# Is ChatGPT Recommending Your Company? How to Measure AI Search Visibility, Traffic and Leads (2026)

> AI assistants now shortlist vendors before a buyer ever visits a website. Here is how to measure it in three layers: are you mentioned, does traffic arrive, and which leads came from it.

Source: https://knkoutbound.com/blog/ki-sichtbarkeit-messen/  
Author: Kenneth Kather, Founder & CEO, KNK Outbound  
Published: 2026-11-06

## Key takeaways

- AI visibility is measured in three layers: whether assistants mention you (prompt tests), whether visitors arrive (referral traffic), and whether leads come from it (a source question in every form and first call).
- In GA4, a custom channel group with a source rule for chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com and claude.ai separates AI visitors from ordinary referrals. App traffic often arrives without a referrer, so the real number is higher.
- Search Console shows the other side: long, conversational queries are often searches an AI assistant runs in the background. In our own export, about 40 percent of the top 1,000 queries looked like prompts.
- The most reliable lead signal is the simplest: ask. Our first meeting sourced by AI search came through a booking form field that asked how the person had heard of us.

## The short answer

Yes, you can measure how much visibility, traffic and how many leads come from ChatGPT and other AI assistants, but not with one number. You need three layers: a monthly prompt test that checks whether assistants mention you, a channel in your analytics that separates AI referrals from other traffic, and a source question in every form and first call. Each layer misses something on its own. Together they give a reliable picture.

It matters because the buying process has moved. A B2B buyer who asks an assistant for "the best providers for X" gets a shortlist of three to five names before visiting a single website. If you are not on it, you never see that buyer in your analytics at all. The underlying question of why assistants pick some companies and not others is covered in [why ChatGPT does not recommend your company](https://knkoutbound.com/blog/get-recommended-by-chatgpt/).

## Layer 1: Are you mentioned? The prompt test

Analytics only sees people who click. To see whether you are recommended, you have to ask the assistants yourself, systematically.

**Collect 20 to 30 real buyer prompts.** Do not invent them. Take them from three sources: the questions prospects ask in first calls, the long questions in your Search Console export, and the wording your customers use in emails. A good prompt sounds like a person: "Which agency handles email and LinkedIn outreach including booking meetings in B2B?"

**Run them once a month** in ChatGPT, Perplexity, Gemini, Copilot and Claude, ideally logged out or in a fresh chat without history. Answers vary from run to run, so run each prompt three times and count how often you appear.

**Record four things per prompt:**

| Field | What you note | Why it matters |
|---|---|---|
| Mentioned | Yes or no, in how many of three runs | Your share of answers |
| Position | First name, middle, last | Earlier names get chosen more |
| Description | What the assistant says about you | Wrong descriptions need fixing at the source |
| Sources | Which pages it cites | Shows which content and which third-party sites carry you |

The source column is the most useful. If assistants cite a trade portal, a comparison list or a directory instead of your site, that is where the next lever is.

## Layer 2: Does traffic arrive? Analytics

**Google Analytics 4.** By default, AI visits end up in Referral and get lost among other sites. Create a custom channel group (Admin, Data display, Channel groups), add a channel called AI assistants with the condition that the session source matches this regular expression: chatgpt.com|chat.openai.com|perplexity.ai|gemini.google.com|copilot.microsoft.com|claude.ai. Then move the channel above Referral, because GA4 assigns each session to the first matching channel. A custom channel group also applies to data collected before you created it.

ChatGPT often adds utm_source=chatgpt.com to the links it shows. That helps, but traffic from the mobile apps frequently arrives without any referrer and lands in Direct. So read the AI channel as a minimum, not the full number.

**Google Search Console.** Since June 2026, Search Console has a separate performance report for generative AI features. It groups AI Overviews and AI Mode together and cannot separate the two. The ordinary performance report holds a second clue: long, conversational queries. When an assistant searches the web to answer a prompt, it often sends very long questions that no human would type into Google. In our own export for August to October 2026, about 400 of the 1,000 most frequent queries looked like this, for example "which agency does lead generation and appointment setting in B2B?". These queries tell you which prompts you are already a source for.

**Bing Webmaster Tools.** ChatGPT search also draws on the Bing index. Bing Webmaster Tools shows how your pages perform there and has its own AI performance report for Copilot. If you only have Search Console, add Bing via the import from Google, it takes a few minutes.

## Layer 3: Which leads came from AI? Ask

The most reliable method is the oldest one. Put a field "How did you hear about us?" in every contact and booking form, with an option such as "AI assistant (ChatGPT, Perplexity, etc.)" and a free-text field. Ask the same question in the first call and store the answer in a source field in your CRM.

This catches what analytics cannot: the buyer who saw your name in a ChatGPT answer, searched for your company on Google two days later and booked through your homepage. In analytics that is brand search. In reality it is AI. Our own first meeting from AI search, in August 2026, came in exactly this way: the booking form answer said "AI research", and analytics would have counted it as a direct visit.

## What to report every month

Keep it to three numbers and their trend:

- **Mention rate:** in what share of test prompts and runs you are named
- **AI sessions:** visits in the AI assistants channel, plus the number of conversational queries in Search Console
- **AI-sourced leads:** inquiries and meetings whose stated source is an AI assistant

Do not expect clicks to grow as fast as mentions. Assistants answer the question themselves, so many buyers only click when they are close to a decision. A low click count with a rising mention rate is normal. Leads are the number that counts.

## How to improve the numbers

Measurement only helps if something changes afterwards. The pattern we see in our own data: pages that answer a real buyer question in the first two sentences, with concrete numbers and a clear statement of who it is for, get cited. Generic service pages do not. The full approach is in our guide to [generative engine optimization](https://knkoutbound.com/blog/generative-engine-optimization-geo/), and how to turn search and AI visibility into steady inquiries is in [predictable B2B inquiries from Google and AI search](https://knkoutbound.com/blog/b2b-anfragen-ueber-google-und-ki/).

## FAQ

### Can you measure how much traffic and how many leads come from ChatGPT and other AI assistants?

Yes, in three layers. A monthly prompt test shows whether assistants mention you, a custom channel group in GA4 separates AI referrals such as chatgpt.com and perplexity.ai, and a source question in every form and first call shows which leads came from AI. App traffic often has no referrer, so analytics shows a minimum.

### How do I set up an AI traffic channel in Google Analytics 4?

Create a custom channel group under Admin, Data display, Channel groups. Add a channel named AI assistants with the condition that the session source matches chatgpt.com|chat.openai.com|perplexity.ai|gemini.google.com|copilot.microsoft.com|claude.ai, and move it above Referral. It also applies to past data.

### How do I find out whether ChatGPT recommends my company?

Ask it, systematically. Collect 20 to 30 real buyer prompts, run each one three times a month in ChatGPT, Perplexity, Gemini, Copilot and Claude in a fresh chat, and record whether you are mentioned, in which position, how you are described and which sources are cited.

### How do I know which leads found me through ChatGPT?

Ask them. A field 'How did you hear about us?' with an AI assistant option in every form, plus the same question in the first call, stored in your CRM. Many AI-sourced buyers later search your name on Google, so analytics alone counts them as brand or direct traffic.

### Does Google Search Console show AI Overviews and AI Mode?

Since June 2026 there is a separate performance report for generative AI features, which groups AI Overviews and AI Mode together and cannot separate them. Long conversational queries in the ordinary performance report are a second signal, because they often come from AI assistants searching the web.

## About KNK Outbound

KNK Outbound builds and runs outbound systems for B2B companies in Europe and the US: cold email, LinkedIn, content and ads, fully done for the client, including reply handling. From EUR 3,300 a month, all-inclusive (tools, data, domains, mailboxes), three-month build phase, then monthly. Pricing: https://knkoutbound.com/pricing/ · Book a call: https://knkoutbound.com/book-a-call/ · Agent access (MCP): https://knkoutbound.com/mcp