Growli

方法论

Growli 如何衡量 AI 可见度——以及它不声称能衡量什么

Growli 里的每一个数字,都来自真实的助手回答或我们实际读取过的页面,并标明采集日期。下面说明具体做法,以及它的局限在哪里。

能见度60+8
提及丢失赢得提及

哪些是实测、哪些是推断——以及哪里缺少数据

  • Measured: mentions, competitors, cited sources and dates come straight from stored assistant answers.
  • Measured: website and local signals come from pages we actually read and Maps data we actually fetched.
  • Inferred: prioritization ranks actions by impact and effort — a judgement on top of measured signals, not a measurement.
  • Assistant answers are non-deterministic and location-sensitive: the same search can name different businesses on different days, so we re-run on a schedule and show the trend, not a single reading.
  • When a stage cannot be collected — a site blocks our crawler, an assistant returns nothing — we show it as not measured, never as a zero or an estimate.
  • We do not invent data: no fabricated scores, competitors, reviews, statistics or customer results, ever.

01

We build the searches your customers actually make

From your website, category and market we generate buyer-intent searches — category, comparison, alternative, problem and, for local businesses, near-me questions. You can edit, add and group them; you own the set.

02

We ask real AI assistants and record every answer

Each search runs against real assistants (ChatGPT, Gemini, Claude, Perplexity and others). We store the full answer, the assistant, the date and the sources it cited — the raw evidence behind every metric.

03

We detect mentions, competitors and sources — and rank actions

In each stored answer we detect whether you are mentioned and where, which competitors are named, and which websites are cited. Gaps become prioritized actions, each linked back to the answer that justifies it.

每个数字的含义

Share of AI answer

Of your tracked searches, the share where an assistant mentions you — per assistant, over time. Definition: your mentions ÷ tracked answers. Not a guess: it is counted from stored answers.

竞争对手

Businesses in your category named in the same answers as you (or, for local businesses, same-category Maps neighbours in your market). Ranked by how often answers repeat them. Review platforms are treated as sources, never competitors.

有影响力的来源

The websites assistants cite when answering your searches, with how often each appears and whether it mentions you — extracted from the answers themselves.

证据

Every finding opens to the exact search, the assistant, the answer text, the cited URLs and the collection date. Nothing is shown without the record behind it.

常见疑问,一一解答

One audit runs your tracked searches across the assistants once, reads your website, fetches local data where relevant, and produces findings and actions — all stored with their collection date. Monitoring re-runs the searches on a schedule so you see change over time.

AI assistants generate answers rather than returning a fixed list, and their answers depend on time, location and phrasing. A brand can appear on one run and not the next. That is why we sample on a schedule and report the trend, with every individual answer kept as evidence.

The set depends on your plan; the exact assistants probed are shown in-app and never implied by icons alone. Google-SERP co-occurrence is available only when a search-data provider is connected.

A competitor is a business in your category that appears in the same answers (or, for local businesses, the same-category Maps neighbourhood). We rank them by how often answers repeat them and exclude review or community platforms, which are sources rather than competitors.

The relevant section says so — 'not measured' or 'no competitors named in the measured answers' — rather than filling the gap with an estimate. Absence is reported as absence.

看看这套方法用在你自己的业务上是什么结果

运行你的第一次检查——每个数字都可追溯到一条真实的助手回答,并附带采集日期。

每周通讯

每周 AI 搜索增长技巧

和众多企业主一起,获取被 Google 收录、被 AI 推荐的实用建议。