Overview

Most teams discover AI search the same way: someone types a question into ChatGPT, sees a competitor recommended, and wonders why it is not them. A single screenshot is a fine alarm bell, but it is a poor measurement of brand visibility in ChatGPT. AI answers change between runs, between engines and between locations.

This playbook shows how to build a measurement setup you can trust: what to track, how to set a baseline, how often to run it, and how to read the results without chasing noise.

What you are actually measuring

Visibility in an AI answer has several parts, and it helps to keep them separate.

  • Mentions: is your brand named in the answer at all, and in what position relative to other brands?
  • Citations: does the answer link to one of your pages as a source? A citation can happen without a mention, and the reverse.
  • Sentiment and description: how does the answer describe you? Is it accurate, and is it favorable?
  • Share of voice: across all tracked prompts, how often do you appear compared with each competitor?
  • Referral traffic: how many real visits arrive from AI assistants?

Tracking only one of these misleads you. A brand can be mentioned often and never cited, which means it is known but not credited, or cited often but described poorly.

Step 1: build your prompt set

Your tracked prompts are the foundation, so spend time on them. Aim for a realistic mix rather than a long list.

Cover the buying journey

Include prompts for each stage: problem-aware questions (“how do I track AI search visibility”), solution-aware questions (“best tools for X”), comparison questions (“X vs. Y”) and branded questions (“is X any good”). Weight them toward the moments where a recommendation changes a decision.

Write them like people talk

Assistants are used conversationally. Use full questions and natural phrasing, and add context where your buyers would, such as company size, industry or country. If you sell in several markets or languages, track prompts for each, since answers differ by region and language.

Keep the set stable

Once you pick a set, keep it fixed so that trends mean something. Add new prompts deliberately and note the date, rather than swapping them out whenever you feel like it.

Tip: Start with 25 to 50 prompts that really matter to revenue. A focused set you review weekly beats 500 prompts nobody looks at.

Step 2: choose the engines

Pick the engines where your audience actually asks questions. For most brands that means ChatGPT, Gemini, Google AI Overviews and AI Mode, Claude and Perplexity, with Grok and others as relevant. Track them separately, because each has its own sources and habits. A rollup is helpful for reporting, but the per-engine view is where you find what to fix.

Step 3: set a baseline

Run the full prompt set and record your starting position before you change anything. Capture mentions, citations, which domains are cited, and who your competitors are in these answers.

Because answers vary, a single run per prompt is noisy. Tracking daily over a couple of weeks smooths this out and shows the spread, which tells you how much movement is normal. That spread matters later: a change smaller than the usual swing is not a result.

Step 4: add your competitors

Choose the five to ten competitors that show up most in your category, including any that surprise you. The AI engines often name brands your sales team never mentions. Track them on the same prompts so you can compare head to head and see where each one is cited.

Step 5: connect visibility to traffic

Mentions are a leading indicator, but traffic and pipeline are what the business cares about. Make sure visits from AI assistants are recognizable in your analytics. Referrers from assistants are inconsistent, so a lightweight tracking snippet on your site that classifies AI visits is more reliable than relying on referrer headers alone.

With that in place you can see which assistants send visitors, which pages they land on, and whether those visitors convert. If you use GA4 or a CRM, send the AI source through so reports separate it from other channels.

Step 6: set up a review rhythm

Measurement only helps if someone looks at it. A simple cadence works well:

  • Daily: alerts for sharp drops, lost citations, or a competitor surge. You do not need to read a dashboard every morning.
  • Weekly: a short review of score and share-of-voice trends, new competitor gains and the top three things to fix.
  • Monthly: a scorecard for stakeholders showing movement against the baseline, wins, and what you plan next.

How to read the results

A few habits keep you honest.

  1. Compare like with like. Look at the same prompts, engines and regions across time.
  2. Respect the noise. If your score normally moves a few points day to day, do not celebrate or panic over a few points.
  3. Ask why before acting. When a metric moves, break it down by engine, prompt and cause. A drop in one engine often has a different fix than a drop everywhere.
  4. Check accuracy. Review how you are described. Wrong claims about pricing, features or availability are worth fixing at the source, even if your visibility is otherwise fine.
  5. Close the loop. After you change a page, re-check the affected prompts after a week or two and see whether it helped.

Common mistakes

  • Tracking brand-name prompts only, which mostly measures people who already know you.
  • Changing the prompt set constantly, which destroys comparability.
  • Treating one engine as representative of all of them.
  • Measuring mentions but not citations, or the reverse.
  • Collecting data without assigning anyone to act on it.

Doing this with aeoora

aeoora handles the setup described above in one place. You add prompts (or accept suggestions), pick competitors, and it runs them daily across engines including ChatGPT, Gemini, Claude, Perplexity, Google AI Overviews and AI Mode. You get an AI Visibility Score from 0 to 100 built from mentions, citations and position, a breakdown of why it changed, citation and sentiment reports, and AI referral traffic from a cookie-free snippet. Daily Pulse emails and Slack alerts keep the review rhythm light.

See which plan fits on the pricing page, or start your 2-day free trial. If you only want a quick look first, the free AI visibility report checks 25 buyer questions and emails you a PDF.

Written by

AI search research

Share this article