Skip to content

Benchmark Pillars

Paceghost measures AI visibility across three benchmark channels, or pillars. Each pillar answers a different question about how your domain appears in AI-driven search and discovery.

The question: Where does my domain rank in traditional search engine results?

How we measure it: Paceghost runs your key search terms through traditional search engines — Google Search, Brave Search, and Bing Search. We record whether your domain appears (“Found”), at what position, and with what snippet.

Signal strength: A ranking in traditional search is the baseline — if you’re not visible here, you’re invisible to a large portion of your audience.

The exact endpoints available depend on your tier — see SEO Endpoints.

The question: Does my domain appear in AI-augmented search results?

How we measure it: Paceghost queries Google AI Overview, Brave AI Answers, Google AI Mode, and You.com Smart with your key terms. These are search engines that use LLMs to generate answers alongside or instead of traditional results. We detect whether the generated answer names your brand or product (“Mentioned”), or cites a page of your domain as a source (“Cited”).

Signal strength: AI-augmented search is the fastest-growing discovery channel. A domain that ranks well in SEO but is invisible in GEO is missing the new front door to information.

The exact endpoints available depend on your tier — see GEO Endpoints.

The question: What do AI models say about my domain when asked directly?

How we measure it: Paceghost sends your key terms as direct prompts to multiple large language models — GPT, Claude, Gemini, and more. We record whether the model mentions your domain in its response and at what position.

In addition to the string match, every model answer is classified by how it engages with your brand: an answer counts toward visibility only when it recommends, discusses, or critically engages with your brand — answers that merely name your brand while declining or disclaiming knowledge of it are not counted. That sorting is done by a machine-learning model built for classification — not a large language model — and we use a fixed version of it. Each report’s methodology section states which definition its figures follow.

Signal strength: Direct LLM queries represent how users will discover information in the near future — asking AI assistants directly rather than searching. If your domain is never mentioned, you don’t exist in this channel.

The exact models available depend on your tier — see AEO Endpoints.

The three pillars form a complete picture of AI visibility. A domain might rank #1 in SEO (strong traditional search) but never appear in GEO or AEO (invisible to AI-augmented search and LLMs). Another domain might have poor SEO but strong AEO presence because multiple LLMs mention it favourably.

The Visibility Benchmark runs all three pillars simultaneously on the same keywords from your Synthetic Domain Profile, giving you a point-in-time snapshot of your AI search presence.

How the Visibility Percentage Is Calculated

Section titled “How the Visibility Percentage Is Calculated”

Every visibility figure is a percentage of the engines that returned results: it counts the measurements in which your domain or brand appeared, divided by the measurements that returned a usable answer. Beside every percentage, its own fraction states that basis directly — for example “44% · 4 of 9 hits” means 4 of the 9 engines that returned usable results found you.

Engines that errored, timed out, or returned no usable answer are excluded from the percentage — never silently: the number of sources attempted is shown in each figure’s explanation and stays visible wherever nothing was measured. On SEO query rows the drill-down adds a separately labelled SERP-occupancy line — your share of the actual search-result slots (“10% of results”) — alongside a “what the hit actually was” line, for example “1 of 10 results · rank #6” for a domain appearing once on a ten-result search page. Occupancy is a different metric from the hit rate and is never blended into it.

To find your domain and brand in a result, Paceghost checks each recorded answer in this order:

  1. Link to your domain — a link to any page of your domain (apex level).
  2. Domain in the text — your domain name, for example example.com, written in the answer or the snippet.
  3. Brand name — your brand name as a whole word. We only use brand names that your own site states, for example in its schema.org markup or with a ® or ™ mark that matches your domain.

There is no partial or fuzzy match on parts of the domain name. A brand name that is also a common word (for example “Apple” in a recipe) can still produce a false positive. The raw results include the surrounding text so that you can make sure that a hit is relevant. See Entity & SERP Matching for the details.