How to Track Leads That Came From AI Answers: Forms and Call Tracking

Why is AI lead tracking hard?

• Many answers need no click. Pew Research found that people who saw a Google AI summary clicked a link inside it in only about 1% of visits.
• Phone numbers travel without a referrer. When an assistant shows your number and the user taps it, no website visit happens.
• Tags are incomplete. ChatGPT adds utm_source=chatgpt.com to many outbound links, but analysts have found it is not on every link, and app traffic often arrives with no referrer.
• Journeys are long. A buyer may ask an assistant, then search your brand a week later, then call. Last-click analytics credits the brand search.
• Memory is imperfect. People who answer "how did you hear about us?" mix up sources.

The answer is triangulation: several imperfect measures that together give a defensible estimate.

What can be measured, and with what?

Can AI visibility increase business calls without increasing website traffic?

Yes, and for local and service businesses it is common. Assistants often answer "who should I call for [service] near me?" with a business name and phone number, drawn from Business Profiles, directories or websites. Google's AI features link to Business Profiles directly. Google's Business Profile performance report counts "calls" as the number of times a customer clicked the call button on the profile. That happens without any visit to your website. A business can see calls rise while web traffic stays flat, and the AI influence shows up only if callers are asked or the profile uses a tracking number.

How to add an AI referral question to lead forms

Add a required or optional field, "How did you first hear about us?", with options that separate AI from search:

• Google search
• Google Maps or Business Profile
• An AI assistant (such as ChatGPT, Gemini, Perplexity, Copilot or Claude)
• A Google AI summary (AI Overview or AI Mode)
• Social media
• Recommendation from someone
• Review site or directory
• Other

Then add a conditional follow-up when "AI assistant" is chosen: "Which one, and what did you ask?", as a short free-text field. Keep the list short, put AI as its own option rather than inside "search", and record answers in your CRM against the lead. Many teams find more AI mentions than their analytics suggested, because the field captures people who never clicked a link.

How to ask leads which page they saw cited by ChatGPT or Perplexity

People rarely remember a URL, but they often remember what the answer said. Ask sales or intake staff to note the exact phrasing:

• "What did the assistant say about us?"
• "Did it mention a specific article, review or page?"
• "What were you asking when we came up?"

Write the answer down word for word. Then reproduce the question in the same assistant to see which sources it cites and which of your pages, if any, carries that phrasing. Distinctive wording in a lead's memory, such as "they're the ones who do same-day repairs", often matches a specific page or review. That tells you which content is working and which third-party sources are shaping your description.

How can call tracking capture inquiries driven by AI recommendations?

Call tracking assigns phone numbers to sources, so calls can be attributed.

• Dynamic number insertion shows a different number to website visitors depending on how they arrived. Configure a number pool for visits with AI referrers or AI UTM tags. Calls from those visitors are then attributed to the AI source.
• Static numbers per channel can go on specific listings or campaigns, such as a directory profile that AI answers often cite.
• Call recording and keyword spotting, where legal and disclosed, can flag calls where the caller mentions ChatGPT or an AI assistant.

Check recording consent rules in your jurisdiction. Some states and countries require all parties to consent, and a disclosure at the start of the call is standard.

Can CallRail track forms and calls referred by LLMs?

Yes, within the limits above. CallRail announced in December 2024 that customers could track leads from AI-driven sources such as ChatGPT and SearchGPT, Perplexity and Gemini. It works by recognising those referrers and assigning tracking numbers and form attribution accordingly. In June 2026 CallRail added an integration for ChatGPT ads, matching calls, texts and form fills to ChatGPT ad campaigns. Like any tool, it can only attribute what arrives with a detectable source. Calls placed directly from a number shown in an answer still need a tracking number on the listing, or a question to the caller. Other call tracking platforms offer similar referrer-based rules. Check that yours recognises the AI referrers you care about.

How to track leads when ChatGPT gives users your phone number instead of your website

When an assistant shows your phone number, it usually comes from your Business Profile, a directory such as Yelp, or your own website. To attribute these calls:

1. Find where the number comes from. Ask the assistant for your contact details and check which source it cites.
2. Use a tracking number on that source, where the platform allows it. For Business Profiles, CallRail and Whitespark both recommend putting the tracking number in the primary phone field and your main number in the first additional phone field. That keeps your real number associated with the listing.
3. Keep your name, address and main number consistent everywhere else, so the tracking number does not fragment your business information.
4. Ask every caller how they found you, with an AI option.
5. Log the answers in your CRM or call tracking notes.

Avoid putting different tracking numbers on many directories without planning. Inconsistent phone numbers across sources can confuse both people and assistants.

How do I attribute a phone call after a Google AI Overview recommendation?

Directly, you usually cannot. Google does not report which AI Overview led to a call. What you can do:

• Put a tracking number on your Business Profile as the primary number. Calls from Google's AI Overviews, AI Mode and local results that link to the profile will be tagged as Business Profile calls.
• Watch Business Profile performance for trends in calls, website clicks and direction requests.
• Check Search Console's generative AI report for impressions of your pages in AI Overviews and AI Mode, and compare with call trends.
• Test the questions your customers ask and note when your business appears in an AI Overview or AI local pack.
• Ask callers, with "Google AI summary" as a distinct option from "Google search" and "Google Maps".

Together these show whether AI features are contributing, even if individual calls cannot be traced to a specific answer.

How medical practices can tell whether new patients found them through AI answers

Medical practices can use the same methods, with privacy care:

• Ask at intake: add "an AI assistant, such as ChatGPT" to the referral source question on new-patient forms and at the front desk.
• Use call tracking carefully. If calls are recorded or transcribed, patients may share health information. Use vendors that will sign a business associate agreement where HIPAA applies, or avoid recording.
• Keep tracking off patient portals and appointment forms. HHS guidance warns that tracking technologies on authenticated pages can disclose protected health information to vendors.
• Record the source as a category, such as "AI assistant", not with clinical details, in marketing systems.
• Track Business Profile calls for each location.

How healthcare providers can audit their AI visibility is covered in how healthcare providers audit AI search visibility step by step.

How to survey customers about AI search referrals and compare answers with traffic data

1. Ask consistently for at least two to three months, on forms and calls, with the same options.
2. Count self-reported AI leads as a share of all leads.
3. Count analytics-attributed AI leads, those with an AI referrer or tag, for the same period.
4. Compare the two. Suppose that of 120 leads in a month, 22 (18%) said they found you through an AI assistant, while analytics attributed 6 leads to AI referrers. Self-reported AI influence is then about 3.7 times what analytics shows.
5. Report a range: "AI assistants contributed to between 5% and 18% of leads", with the method explained.
6. Watch the trend in both numbers rather than obsessing over the gap.

The gap is expected. It reflects no-click influence, app traffic without referrers, and memory errors in both directions.

What mistakes distort AI lead data?

• Counting all "Google" answers as search. Without a separate option for Google's AI summaries, AI influence on Google disappears into search.
• Changing the form options mid-way, which breaks trends.
• Letting staff skip the question on busy days, which biases the sample towards quieter periods.
• Attributing everything to the last click, which hides AI's role early in the journey.
• Treating one month as a trend. Lead volumes are small for many businesses, so month-to-month swings are often noise.

How should AI lead data be reported?

A simple monthly view works:

• AI-referred sessions and their conversion rate.
• Leads with an AI referrer or tag, by form and call.
• Self-reported AI leads and the phrasing they quoted.
• Business Profile calls and direction requests, for local businesses.
• Visibility trends for your tracked questions, so lead changes can be read against visibility changes.

How to track AI referrals in analytics is covered in how to track ChatGPT and other AI referrals in GA4.

What might an illustrative case look like?

Consider an illustrative roofing company that saw almost no ChatGPT traffic in analytics but kept hearing callers say "ChatGPT recommended you". It added an AI option to its web form and trained staff to ask every caller how they found the business. It put a tracking number on its Business Profile as the primary phone, with the main line as an additional number, and configured its call tracking to recognise AI referrers. Over three months, 15% of leads said an AI assistant had pointed them to the company. Only 3% arrived with an AI referrer. Most of the rest called the number the assistant showed. The owner kept investing in the directory profiles and pages the answers cited, and reported AI influence as a range.

What do people ask about AI lead tracking?

Is the "how did you hear about us" question reliable?

Not perfectly, but it captures influence analytics cannot see. Use it alongside referrer data.

Should we use a separate number for AI?

On your website, yes, through dynamic insertion for AI-referred visits. On listings, one tracking number on the profile assistants cite most is usually enough.

Do call tracking numbers hurt local rankings?

Used as the primary number with your main number as an additional one, as call tracking providers recommend, the risk is low. Keep other listings consistent.

How long before the data is useful?

Two to three months of consistent collection usually gives a stable enough picture to act on.