A content audit begins with simple inputs: a page and a question a buyer might ask an AI model. Before you add a company URL, it helps to know what enters the analysis and what to ask if your organization has its own data requirements.
This article explains the audit inputs and outputs described by the product. It does not replace a privacy policy, data processing agreement or security documentation.
What you provide
TextRank uses the selected page URLs (up to three per prompt), one or more prompts, selected models and project context: your brand, product, audience, market and language. That context helps determine whether a question and page relate to a real buying decision. Your team can review and correct it before the audit.
In TextRank, you can audit one page or a group of pages, with a maximum of three of your own page URLs per prompt (buyer question). For example, one question about your offer can cover a product page, pricing page and documentation page.
For a first audit, choose a public URL and a question a buyer could reasonably ask. Do not put passwords, API keys, private links, personal data or confidential company information in a prompt.
What your page is compared with
For the selected question, TextRank records the chosen model's answer and the available sources it cites or mentions. It reads your page and available source pages, then splits their text into passages for comparison. Results from different models stay separate.
Passage analysis also uses AI models. The specific model used for this work depends on the audit configuration. Choosing ChatGPT for the answer does not by itself mean that every processing step uses only a GPT-family model. The method documentation explains passage selection and how to interpret the result.
What appears in the result
The report keeps the prompt, model, date, answer, available sources, your page passage and the suggested edit together. A writer can see the observation behind the task. A product expert can verify the facts before publication. After an edit, you can return to the prompt and compare a new observation with the earlier one.
If a source cannot be read, the result should not present a full comparison of its content. If an audit identifies a missing fact on your page, that is not permission to publish the fact without confirmation.
Interpreting answer provenance and scope
A report records an answer obtained in a particular context, not every buyer conversation. When comparing results, check the prompt, date, market, language and any available model identifier and answer acquisition mode. If a report does not specify a model version or acquisition method, do not infer either from the ChatGPT or Gemini brand name; ask for clarification before using the result in a study.
Gemini / Google AI Overviews shares one TextRank audit result at 7 credits. It is one observation and one charge, not two independent measurements. Do not treat the shared result as evidence that every Gemini and Google Search session will return an identical answer.
A mention contains the brand name; a citation references a specific source; a recommendation suggests a product in a choice context. These signals can occur separately. Retrieval, Selection and Citation Opportunity are TextRank assessments for prioritizing work, not readings of a provider's internal ranking.
What to clarify before wider use
Before auditing material covered by your company's internal data policy, ask for current written answers about retention, deletion, processing location, AI service providers and TextRank team access. A feature description or example report cannot establish those terms.
If you need that information to evaluate TextRank, email contact@textrank.ai before starting an audit. A public page and a non-sensitive prompt are enough to examine the method itself.
Frequently asked questions
Do I need to submit confidential material to start?
No. Start with a public page and a buying prompt that contains no confidential information.
Where can I find retention and deletion details?
If your company needs to confirm those terms before an audit, ask the TextRank team for its current data documentation and terms.