AI Visibility Audit Checklist

Use five steps to check whether AI answers name your brand, see which pages they cite, and rerun the same test after an update.

Five checks for a repeatable audit

  1. Freeze the buyer questions

    Write five questions a buyer might ask without naming your brand. Cover discovery, use case, and evaluation. Choose a market, target brand, and set of platforms. Save the exact wording before opening an answer.

  2. Run the same questions on each platform

    Use a fresh conversation for each question. Record the platform, model or interface, search settings, market, and date. Keep branded questions in a separate group so they do not change the unbranded baseline.

  3. Record mentions and citations separately

    Save each answer. Mark whether it named the target brand and separately log the URLs that appear in its citations. Open the cited pages to see what each supports. A brand mention can appear without a link to the brand.

  4. Score valid runs by platform

    Divide valid unbranded answers naming your brand by all valid unbranded answers for that platform. A valid answer without your brand stays in the denominator. Mark errors and timeouts as missing and report their count alongside the score.

  5. Assign a change and schedule a retest

    Pick one documented gap and assign a page, owner, and date. For example, if buyers ask for proof of comparable work, verify whether the relevant service page links to a case study with a clear scope and outcome. Retest the same questions and record any change in model or search settings.

Use the buyer question planner to draft your set and the measurement methodology to define scoring groups.

Worked example: two valid answers, one mention

In the October 13, 2026 US benchmark, question DSA-02 asked which agencies a mid-market B2B software company should consider to grow non-brand organic demand. For Siege Media, the ChatGPT-style run returned a valid category-only answer and Perplexity named the brand.

ChatGPT-style
0 of 1 valid runs
Perplexity
1 of 1 valid runs

This question produced one mention across two valid runs. Use each platform’s own result as its baseline, and review the cited URL separately. Two runs give a worked example for scoring. Repeat across questions and dates before reporting a trend.

Inspect the public benchmark·Open the filled CSV example

When you have answers to review, use the AI citation gap analysis guide to inspect competitor sources and document what you can verify.

Start with five buyer questions

The free check samples five organic buyer questions across ChatGPT-style answers and Perplexity. After reviewing the result, you can choose the one-time Snapshot for more questions and evidence. View a sample report.

Run the free check