AI Transparency Policy

1. Which AI providers we use

Model access is routed through the Lovable AI Gateway, which passes prompts to the underlying providers above. The specific assistants tested on a given run are recorded in the report.

Provider coverage changes as models are released and retired. We record the platform and model family used for every observation so historic reports remain interpretable.

• OpenAI (ChatGPT models)
• Google (Gemini models)
• Anthropic (Claude models)
• Perplexity, where included in your plan

2. Where AI is used in the product

• Question generation: AI drafts a representative set of real customer questions for your sector, services and locations. The set is grouped by intent and stored with your report.
• Visibility testing: those questions are submitted to AI assistants, and every business named in each response is recorded.
• Free preliminary check: a single lightweight AI estimate of visibility for a submitted domain.
• Summarisation: AI assists in drafting plain-English explanations of findings.
• Scoring: performed by Visora's own deterministic scoring service, not by a language model.

3. What information may be sent to AI providers

Testing prompts contain business-level information: your business name, website, sector, services, locations and the customer question being tested. We do not send account credentials, billing details, payment data or your customer records.

Do not enter personal data into business fields if you do not want it included in testing prompts. Where a provider offers the control, we do not permit your data to be used to train their models.

4. AI outputs are probabilistic

Large language models generate responses probabilistically. The same question, asked twice, can return different businesses, different wording and different ordering. Responses also vary by model version, prompt wording, date, location, user context, platform updates and the information available to the model.

Because of this, Visora tests a set of questions repeatedly and reports observed rates and patterns rather than treating a single answer as fact. Every report is a snapshot of the moment it was produced.

5. Human oversight

AI is used to generate test questions, to record assistant responses and to help draft plain-English explanations. Scores are calculated by Visora's own deterministic scoring service, not by a language model. Methodology changes, scoring weights and published guidance are reviewed and approved by a person before release. Individual report text is produced automatically and is not reviewed line by line before you receive it.

6. Known limitations

• AI providers do not publish their selection or ranking logic. We report observed differences and their likely contribution, never confirmed causes.
• AI responses can be inaccurate, out of date, or describe a business incorrectly. Recording such an output is evidence of what the assistant said, not confirmation that it is true.
• A score measures visibility across the questions in your approved set. Change the set and the score changes with it.
• Sampling is finite. Rare or highly personalised answers may not appear in a run.
• Results in one region, language or assistant do not necessarily transfer to another.
• The scoring model version applied to your run (currently Visora Visibility Score v1.2) determines how observations are weighted.

7. Outputs are recommendations, not guarantees

Recommendations describe observable changes that may strengthen the signals AI systems appear to rely on. They are professional observations, not legal, financial or marketing advice, and outcomes remain outside our control.

• We do not guarantee rankings in any AI assistant or search engine.
• We do not guarantee that your business will be mentioned, cited or recommended by any AI system.
• We do not guarantee traffic, leads, enquiries, revenue or business growth.
• We do not guarantee that a result observed today will be observed again tomorrow — AI outputs change over time.
• We do not guarantee that implementing a recommendation will produce any particular outcome.

8. Competitor comparisons

Competitor comparisons reflect Visora's testing methodology and the results observed during the runs shown. They are observations about visibility in AI responses only, and are not statements about the quality, suitability, legality or superiority of any business.

9. Use of AI outputs in reports

Where AI responses are shown, we display only the extract necessary to support a finding and identify it as an observed output recorded during testing. Findings are summarised rather than reproduced at length.

10. Independence

Visora is not affiliated with, endorsed by or sponsored by any AI provider. We do not optimise, influence, manipulate or control AI systems. We measure observable outputs and advise on observable business signals.

11. Questions

Questions about how AI is used in your reports can be sent to hello@visora.app.