How to Actually Measure Your AI Search Visibility and Performance
Asaf Fybish · March 2026 · 7 min read
Most AI visibility tools give you a score. Not leads, not traffic, not revenue. A score.
From our experience working with 50+ B2B clients on SEO and GEO, we only care about one thing: can you trace actual leads back to AI search? Almost no one can answer that today. Not because the data doesn't exist. Because no one set up the right systems to capture it.
Here's how we do it at Growtika. Six methods, all free except one.
TL;DR
Why most AI visibility tools mislead you
Before the methods, it helps to understand how the popular tools work. Because once you see it, you can't unsee it. (For a full breakdown of what each tool actually does, see our AI SEO tools review.)
The standard approach is this: take your keywords, use an LLM to generate hundreds of related questions, send those questions to the ChatGPT or Gemini API, check if your brand appears in the answers, and turn that into a score.
How most AI visibility tools generate their score
Seed keywords
LLM generates questions
Tool asks ChatGPT those questions
This score measures how often you appear in questions nobody actually asked. No real buyer is in this pipeline.
The problem isn't the tools. It's what the score represents. A high score can coexist with zero AI-sourced leads. We've seen it. These tools are useful for competitive benchmarking. Not for measuring business impact. If you're wondering whether AI visibility even matters when you already rank well on Google, this guide explains why it does.
Don't report a visibility score to leadership as proof GEO is working. That question will eventually follow: "how many leads came from AI?" You need a real answer for that.
The 6 methods we actually use
Lead Form Qualification
Your lead form already asks "how did you find us?" Most companies list Google, LinkedIn, Referral, Other. Just add the AI platforms to that list.
When a prospect picks "ChatGPT" on their way in, you know exactly where they came from. No guessing. Make sure this field flows into your CRM from day one or the data disappears. For context on what buyers are actually asking AI before they reach your form, see our guide to buyer AI queries.
How did you hear about us?
Select all that apply
Sales Team Qualification
Enterprise buyers skip forms. They book calls directly. They write "Google" out of habit. Your sales team catches what forms miss.
Train BDRs and AEs to ask this on every first call: "How did you first hear about us? Was it Google, a colleague, or maybe an AI tool like ChatGPT or Perplexity?"
Naming the options gets more accurate answers. Log responses in a structured CRM field, not freetext notes. Freetext can't be filtered or reported on.
"How did you first hear about us? Was it Google, a colleague, or maybe an AI tool like ChatGPT or Perplexity?"
Log answer in CRM field: Source → AI Platform → [ChatGPT / Perplexity / Claude / Copilot / Grok / Gemini]
CRM Attribution Mining
HubSpot logs the referring domain for every contact. When someone clicks a link to your site from ChatGPT or Perplexity, that gets recorded automatically. Most teams never look at it.
Filter Contacts by Original Source = Referrals, export with the "Drill-Down 1" field, and filter for chatgpt.com, perplexity.ai, claude.ai. You may already have AI-sourced leads in your CRM that nobody noticed.
| CONTACT | SOURCE DRILL-DOWN | DEAL STAGE |
|---|---|---|
| Sarah Chen | chatgpt.com | Closed Won |
| Marcus Webb | perplexity.ai | In Negotiation |
| Dana Park | google.com | Discovery |
GA4 Referral Dashboard
When ChatGPT or Perplexity include a link to your site in their answer and someone clicks it, that shows up in GA4 as referral traffic. Most teams never filter for it.
Add this regex as a custom channel in GA4 called "AI Search". Set it up once and it applies to all your reports going forward.
chatgpt\.com|perplexity\.ai|claude\.ai|copilot\.microsoft\.com|bing\.com\/chat|gemini\.google\.com|grok\.x\.com| CHANNEL | SESSIONS | CONV. RATE |
|---|---|---|
| AI Search (new) | 847 | 4.2% |
| Organic Search | 12,340 | 2.8% |
| Direct | 8,910 | 1.9% |
Bing Webmaster Console
Bing has 3% of consumer search. Most people ignore it. That's a mistake, because Bing powers Microsoft Copilot, which is installed by default on every Windows machine in most enterprise IT environments.
Bing Webmaster Console is the only free tool that shows you the actual questions users typed before your page was cited. Real prompts, not generated ones. Setup takes 10 minutes. If you're finding that Copilot cites your competitors instead of you, here's why that happens and how to fix it.
| QUERY | IMPR. | CITED |
|---|---|---|
| what's the best siem tool for a 50 person security team? | 1,240 | 38 |
| how do i choose between splunk and alternatives in 2025? | 980 | 27 |
| we failed an iso audit. what security tools should we look at? | 640 | 19 |
These are verbatim user questions. No AI tool that fires fake prompts will ever show you queries like the third one.
Ahrefs AI Overviews Data
We use Ahrefs at Growtika. What most teams don't know is that Ahrefs now shows you which keywords trigger a Google AI Overview, whether your page is cited in it, and how many extra clicks that citation drives.
The most useful filter: keywords where you rank in the top 10 but are not cited in the AI Overview. That's your optimization list. You already have the authority, the page just isn't structured in a way AI picks up as worth citing. Our ChatGPT SEO guide covers how to structure content so AI models actually cite it.
| KEYWORD | POS. | VOL. | AI OVERVIEW | AI CLICKS |
|---|---|---|---|---|
| best SIEM tools for startups | 4 | 2,400 | Cited | +142 |
| cloud security posture mgmt | 7 | 5,900 | Cited | +89 |
| endpoint detection comparison | 3 | 1,100 | Not cited | -- |
| siem pricing enterprise | 2 | 880 | Cited | +67 |
Row 3 is ranked position 3 but not cited in AI Overview. That's your optimization target.
AI clicks are additive. A page at position 7 that gets cited can outperform position 2 without citation.
What's coming in the next 12 months
The measurement tools are catching up. Here's what we're watching:
- OpenAI's Search Console is expected to launch. When it does, you'll see which ChatGPT prompts cited your pages and how often. The teams already measuring today will have baseline data to compare against.
- Google Search Console is expanding AI Overview reporting. Right now it's thin. It's getting better.
- Perplexity is building out publisher attribution. Better referral data is coming.
Start measuring now. When these tools launch, you'll have historical data. Everyone starting fresh after the launch will be flying blind by comparison.
No single method covers everything. Forms miss enterprise leads. GA4 misses direct traffic. Bing covers one platform. Start with the lead form and GA4 dashboard this week. Add the others over the next month. When multiple methods point to the same trend, you can trust it. If you want help setting all of this up, let's talk.
Frequently Asked Questions
Asaf Fybish
Building organic growth engines for startups | SEO, AI Search & Relevance Engineering. Co-Founder & GM at Growtika.