15 / Improve
AI Visibility asks ChatGPT, Gemini, Claude and Perplexity the questions your customers ask, with web search on, and records whether each answer mentions your brand and cites your site.
Search visibility now includes AI answers. The AI Visibility page measures it directly: you choose the prompts, SiteAuditLint sends them to each model through its official API with web search enabled, and parses the citations each provider returns.
1. Setting up a check
- Add API keysSettings > Connections > AI models: paste keys for OpenAI, Google AI Studio, Anthropic or Perplexity, then Test keys and Save.
- Enter brand namesThe names people use for your brand, plus optional competitors by name or domain.
- Write promptsUp to 50 questions, one per line, written the way a customer asks. Leave your brand out of them.
- Choose models and runs1 to 3 runs per prompt; 2 is a good default.
- Run checkIt runs in the background. Stop keeps the answers so far.
2. Models and citation sources
| Model | Default | API used | Citations from |
|---|---|---|---|
| ChatGPT | gpt-5-mini | OpenAI Responses API with web_search | url_citation annotations |
| Gemini | gemini-flash-latest | Gemini API with Google Search grounding | Grounding chunks |
| Claude | claude-haiku-4-5 | Claude Messages API with web search | Web search result citations |
| Perplexity | sonar | Sonar chat completions | search_results |
3. What you see
- Headline: cited in X of Y answers, the change since the previous check, and mentioned in Z.
- By prompt: each model's result, for example Cited #2 (2/2), Mentioned or Not mentioned.
- By model: citation and mention rates, average citation position and any errors in plain words.
- Cited sites: every domain the answers cited, ranked. This is who wins your prompts.
- Your cited pages: which of your URLs AI cites, with their critical issues and warnings from the latest audit.
- Competitors: how often each one was mentioned and cited.
4. Scheduled AI visibility checks
Tick Include an AI visibility check in scheduled audits, at most every N days (default 7). Scheduled runs then check each site that has prompts, and the Slack message adds an AI visibility line plus the prompts where you are not cited. Audits run in the app window never start a paid check on their own.
API keys are stored in Windows Credential Manager. Prompts and answers stay in your local database, and you can export one row per answer to CSV or post the summary to Slack.
Frequently asked questions
What is the difference between mentioned and cited?
Mentioned means a brand name appears as a whole word in the answer. Cited means your domain or a subdomain is among the sources the answer used.
How much does a check cost?
Usage is billed to your own provider accounts. The page shows an estimate before each run, based on prompts x models x runs at roughly US$0.01 to $0.04 per answer depending on the model.
Are API answers the same as the ChatGPT app?
Close, but not identical, and answers vary between runs. Read the numbers as a trend, and use 2 runs per prompt for steadier results.
Want the background first? Browse the SiteAuditLint Academy, or see all features.