How the AI Answer Tracker works
You choose the questions your clients really ask. On a set schedule the tracker sends each question to the engine through its official API, takes several fresh samples, and saves every answer word for word, with its citations and the search queries the engine ran. Grok runs today. ChatGPT, Perplexity, Google AI Overviews and AI Mode, and Copilot are rolling out.
To be confirmed: check frequency for customer plans
Your questions
The tracker works from a fixed, versioned set of questions, so results can be compared over time. Questions that name your brand are reported separately, so they don't inflate how often you're discovered for general questions.
Engine by engine
Each engine is checked in the way its official tools allow. The table shows how.
| Engine | Status | How it's checked | Location handling | What's stored |
|---|---|---|---|---|
| ChatGPT | Rolling out | OpenAI API with web search | Location set to Australia by API setting | Each answer, its citations, the sources consulted and the search queries run |
| Perplexity | Rolling out | Perplexity Agent API with web search | Australian location set by API setting | Each answer, its search results and the queries run |
| Grok | Live | xAI API with web search | Set in the prompt only, because the API has no location setting; results are labelled to show this | Each answer, its citations and the sources it found |
| Google AI Overviews and AI Mode | Rolling out | Search Console Generative AI report To be confirmed: whether manual Google checks are included | Australia filter in Search Console | Impressions by page |
| Copilot | Rolling out | Monthly Bing Webmaster Tools AI Performance export | Global data; there's no Australia filter | Citation activity by page and grounding query |
What's live today
Grok is live. ChatGPT, Perplexity, Google AI Overviews and AI Mode, and Copilot are rolling out. Automatic checks against your facts are coming soon. Each is marked live only once it's running. The table above shows how each engine is, or will be, checked.
Why several samples
AI engines can give a different answer each time you ask the same question. So the tracker asks each question several times per engine, each in a fresh session, and reports trends rather than single answers. A "win" only counts once it repeats.
Forced search vs auto
Most checks tell the engine to search the web before answering. These results are labelled "forced-search": they show how visible you are when the engine searches. Once a month an "auto" check lets the engine decide for itself whether to search, and records whether it did. Auto results are shown as their own series.
Accuracy checks (coming soon)
You'll keep a facts file with the details that should be right in any answer about you: your fees, the services you do and don't offer, your service area, and claims you never make. Each answer will be checked against it. A mismatch will get a reason and go to a review queue, where a person confirms it before it counts. This is switched off for now. Until it's on, every answer is still saved word for word for you to read.
What it can't tell you
- API answers aren't the consumer app. They're a consistent way to track change over time, not a copy of what a person sees in an AI app.
- It can't promise placement. It measures whether you're named and cited; it doesn't change what engines say.
- Search Console's Generative AI report gives impressions only, with no clicks or queries.
- It isn't a compliance tool. It records what engines say; it doesn't assess whether your business or your pages meet any regulator's or professional body's rules, and it doesn't give legal advice.
Questions about the method
Is this the same as what I see in the app?
No. Consumer AI apps can use different models, settings and personalisation from the API, so the tracker's results are a trend signal rather than a copy of your own screen.
Why does Grok use a prompt for location?
xAI's web search tool has no location setting, so the tracker tells Grok in the prompt that the user is in Australia, and labels Grok's results accordingly.
How are AI Overviews measured without scraping Google?
Through Google's own Search Console Generative AI report, which shows AI Overviews and AI Mode impressions for your pages. The tracker never scrapes google.com.
More answers are in the FAQ.
Official sources
Find out what AI says about your practice or firm
Join the waitlist for early access. We'll let you know when the AI Answer Tracker opens to Australian health and legal businesses, and the agencies that work with them.