Sign up

Features

AI visibility: what assistants say about you

On this page

The AI crawler report shows which assistants read your pages. AI visibility shows what they then say: MetricRaven asks your tracked prompts to each assistant every week and measures whether your brand appears, where, next to which competitors, and which pages get cited.

AI visibility is part of the Growth plan, with an allowance included in the price (no top-ups). The free trial includes it with the 100k allowance, so you get about two weekly checks before choosing a plan. Starter shows a sample preview with an upgrade link.

How it works

  • Prompts. You list questions a buyer might ask, such as best privacy-friendly analytics with revenue tracking. Suggest prompts drafts ideas from your brand, domain, competitors and mention keywords.
  • Assistants. Each prompt is sent to ChatGPT, Claude, Perplexity, Gemini and Grok (how many depends on your tier) with web search on, the way their apps answer. Requests go through OpenRouter to each provider's current fast model.
  • Google's own answers (optional). Switch on Google AI Mode and Google AI Overviews in the settings to see what Google itself answers for the same prompts. They come on top of your tier's assistants and use 1 credit each per prompt.
  • Weekly checks. Every active prompt runs on every enabled assistant once a week. Run now starts a check immediately, once a day.
  • Parsing. Each answer is read for the brands it names (in order, with the sentence around each and its tone) and the sources it cites. Your brand counts when the answer names it, an alias, or your domain as a whole word.

Set it up

  1. Open the AI visibility tab and choose Prompts

    Admins and the owner can edit; viewers and share links see the results.

  2. Add three to five prompts

    Ask the way a buyer would, without your brand name (except comparisons like X vs Y). Group them if you like, for example Pricing or Alternatives.

  3. Check your brand name and add competitors (optional)

    The brand defaults to your website's name. Every brand the assistants name appears in the leaderboard anyway; listed competitors are always matched and their pages are flagged in citations.

  4. Wait a few minutes

    The first check runs shortly after you add prompts, then weekly.

Gemini, Google AI Mode and AI Overviews

Three ways Google can answer a buyer, each measured on its own:

AssistantWhat it measures
GeminiThe Gemini model answering with Google Search grounding, as in the Gemini app.
Google AI ModeThe answer in the AI Mode tab of Google Search, with the sources it links.
Google AI OverviewsThe AI Overview above the regular search results, for your prompt typed as a search. Google shows one only for some searches.
  • Turn them on under Settings → Assistants → Google's own answers. They're off by default and don't count toward your tier's number of assistants.
  • Search from sets the country Google answers for, in that country's language. The default is the United States, in English. The other assistants have no location.
  • When Google shows no AI Overview for a search, the prompt grid shows n/a for it. That answer is left out of your scores and of the Not mentioned filter, because there was nothing to mention you in. It still uses its credit, since the search ran.

Metrics

MetricMeaning
Visibility score0–100, position-weighted: named first = 100, second = 50, third = 33, not named = 0, averaged over every answer.
Mentioned inShare of answers that name your brand.
Average positionYour place among the brands an answer names, when it names you (#1 is first).
Named firstShare of answers that name you before anyone else.
Pages citedShare of answers that cite one of your pages as a source.
ConsistencyWhen one assistant names you for a prompt, the share of assistants that do too.

Below the headline figures: the score over time (overall or per assistant), results per assistant, the brands assistants mention ranked by the same score, a prompt-by-assistant grid (select a prompt to read the answers, brands in order and sources), and the domains and pages cited most.

Every row opens a detail:

  • An assistant: its score, mention rate, position and citation rate, how its score moved across your checks, and its latest answer to each prompt.
  • A brand in the leaderboard: the share of answers naming it across your checks next to yours, its average position, and the prompts where it's named and you aren't.
  • A cited domain or page: how often answers cited it across your checks, which prompts and assistants cite it now, and (for a domain) which of its pages.
  • A prompt: every assistant's latest answer, the brands in the order named and the sources. Answers listed in the other details open their prompt too.

Credits and limits

One credit is one prompt asked to one assistant; Gemini costs 2 credits per prompt because Google charges for each web search it makes, and Google AI Mode and AI Overviews cost 1 each. A check costs active prompts × the credits of each chosen assistant. Your monthly allowance covers weekly checks of every prompt on every assistant your tier includes, Gemini among them, plus both Google answers; Run now uses the same credits.

Growth tierPrompts · assistants · credits a month
10k10 · 3 · 265
100k20 · 3 · 525
200k30 · 4 · 915
500k50 · 4 · 1,525
1M75 · 5 · 2,610
2M100 · 5 · 3,480
5M150 · 5 · 5,215
10M and up200 · 5 · 6,955
  • The prompt limit counts active prompts across all your websites. Pause a prompt to free its slot.
  • Credits reset with your monthly usage. When they run out, checks pause until the reset; nothing is ever billed extra. A larger tier is the way to get more.
  • Answers that fail (an assistant times out or errors) don't use credits.

Choose the period

The tab uses the same toolbar as Traffic and Revenue: the date range (presets, pinned periods or custom dates), the arrows for the previous and next period, Compare, Filter, the reporting timezone (the globe) and Refresh. The period is shared across the tabs, so switching tabs keeps it. The headline figures and the cards describe the latest check up to the end of the period; Visibility over time shows every check in it. Checks run once a week, so a period of a few days often has none: the tab then says so and offers a longer range.

Compare

The % changes under the headline figures compare the latest check with the one before by default. Choose vs same period last year or vs custom start to compare with the latest check in that window instead, or No comparison to hide the changes. The choice is shared with Traffic and Revenue.

Filter the answers

Choose Filter in the toolbar to look at some of the answers: an Assistant, a Prompt, a Topic (the prompt group), a Brand named in the answer, Your brand (mentioned, not mentioned, or your page cited) or a Cited domain, each with is or is not. Every card is recomputed from the matching answers: the scores and their change, the trend, assistants, brands, prompts, citations and the details behind each row. Filters live in the page address, so a filtered view can be bookmarked or shared.

API and MCP

Read the results with GET /api/v1/stats/ai-visibility or the MCP tool get_ai_visibility. Both read stored checks and never start one. Add competitor, domain, page or model for one brand, cited site, page or assistant in detail, or prompt for a prompt's latest answers.

terminal
curl "https://metricraven.com/api/v1/stats/ai-visibility?range=90d" \
  -H "Authorization: Bearer $METRICRAVEN_SECRET_KEY"

Details

  • Answers vary from run to run; a weekly trend across several prompts is more telling than one answer.
  • Full answer text is kept for 30 days. The brands, positions and sources parsed from it are kept with your data.
  • The public demo shows sample data: fictional brands, answers and sites. Compare it with the AI crawler report, which shows which assistants fetch your pages.