How to Measure Your AI Visibility: A Step by Step Guide

Why AI visibility measurement matters

You cannot improve what you do not measure. AEO has matured to the point where citation share is as trackable as traditional rank position used to be, and any program operating without measurement is flying blind.

A serious measurement program does four things.

It establishes a baseline so future change can be quantified.

It surfaces the prompts where the brand is winning, losing, or absent.

It reveals where competitors are eating share that should be yours.

It produces a monthly report your leadership team can act on with confidence.

Without measurement, AEO is a hope. With measurement, AEO is a managed program.

Step 1: define your prompt set

The first step is to pick the set of prompts you will run every month. The right prompt set is.

Representative. It mirrors what your buyers actually ask AI, including category questions, comparison questions, and recommendation questions.

Stable. The same prompts run every month so you can track change over time.

Sized for signal. Typically 50 to 200 prompts is enough to produce a reliable mention rate across the major surfaces.

A good way to assemble the set is to mix three sources: prompts derived from your top buyer keywords, prompts derived from your sales team's most common discovery questions, and prompts that name your category and competitors explicitly.

Step 2: run the prompt set across every major AI surface

The second step is to actually run the prompts. The major surfaces to cover in 2026 are.

ChatGPT, including the most common product configurations your buyers are likely using.

Claude, including the consumer surface and the enterprise surface where relevant.

Gemini, including the standalone product and Google AI Overviews appearing in search.

DeepSeek, where it has traction in your buyer audience.

Perplexity, where research first buyers tend to be active.

Running 200 prompts manually across five surfaces every month is impractical. Most serious programs use a tool like Gumshoe to automate the runs.

Step 3: track the five core metrics

Five metrics matter consistently across categories.

Mention rate. The percentage of prompts where the brand appears at all.

Citation share. Across the prompt set, the percentage of total brand mentions that belong to the brand versus competitors.

Recommendation frequency. How often the model places the brand inside a direct recommendation versus a hedged or generic mention.

Source attribution. Which URLs are driving citations so you know which content formats are working.

Sentiment. Whether the brand is described positively, neutrally, or negatively when mentioned.

These five metrics together give a full picture of AEO performance.

Step 4: benchmark against competitors

Single brand metrics are interesting. Competitive benchmarks are actionable.

A strong measurement program runs the same prompt set for the brand and the top three competitors. The resulting citation share comparison shows where competitors are taking share that should be yours, where your differentiation is being recognized, and where new entrants are gaining ground.

Competitive benchmarking is often what turns AEO measurement from a report into a strategy.

Step 5: report monthly with clear actions

Measurement without action is reporting theater. A useful monthly report includes.

The headline metrics for the brand and top three competitors, with month over month change.

The five prompts with the biggest mention rate gain and the five with the biggest loss.

The URLs driving the most citations and the URLs that should be driving more.

A prioritized next month action list tied to specific prompts and pages.

OnlyAEO delivers this report monthly to every retained client so each cycle of work is informed by the prior cycle's signal.

Common measurement pitfalls

Several common pitfalls undermine AEO measurement programs.

Inconsistent prompt sets. Changing prompts between cycles destroys comparability. Pick the set, freeze it, and only revise on a planned schedule.

Single surface tracking. Only measuring ChatGPT misses Claude, Gemini, DeepSeek, Perplexity, and AI Overviews. Cross model coverage is the point.

Manual sampling without rigor. Running ten prompts once does not produce a reliable baseline. The math requires a real sample.

No competitive benchmark. Without competitor data, share gains and losses are invisible.

No action loop. If the report never feeds back into next month's content and entity work, the program stagnates.

What to do with the first three months of data

The first three months of measurement establish the baseline trend line. Use them to.

Identify the prompts you already win and protect them with refreshed content and schema.

Identify the prompts you should win, given your authority, and prioritize content for them.

Identify the prompts where a competitor has a clear and probably durable lead, and decide whether to compete head on or pivot to adjacent prompts.

Identify the prompts that are not strategic and deprioritize them so resources go to the work that moves the share number.

A disciplined three month cycle usually produces a clear roadmap for the next two quarters.