Track leads and customers from AI in Close
Buyers increasingly start their research inside ChatGPT, Claude or Gemini, then click through to a business site and fill in a quote or call-booking form. Website analytics can show that visit, and AI visibility tools can show that a brand was mentioned in a chat, but neither tells you how many leads or paying customers that mention produced. This guide shows a four-step setup that carries the AI source all the way into Close CRM, then four reports that turn it into pipeline and revenue answers.
Why AI traffic needs to be tracked at the lead level
A new lead arriving in Close CRM raises an obvious question: did it come from a paid campaign, an organic search result, or a recommendation inside an AI assistant? The tools most teams already run answer a narrower question. Mention trackers count appearances in AI chats; web analytics counts sessions. Neither one closes the loop to a named lead, a won customer or a booked amount of revenue.
The gap matters because the numbers behave differently at each stage. A page can collect plenty of AI-referred visits and still convert poorly, while a smaller stream of AI referrals can close at a much higher rate because the visitor already had basic questions answered before landing. Only lead- and customer-level data in the CRM settles that.
The four-step setup at a glance
The workflow has one job at each stage, and each stage can be checked before moving on:
- Place the tracking snippet on the site and add four hidden fields to every lead capture form.
- Let the snippet classify each visitor into a channel and remember it in the browser.
- Capture the hidden field values with the form submission, alongside name, email and the rest.
- Push the submission into Close CRM as a Lead, where the AI data sits on the record and becomes reportable.
If a later report shows blank channel values, the break is almost always at step one (missing hidden fields) or step four (the form-to-CRM handoff), not in how the visitor originally arrived.
Step 1: Install the snippet and add hidden fields
Once a trial account is active, the setup begins with a small code snippet placed on the website; the installation instructions explain exactly where it belongs. The next task is to add four hidden fields to each lead capture form on the site, named precisely as follows so the values line up later in the CRM: Channel, Channel Drilldown 1, Channel Drilldown 2 and Channel Drilldown 3. This normally takes only a couple of minutes. Depending on the form builder, you may find a dedicated hidden field type to drag into the form, or you may need to add a standard text field and switch off its visibility in the settings panel. Written walkthroughs are available for Gravity Forms, Wix Forms, Webflow Forms and a wide range of other builders.
Step 2: How the visitor gets classified
On each visit, the snippet reads the same technical signals a tool like Google Analytics uses to determine a referrer. It assigns the visitor to a channel such as AI Referral, Organic Search, Paid Search or Paid Social, and stores that result in the browser.
The classification happens before any form is filled in, which is why a visitor who browses several pages and returns later still carries the original source into the form.
Step 3: Capture the data with the form submission
When the visitor submits a form, the values written into the hidden fields are saved by the form tool together with the name, email address and everything else the visitor typed. Supported form builders, scheduling tools and chat products number more than 50, including Gravity Forms, Contact Form 7, Wix Forms, Webflow Forms, Typeform, Jotform and Calendly.
Step 4: Send the submission into Close CRM
Many form builders connect natively to Close; where no native integration exists, Zapier can create a new Lead from the submission instead. Once the handoff works, the AI referral values sit on the Lead record next to the contact details.
Close's own reporting is a reasonable starting point. If you need to slice the data further, connect an analytics tool such as Looker Studio, Databox or Microsoft Power BI and build the reports there.
What data is captured, field by field
Two kinds of information travel with each lead: how the visitor reached the site (the channel) and which page brought them in (the landing page).
Take a roofing contractor. A buyer asks ChatGPT for a local roofer, clicks through and requests a quote. The hidden fields then hold: Channel = AI Referral; Channel Drilldown 1 = ChatGPT; Channel Drilldown 2 = chatgpt.com; Channel Drilldown 3 = None, because an AI referral has no deeper level to record.
The landing page data arrives separately: Landing Page = meridianroofing.com/services/roof-replacement and Landing Page Group = /services. That page-level detail is what reveals which pages AI tools actually cite, and therefore which content is worth expanding.
Four reports that answer the real questions
Each report answers a different question, and together they move from volume to money:
- Leads by Channel shows monthly lead counts split by source, so AI sits beside every other channel.
- Customers by AI tool splits new customers by which assistant sent them, so ChatGPT can be compared with Claude and Gemini on outcomes rather than mentions.
- Close rate by channel shows lead-to-customer conversion per channel, which often reverses first impressions from volume alone.
- Revenue from AI by landing page ties customers and revenue to the content that produced them.
A useful reading order is volume first, then conversion, then revenue. If a channel brings few leads but converts far better, the budget question changes.
Reading the close-rate report without being misled
Lead volume alone can point you the wrong way. Suppose AI Referral leads convert at 35%, Organic Search at 10% and Paid Search at 7%. On those numbers, a single AI-sourced lead carries roughly the same commercial weight as several paid search leads, so the conversion rate — not the raw lead count — is the better guide to where the next budget increment belongs.
Treat such figures as an illustration of the pattern, not a benchmark. Your own report is the only number that should drive a decision, and it needs enough closed deals per channel before the percentages mean anything.
A pre-launch checklist
Before trusting the first report, confirm each item:
- The snippet is present on every page that hosts a lead capture form.
- All four hidden fields exist on each form and are spelled exactly as Channel, Channel Drilldown 1, Channel Drilldown 2 and Channel Drilldown 3.
- A test submission from a known AI referral shows AI Referral in the Channel field and the tool name in Drilldown 1.
- The same test lead appears in Close with those values intact.
- A submission from a direct visit shows a channel value that reflects direct traffic rather than a blank field.
A blank channel on a real lead points to a setup fault, not to a visitor with no source.
What this setup does and does not tell you
It tells you how many leads and customers arrived via an AI tool, which tool sent them, which landing pages they first saw, and how their conversion and revenue compare with other channels.
It does not measure how often your brand appeared in AI answers without a click, nor what was said about you inside the chat. Those are separate visibility questions. The value here is the bottom of the funnel: the leads and customers the CRM can actually count.
| Field | Example value | What it tells you |
|---|---|---|
| Channel | AI Referral | Confirms the lead arrived from an AI tool rather than search or ads |
| Channel Drilldown 1 | ChatGPT | Identifies which assistant sent the visitor |
| Channel Drilldown 2 | chatgpt.com | Records the domain the visitor arrived from |
| Channel Drilldown 3 | None | No deeper level exists for an AI referral |
| Landing Page | meridianroofing.com/services/roof-replacement | The first page the visitor saw |
| Landing Page Group | /services | Groups pages for content-level reporting |
Example values follow the roofing scenario described in the guide; your own domains and tool names will differ.
| Report | Question it answers | Watch for |
|---|---|---|
| Leads by Channel | How many leads each channel produced per month | Volume alone can mislead about value |
| Customers by AI tool | Which assistant produced paying customers | Compare ChatGPT, Claude and Gemini on customers, not mentions |
| Close rate by channel | How well each channel converts leads to customers | Needs enough closed deals per channel to be meaningful |
| Revenue from AI by landing page | Which content generated customers and revenue from AI | A few pages often account for most of the revenue |
Report names and purposes follow the guide's examples; the specific charts depend on how you build them in Close or an external analytics tool.
Frequently asked questions
Do I need a separate AI tracking tool as well as this setup?
Not for lead and customer counts. AI visibility tools report mentions and analytics tools report sessions; this setup carries the referral into the CRM so leads, customers and revenue can be counted. If you also want to know how often you appear in AI answers without a click, that is a different measurement.
Why is Channel Drilldown 3 empty for AI leads?
An AI referral has no deeper level of detail to record, so the third drilldown comes through as None. The tool name sits in Drilldown 1 and the referring domain in Drilldown 2.
What if a lead in Close has no channel value at all?
That usually means the hidden fields are missing from the form, are misnamed, or the form-to-CRM handoff dropped them. Check the four field names on the form first, then submit a test lead and follow it into Close.
Can I report on this data outside Close?
Yes. Close's built-in reports are a starting point, and you can connect an analytics tool such as Looker Studio, Databox or Microsoft Power BI for deeper slicing.
Does this work with forms other than the ones listed?
The setup supports more than 50 form builders, scheduling tools and chat products. If your tool is not among them, check whether it can pass hidden field values through to Close, directly or via Zapier.