Skip to content

Track leads and customers from AI in Formidable Forms

Visitors increasingly arrive at your site after asking an assistant such as ChatGPT, Claude or Gemini for a recommendation. Those visits look like ordinary direct or referral traffic unless you deliberately record where they came from. This guide shows how to attach AI source data to each Formidable Forms submission, carry it into your CRM, and turn it into reports that answer a practical question: is AI actually producing leads and paying customers for you?

What you are measuring, and what you are not

A mention or citation inside an AI answer is not the same thing as a tracked lead. A citation tells you an assistant surfaced your content. A tracked lead tells you a specific person clicked through, reached your site, and submitted a form — and only the second one can be joined to a customer record and revenue.

The setup below produces the second kind of evidence. It records, on the form entry itself, the channel the visitor arrived through, the AI tool involved, the referring domain, and the first page they landed on. Everything downstream — CRM records, close rates, revenue — is built on that entry-level data.

Step 1: Install the tracking script and add six hidden fields

Begin with the site-wide script. Once it is in place, the remaining configuration happens inside each Formidable form you want to measure.

Add six Hidden Field controls to the form and give each one a label and a default value. The six fields are:

  • Channel
  • Channel Drilldown 1
  • Channel Drilldown 2
  • Channel Drilldown 3
  • Landing Page
  • Landing Page Group

In the WordPress admin, open Formidable Forms, select the form, and drag six Hidden Field elements into it. Click each field and set its label and default value in the options panel. The precise values for each field are published in the setup guide; alternatively, the WordPress plugin can create the whole configuration for you.

A common failure at this stage is adding the fields but leaving default values blank or mismatched, which produces entries that look tracked but carry no usable source data. Verify one test submission end to end before moving on.

Step 2: How the visitor's origin gets classified

The classification happens before the form is ever touched. As a page loads, the script reads the clues a visitor's arrival leaves behind — the same family of clues that conventional web analytics relies on — and uses them to decide which bucket that visit belongs in. The result is a channel label such as AI Referral, Organic Search, Paid Search or Paid Social, and it is parked in the visitor's browser rather than written to the server immediately.

Parking it there is the point. A visitor may browse several pages before converting, and the browser copy survives that journey. When the person finally submits a quote request or contact form, the stored label and its drilldowns are poured into the hidden fields, so the entry documents its own origin instead of depending on a later reconstruction.

Step 3: Read the data on the entry and forward it

Every submission becomes an entry that holds two kinds of information side by side: what the visitor typed (name, email, message) and where they came from (the six hidden field values). To see both together, open Formidable > Entries from the WordPress dashboard. If you want sales to see provenance the moment a lead lands, add the source fields to the notification email Formidable sends on each submission.

Getting the data out of WordPress is a separate decision, and there are three routes:

  • Native add-on: Formidable offers add-ons for destinations such as HubSpot, Salesforce and ActiveCampaign, which can carry the hidden field values across with each lead.
  • Automation platform: if your CRM has no native add-on, Zapier or Make can relay the fields on every new submission.
  • Export: a spreadsheet or email marketing platform can receive the data for lighter-weight reporting.

Whichever route you pick, confirm that each of the six fields is mapped to a real destination field. An unmapped field is silently dropped.

Step 4: Build the reports where the data lives

Reporting only becomes possible once the source fields have landed somewhere queryable. From that point you can count AI-sourced leads and customers, identify which assistant produced them, see which content preceded them, and total the revenue attached to them.

Three routes cover most situations. The first is the reporting already built into your CRM, which is enough for many teams. The second is a dedicated BI layer — Tableau, Google Data Studio or Microsoft Power BI — for teams that want custom dashboards. The third is conversational: connect Claude or ChatGPT to the CRM and ask a plain question such as how many leads came from AI referral this month. Pick the route that matches how often you expect to look at the numbers, not the most elaborate one available.

Field map: what each hidden field holds for an AI visit

The clearest way to understand the six fields is to walk a single scenario. A marketing lead at a metal fabrication business is recommended by ChatGPT, clicks through, and submits a quote request form.

  • Channel holds AI Referral — the top-level category.
  • Channel Drilldown 1 holds ChatGPT, the tool that sent the visitor.
  • Channel Drilldown 2 holds chatgpt.com, the referring domain.
  • Channel Drilldown 3 holds None, because AI referrals have no further level of detail.
  • Landing Page holds the first page visited, for example the laser-cutting capabilities page.
  • Landing Page Group holds the section of the site, for example /capabilities.

The first four fields describe how the visitor found you. The last two describe what content did the persuading. Keeping that distinction clear is what lets you separate channel performance questions from content performance questions later.

Four reports and the decision each one supports

Each report below answers a different question, and each should lead to a specific action rather than a general impression.

  • Leads by channel: a monthly tally of leads per channel, so AI referral can be sized against organic, paid search and paid social. The decision it supports is whether AI earns a dedicated line in your planning.
  • Customers by AI tool: new customers split by the assistant that produced them. The decision is where to concentrate optimization effort — the tool that dominates is the one your audience actually uses.
  • Close rate by channel: conversion frequency per channel. The decision is about lead quality rather than lead volume; a smaller channel that converts far better can outweigh a large, weak one.
  • Revenue from AI by landing page: customers and revenue attributed to the first page visited. The decision is what to publish next — when a handful of pages account for most AI revenue, extending that pattern is the obvious move.

Treat illustrative conversion figures as illustrations of the comparison, not as benchmarks for your business. Your own numbers are the only ones that should drive budget.

Pre-launch checklist

Before you rely on the numbers, confirm each of the following:

  1. The site-wide script is present on every page that hosts a tracked form.
  2. All six hidden fields exist on each form, with correct labels and default values.
  3. A test submission from an AI referral shows AI Referral in the Channel field and a tool name in Channel Drilldown 1.
  4. Channel Drilldown 3 correctly shows None for AI referrals rather than a blank or error value.
  5. Landing Page and Landing Page Group are populated and the group reflects your actual URL structure.
  6. Every hidden field is mapped to a destination field in your CRM or automation.
  7. Notification emails include the source fields so sales can see provenance immediately.
  8. At least one report is live before you start drawing conclusions from the data.

Where this setup commonly breaks

Most failures are configuration failures rather than tracking failures. The recurring ones:

  • Fields added without default values, so entries are stored but carry no source information.
  • Fields added to some forms but not others, producing partial coverage that understates AI's contribution.
  • Hidden fields not mapped in the CRM connector, so the data exists in Formidable but never reaches reporting.
  • Landing page groups that do not match the site's URL structure, which fragments the revenue-by-page report.
  • Reporting on AI leads only, without joining to customer and revenue records, which answers a volume question but not a value question.
Hidden field values for a ChatGPT-referred quote request
Hidden fieldExample valueWhat it tells you
ChannelAI ReferralTop-level channel category
Channel Drilldown 1ChatGPTThe AI tool that sent the visitor
Channel Drilldown 2chatgpt.comReferring domain
Channel Drilldown 3NoneNo further detail exists for AI referrals
Landing Pagekestrelfabrication.com/capabilities/laser-cuttingFirst page visited
Landing Page Group/capabilitiesSite section of the first page

Values shown are the illustrative example used to explain the field structure; your own entries will contain your domain and pages.

Reports and the decision each supports
ReportQuestion answeredDecision it informs
Leads by channelHow many leads per channel each monthWhether AI referral warrants its own planning line
Customers by AI toolNew customers split by assistantWhere to concentrate optimization effort
Close rate by channelHow often leads convert, per channelWhich channels produce quality rather than volume
Revenue from AI by landing pageCustomers and revenue per first pageWhat content to produce next

Report definitions describe the analysis structure; no performance figures are claimed for any business.

Frequently asked questions

Does this measure AI mentions or actual leads?

Actual leads. The data is attached to a form submission made by a real visitor, so it can be joined to a customer record and revenue. A mention or citation in an AI answer is a different signal and is not captured by this setup.

What appears in Channel Drilldown 3 for an AI referral?

None. AI referrals do not carry a further level of detail below the tool and domain, so that field is populated with None rather than left blank.

Do I need a native CRM add-on?

No. Native add-ons exist for destinations such as HubSpot, Salesforce and ActiveCampaign, but if your CRM is not covered, an automation platform like Zapier or Make can forward the fields with each new lead. A spreadsheet or email platform is also an option for lighter reporting.

Where do I see the captured data inside WordPress?

Go to Formidable > Entries in the dashboard. The source values appear with the rest of the entry, and you can also include them in the notification emails sent for each new lead.

Can I skip the manual field configuration?

Yes. The WordPress plugin can set up the hidden fields automatically. If you configure them by hand instead, the exact label and default values are published in the setup guide.

Which report should I start with?

Start with leads by channel to establish whether AI referral is material at all. Once volume is confirmed, move to close rate by channel, because a channel's value depends on conversion quality as much as lead count.