Data use.

What we read.

During onboarding we crawl up to twelve pages of your website (the homepage plus the first eleven internal links we discover) to calibrate your brand voice and infer your category. We retain only the markdown extract of each page; we do not store screenshots or full HTML beyond the duration of the crawl run.

What the LLMs read.

When the audit runs, we send each engine your buyer-intent prompts via the engine's commercial API. None of the four providers we use train on the API endpoints we call (Anthropic Claude, OpenAI gpt-4o, Google Gemini, Perplexity sonar — all commercial tiers with zero-data-retention or no-training defaults). We never send the engines content marked as private or unpublished.

What we publish.

Approved drafts publish under your business's byline on a polhia.com subdomain (e.g. polhia.com/p/your-business/post-slug) and, if you've connected one, to your CMS. We never publish anything you have not approved. We never sell, share, or syndicate your content to a third-party brand.

What we measure.

We measure citation rate, sentiment, competitor citation rate, and per-engine breakdowns. We do not measure traffic, do not plant tracking pixels in your published pages, and do not have a third-party analytics tracker on polhia.com.

What we share.

Cohort-level aggregates only — for example, "the median family- law firm in our cohort hit 32% citation rate by week 12" — and only when the cohort contains five or more customers (the k-anonymity threshold we use). Individual customer data is never visible outside the account it belongs to.