Measuring AI Search Visibility

AI search visibility is how often, and how accurately, AI assistants such as ChatGPT, Gemini, Perplexity, Microsoft Copilot and Google's AI Overviews mention and cite a brand when people ask questions about its category. Measuring it takes three parts: a fixed prompt list checked every month, a few simple metrics, and tracking of visits from AI tools in GA4.

• Mention rate: The share of tracked prompts where the brand is named in the answer.
• Citation rate: The share of tracked prompts where the brand's own website is linked as a source.
• Share of voice: The brand's mentions as a share of all brand mentions across the tracked answers.
• Accuracy: How many answers state something wrong about the brand, such as an old price or a false claim.
• AI referrals: Visits and sales from AI assistants, measured in Google Analytics 4.

This lesson follows one example: a family business in Lucknow that sells hand embroidered chikankari kurtas, sarees and dupattas through its own website and ships across India. Shoppers now ask AI assistants questions like "where to buy authentic chikankari kurtas online" and "how to tell real hand chikankari from machine work", and the owner wants to know whether the store appears in those answers and whether that is improving.

Why AI Search Visibility Matters

• No built-in report: Most AI assistants do not tell businesses how often they were shown or named, so visibility has to be measured by checking answers directly.
• Rankings do not tell the whole story: A page can rank well in Google and still be missing from AI answers, or the reverse.
• Mistakes cost sales: An assistant that says the store "only sells machine embroidery" can turn buyers away, and only regular checks will catch it.
• Proof of effort: Work on getting cited by ChatGPT and on brand mentions in AI search needs numbers to show whether it is working.

Step-by-Step Framework to Measure AI Search Visibility

Step 1: Build a Fixed Prompt List

Write 30 to 50 prompts in the words customers actually use, and group them by type:

• Category: "Where to buy authentic chikankari kurtas online"
• Problem: "How to tell real hand chikankari from machine work"
• Comparison: "Chikankari from a Lucknow store vs a marketplace seller"
• Brand: "Is [store name] genuine?" and "[store name] return policy"
• Language: A few prompts in Hindi or Hinglish, if customers search that way, such as "Lucknow ki asli chikankari kurti kahan milegi".

Keep the list fixed. You can add new prompts later, but report them separately so the trend stays comparable.

Step 2: Fix the Checking Method

AI answers change with account history, location, wording and even between two runs of the same question. To keep checks fair, use the same method every month: a signed-out or fresh session where possible, the same location, web search turned on where the tool offers it, and each prompt run twice. Check Google's AI Overviews and AI Mode on a phone as well, since the layout can differ from a laptop. The workings of these tools are explained in how Google AI Overviews and AI Mode work.

Step 3: Record Each Answer

For every prompt and tool, record whether the brand is named, whether the site is linked, where the brand appears in the answer (first, among others, or last), the tone, any wrong facts, the competitors named, and the pages cited. The cited pages show which sources to work on next.

Step 4: Calculate Four Metrics

Mention rate    = prompts where the brand is named / prompts checked
Citation rate   = prompts where the brand's site is cited / prompts checked
Share of voice  = mentions of the brand / mentions of all brands in the answers
Accuracy issues = number of answers with a wrong fact about the brand

Calculate each metric per tool and in total, and compare it with the previous month rather than with anyone else's numbers, because every business's prompt list is different.

Step 5: Track AI Referrals in GA4

When someone clicks a link in an AI answer, GA4 usually records the tool's domain as the session source, such as chatgpt.com or perplexity.ai. ChatGPT often adds a UTM tag to its links as well. Group these visits into one channel by creating a custom channel group in GA4 under Admin > Data display > Channel groups, adding a channel called "AI assistants" with the condition "Session source matches regex", and placing it above the Referral channel. The basics of GA4 are in the GA4 tutorial.

^(.*\.)?(chatgpt\.com|chat\.openai\.com|perplexity\.ai|gemini\.google\.com|copilot\.microsoft\.com|claude\.ai)$

Then compare this channel's sessions, key events and revenue with other channels each month. Note that some visits from AI apps arrive with no referrer at all and show up as direct traffic, so these numbers are a minimum, not the full picture.

Step 6: Add Search Console and Server Logs

In Google Search Console, watch clicks and impressions for the pages targeted by the prompt list, remembering that AI Overviews and AI Mode traffic is counted inside the web totals. Bing Webmaster Tools may also report how often the site is cited in Microsoft's AI answers. Ask the developer or host for a monthly count of visits by AI crawlers such as OAI-SearchBot and PerplexityBot in the server logs, which confirms they can reach the site.

Step 7: Report Monthly and Act

Share a one page report each month: the four metrics, AI referral sessions and sales, the wrong facts found, and the three actions for next month. Every action should come from the data, such as "pitch the two blogs cited most often" or "fix the return policy page that assistants misquote".

Template: AI Search Visibility Tracking Sheet

Copy these columns into a spreadsheet, with one row per prompt, per tool, per round. The rows below show how entries look; they are not real results.

Add a second tab for the monthly summary: the four metrics per tool, AI assistant sessions and revenue from GA4, and the actions chosen. A connected dashboard can be built later with the Looker Studio tutorial.

Example: A Lucknow Chikankari Store Measures Its Visibility

• Prompt list: The owner writes 40 prompts from customer WhatsApp chats and Instagram messages, including eight in Hinglish.
• First round: The store is named in some problem questions, such as spotting real hand embroidery, because its guide page is cited, but it is missing from almost all "where to buy" answers.
• Wrong fact: One assistant says the store does not accept returns, based on an old marketplace listing. The team updates the listing and makes the return policy page clearer.
• GA4 channel: The new AI assistants channel shows a small but steady number of sessions, and the team watches whether they turn into orders.
• Actions: The team contacts the fashion blogs cited for "where to buy" questions with accurate details about its artisans, and adds a clear "Where to buy genuine chikankari" section to its site.
• Next rounds: The same 40 prompts are rechecked monthly, and the owner reads the trend over several months instead of reacting to one result.

Mistakes to Avoid

• Changing the prompts every month: The trend becomes meaningless when the questions change.
• Checking once: A single run of a prompt can be unusual, so run each one twice.
• Only brand prompts: Asking "is [store name] good?" always names the brand, which says nothing about discovery.
• Trusting one tool's score: Tracking tools use their own prompts and methods, so compare their trend with a hand checked sample.
• Ignoring accuracy: Being named with wrong facts can be worse than not being named.
• No link to actions: A report that does not lead to next month's tasks is just a record.

How AI Changes Measuring AI Search Visibility

What AI Automates Now

Tracking platforms can run hundreds of prompts across several assistants on a schedule, extract brand names and cited URLs, and chart mention rates over time. General AI assistants can also read a pasted set of answers and fill in the tracking sheet columns, which saves hours of copying.

What Still Needs a Human

A person must choose prompts that reflect real buying questions, judge whether an answer is accurate, and decide which actions deserve the team's time. Only the owner knows that "machine chikankari" in an answer is a serious error for this store.

Risk to Watch

AI tools that fill in the sheet can misread an answer, for example counting a competitor with a similar name as the brand. Spot check a sample of rows every month, and never paste customer names or phone numbers into AI tools while preparing reports.

Do It with AI

Use this prompt to turn raw AI answers into rows for the tracking sheet. It works in ChatGPT, Claude or Gemini.

1. Run the month's prompts in each tool and save the full answers.
2. Paste them into the prompt in batches of ten.
3. Check every row marked "check" and a random sample of the rest by hand.
4. Copy the rows into the tracking sheet and update the monthly summary.

Check Before You Use It

• Facts: Recount the metrics for one tool by hand to make sure the AI calculated them correctly.
• Brand fit: Decide yourself whether an answer's tone and description match how the brand wants to be seen.
• Compliance: Keep personal data out of the answers you paste, and follow the company's data policy on which tools may be used.