Build a transparent reporting system based on dated sample questions, cited sources and actual business inquiries. State what the method cannot measure.
What actually matters
Unlike a simple fixed keyword report, AI responses can vary by user, wording, time, location and search access. A custom 'AI visibility score' may be an agency's internal sample index, but it is not an official cross-platform rank. Keep the raw evidence behind any summary: exact prompt, model or product, timestamp, result, citation and whether a recommendation was explicit or merely a link.
A practical action plan
- 01
Choose 10–20 realistic buyer questions across your main services, including nonbranded prompts.
- 02
Run them on a defined schedule and save the full dated observations with citations where available.
- 03
Count observable events separately: named mentions, linked citations, inaccurate details and no mention.
- 04
Combine this with verified Search Console clicks, Analytics referral data where available, and qualified lead records.
An example for a real business
Report 'mentioned in 3 of 12 sampled queries on this date' rather than 'AI authority score: 94'—and explain the sample is not comprehensive.
Don't fall for this.
Don't imply a homemade visibility score is certified by OpenAI, Google or Anthropic.
Look for real progress.
Improve factual accuracy, cited references and real inquiry performance while tracking sampling variance.
Check the source material.
We rely on public guidance for platform mechanics, and we don't claim special access or guaranteed rankings. These references explain relevant policies and practices.
This is educational guidance, not a live audit of your business, a verified search-volume report or a promise that an AI product will recommend you. Platform behavior can change. Published content is intended for ongoing factual review.