AI/ML models without Customer's prior permission or instruction." Google Cloud Service Specific Terms, Training Restriction
- SpaceXAI. Never trains on API inputs or outputs without the customer's explicit permission. API security FAQ
- Mistral. "Mistral AI will not use Customer Data or Outputs to train its artificial intelligence models" except where the customer has chosen to allow it. Commercial Terms of Service, section 4.2
- Perplexity, which answers Presence questions only. "Perplexity shall not use (or authorize third parties to use) Customer Content to train, retrain, fine-tune or otherwise improve any generative artificial intelligence models." API Terms of Service, section 2.3.3
Several providers let a customer opt in to training, usually in exchange for free usage. Silktide has not opted in with any of them, and never will with your data.
Not training on your data is different from not storing it. Providers keep requests for a limited period to detect abuse. See Data, privacy, and retention.
The exception: questions Presence asks public AI assistants
Presence shows how AI assistants answer the questions your customers ask. A business API does not give the same answers the public sees, because the public products choose their own models, search, and settings. So to measure what your customers actually see, Silktide asks the public products.
For these assistants, Silktide scrapes the public product. An automated browser, run by Silktide or by a service that collects public search and AI results for us, asks your question the way a member of the public would and records the answer:
The same goes for the Google and Bing searches Presence runs to measure search rankings.
When Silktide scrapes a public product, it is a visitor to that product, not a business customer with a contract. That product's own policy for public users decides whether a question asked there may be used for training, and those policies differ between companies and change over time. So Silktide cannot promise that a question asked this way is never used to train a model. We would rather say so plainly than make a promise we cannot keep. No tool that measures these assistants as the public sees them can honestly promise otherwise.
Claude, Perplexity, and Grok are different: Silktide asks them through their business APIs, so the provider terms above apply.
What a scraped question contains
Only the question itself, plus the country and language it is asked for. Questions are about your market, such as "What are the best universities for engineering in Ohio?". You write them or Silktide suggests them, and they can name brands, including yours and your competitors'.
Nothing else is sent: no website content, documents, analytics data, account details, or the name of anyone in your organization. The public product is not told which organization the question is for.
Because the question goes to a public product, treat it as public. Do not put anything confidential in a Presence question, such as an unreleased product name.
Governance and Analytics never scrape AI
The exception above applies to Presence only. Governance and Analytics send your data to AI only through business APIs, under the provider terms above. Your administrators choose which of those providers may be used on the AI screen.
If your organization cannot accept any exception, talk to your Silktide contact before you buy Presence.
Your own AI products
If you connect your own AI product, such as ChatGPT or Claude, to Silktide through MCP, what that product does with your data is governed by your agreement with its vendor, not by Silktide's.