How AI Query Analytics Works

This page explains what happens under the hood when Quratic runs a check.

The measurement problem

Traditional SEO tools measure rankings on search engines like Google, where ranking position is well-defined and observable. AI platforms work differently:

There are no positions — the AI writes a free-form response
The same query can return different responses on different runs
Responses are localised — what a user in Japan sees differs from what a user in Singapore sees
AI platforms are not transparent about why they mention certain brands

Quratic solves this by automating the process a human researcher would follow: ask the question, read the answer, note which brands appear.

Real browser, real IP

Quratic uses a genuine Chromium browser — not an API call where available, but a real browser session to simulate what users actually experience. Traffic is routed through a residential IP address in the target country — a real home internet connection, not a data-centre proxy.

This means:

AI platforms see a request that looks identical to a real user
Localisation features (language, region-specific results) are active
Results reflect actual user experience, not sanitised API responses

Response parsing

Once the response is captured, Quratic:

Extracts the plain text of the AI response
Scans for your brand name (case-insensitive exact match)
Scans for each competitor name (case-insensitive exact match)
Extracts all hyperlinks and checks if any point to your domain or competitor domains

The raw response is stored so you can read it any time in Responses.

Data freshness and cadence

Checks run on a schedule determined by your plan (weekly, every 48 hours, or daily). The dashboard shows data from all checks within your selected date range.

Because AI responses are non-deterministic (the same question can get slightly different answers each time), Quratic tracks trends rather than single data points. Share of voice over many runs is more meaningful than any single response.

Variability and sample size

AI responses are non-deterministic. A brand might appear in 9 out of 10 runs of the same prompt and be absent on the 10th. This is normal behaviour.

Quratic handles this by:

Running each prompt on a consistent schedule and accumulating results
Showing moving averages in trend charts
Reporting share of voice as a percentage, not a binary yes/no