Definition
In short
AI visibility is how often, how prominently and how accurately a business appears in answers generated by AI search systems. Unlike a search ranking, it is a rate rather than a position, because the same question asked twice can return different businesses. It is measured by running a fixed prompt set repeatedly across several engines and recording what comes back.
[ 01 ]Metrics
Each one answers a different question. The headline score is a weighted roll-up, and it is never published on its own.
01
How often you are named at all.
The share of prompt runs in which the business is named anywhere in the answer. Each prompt runs multiple times per engine because answers vary, so this is a frequency and never a single observation.
Percentage of runs
02
How often you are put forward as a choice.
The share of runs in which the business is presented as a recommendation rather than mentioned in passing. Being listed as one of three suggested providers counts; appearing in a footnote does not.
Percentage of runs
03
How often your own pages are credited.
The share of runs where the engine links or attributes to a source you control, such as your website or your profile, rather than describing you from a third-party page only.
Percentage of runs with a citation
04
Where you land among the businesses named.
The mean ordinal position of the business among all businesses named in a given answer. First named averages nearer 1. This is measured only across runs where the business appears.
Mean ordinal among named businesses
05
How much of the local conversation you hold.
Your mentions as a share of all business mentions across the prompt set, compared with the named competitors in the same set. Local AI visibility is relative, so this is the figure that shows whether you are gaining ground or standing still while others move.
Percentage of total mentions in the set
06
How many engines actually know you.
The number of measured AI search experiences in which the business appears at least once. Coverage exposes concentration risk: a business visible on one engine and invisible on four is more fragile than the headline score suggests.
Engines out of 5
07
How much of the buying journey you show up for.
The share of distinct prompts in the set that surface the business at least once. A business can score well on one popular prompt and be absent from every other question a customer might ask.
Percentage of prompts in the set
[ 02 ]Reporting standard
This is the disclosure block that accompanies every score on this site, on client reports and in our published research. The example below shows the format using a real measured result.
Measurement context
That figure belongs to North Idaho Archery, a MxdifyLocal client, and it is published with their permission. Read the full engagement in the North Idaho Archery case study.
A score without a date range, a prompt count, an engine list and a geography is not a measurement. It is a claim. We do not publish those, and you should be skeptical of anyone who does.
Current engine set: ChatGPT, Google AI Mode, Google AI Overviews, Gemini, Perplexity. Methodology version 1.0, last reviewed 2026-08-11.
[ 03 ]Engines
Kept in one configuration so a change updates every page, report and case study at once.
OpenAI
ChatGPT
A general assistant that answers local questions with a short list of businesses, often with web citations when browsing is used.
Google AI Mode
Google's conversational search mode, which answers a query end to end and links out to sources.
Google AI Overviews
The generated summary shown above classic Google results for many queries, with linked source cards.
Gemini
Google's assistant, which answers local questions and can ground responses in Google Search.
Perplexity AI
Perplexity
An answer engine that cites its sources inline for nearly every response.
[ 04 ]Questions
It is a single 0 to 100 roll-up of the metrics on this page for one business, one category, one geography and one measurement period. It is only meaningful alongside that context, which is why every score MxdifyLocal publishes carries its date range, prompt count, engine set and methodology version.
Generated answers are probabilistic and can also be affected by personalization and recency. That is exactly why a single check is not measurement. We run every prompt several times per engine and report rates rather than one-off results.
It can be measured consistently, which is the practical version of objectivity. Fix the prompt set, the engines, the geography and the run count, then repeat on a schedule. Numbers produced under different conditions should not be compared, and we do not compare them.
No. It is a leading indicator of discovery, not a sales figure. We report it next to real business outcomes such as calls, form fills and clicks, and we say plainly when a movement in the score has not yet shown up in leads.
Write ten questions a customer would ask an assistant to find a business like yours, run each one three times in each engine, and record whether you were named, where, and which sources were cited. It is tedious but honest, and it will tell you more than any tool that reports a single number.
Weighting, prompt construction and how the roll-up is calculated.
What to do once you know where you stand.
Measured client outcomes with dates and named limitations.
The signals that influence whether you get named.
Part of the Mxdify family
Same studio standards as Mxdify, scoped and priced for businesses that serve a town.
