Metrics
How Ranqo asks each engine
How Ranqo asks each AI engine your prompts, and why an answer in the app you use can differ from the one Ranqo recorded.
Every answer in Ranqo comes from Ranqo asking the engines your tracked prompts itself. This page explains how it asks, and why an answer you get in an AI app can differ from the one Ranqo recorded. How runs are scheduled and how each answer is read are on How Ranqo measures.
How Ranqo asks
- Through each engine's API, not an app. ChatGPT, Claude, Perplexity, Gemini and Grok are asked through their developer APIs, the way software talks to an engine. Google AI Overviews is read from Google's live results page for the prompt.
- With web search on. ChatGPT, Perplexity and Grok search for every answer; Claude and Gemini decide when a search helps.
- No account behind it. No one is signed in, and no chat history, saved memory or personal settings go with a question.
- The same way for every brand. Each engine answers with one fixed model, and every brand's questions go out with the same instructions, written for no brand in particular. That keeps results comparable across brands and from one run to the next.
- From your market, in your language. Each prompt is asked from its own location, and in your brand's answer language when you set one. See Location and language.
Why the app can show something else
An AI app knows things about you that Ranqo's questions never carry, so the answer you get there can name different brands:
- Your account. Signed in, an app can use your saved memories, past chats and custom instructions.
- Your location. Apps estimate where you are, often down to your city. Ranqo asks from the prompt's market instead.
- Your plan's model. The model that answers in an app can depend on your plan, and some apps choose one for each question.
- Whether it searches. Most apps decide for each question whether to search the web.
- Ads and shopping cards. These can appear around an answer in the app. Ranqo records the answer itself and the sources it cites.
Answers also vary from one asking to the next, even for the same person. A single answer in the app is one sample; Ranqo asks the same questions the same way on every run, which is what makes runs comparable. To tell a real change from normal variation, see Noise bands.
Reading the answers Ranqo recorded
Open a prompt's insights page from Prompts and read AI Response History for its recent answers, or open a run in Snapshots to see every answer from that run. Selecting an answer shows its text, usually in the layout Ranqo reformats answers into (see How an answer is read), and the sources it cited.