UI vs API

Genwolf collects AI answers by scraping live assistant surfaces.

That is the main method. In some cases we also use official model APIs — they are generally more expensive.

People see ChatGPT, Gemini, Perplexity, Google AI Overviews, and Google AI Mode in the product UI. We measure those rendered answers.

The layers behind an AI answer

An AI chat response is not a single system.
It's a stack of layers:

1

Model knowledge

What the model has learned and internalized.

2

External sources & citations

Which documents, domains, or references are pulled into the answer.

3

Chat software & personalization

UI logic, conversation history, location, user profiling.

Official APIs mostly expose the first two layers. The chat or search UI is the third — and that is what users actually read.

What Genwolf measures

Genwolf still tracks the underlying signals:

1. Model-level knowledge & recognition

Whether the model:

  • knows your brand
  • associates it with the right category
  • considers it relevant to a given question

This is still the foundation of AI visibility.

The product UI can still add sources, formatting, and retrieval that a raw API response does not show.

2. Source-level signals

Which sources are cited or referenced:

documentation
blogs
review sites
authoritative domains

Sources strongly influence:

  • which brands appear
  • how they are framed
  • who is compared against whom

Tracking sources reveals why visibility happens, not just that it happens.

Why Genwolf scrapes the live UI

ChatGPT, Gemini, and Perplexity in the product are not the same as their developer APIs. Google AI Overviews and Google AI Mode have no equivalent official API at all.

UI scraping lets Genwolf:

capture the answer a user would actually see
include engines that only exist as rendered surfaces
use official APIs only where they actually help

Scraping is still a clean, repeatable session — not a logged-in user with history. The point is the rendered assistant surface, not fake personalization.

When Genwolf uses APIs

For ChatGPT, Gemini, and Perplexity, the official API is the automatic fallback: if the live surface cannot be captured on a given run, Genwolf retries the same prompt through the provider API so the run still produces a measurement. An API answer is a valid measurement of what the model knows and which sources it cites. It is not a 1:1 copy of the chat UI.

Google AI Overviews and Google AI Mode have no equivalent official API, so those engines are scraped only.

Why APIs are generally more expensive

Provider APIs bill per token, often with extra cost for web search or grounding. Running the same prompt set every day across several engines adds up quickly.

Official APIs charge for input, output, and often search tools
UI scraping avoids that per-token meter
That is one reason scraping is the main collection method

APIs are useful in some cases. They are also the costlier way to collect the same kind of answer.

Trade-offs & limitations

What Genwolf captures well

  • the rendered assistant answer
  • brand mentions and citations
  • source-driven visibility
  • relative competitive presence

What Genwolf does not capture

  • individual personalization
  • logged-in conversation history
  • a guaranteed 1:1 match to every user's screen

These limits are explicit and intentional. The goal is repeatable measurement of live surfaces, not a clone of one person's chat.

Summary

Collection is primarily live-surface capture, with APIs as fallback:

Scrape the live UI→Fall back to APIs→APIs cost more

Genwolf measures the answer users see.

Live-surface capture is the main method. APIs are the fallback, and they cost more.

Documentation