Lead Created (lead_created) is the standard conversion event you send to ChatGPT Ads when a visitor submits a lead form or requests a demo after clicking your ad. It records a lead submission specifically, so it sits below order_created in the funnel and should not be confused with registration_completed, which marks a finished sign-up.
If you run lead-generation campaigns in ChatGPT, this event is how you tell OpenAI which clicks turned into real prospects, and it is what a conversion-optimized campaign bids toward. Skip it and your campaigns optimize blind, spending budget on clicks that never become pipeline while you have no reliable count of the leads your ads produced.
Lead Created (lead_created) marks the moment a ChatGPT Ads data source records a completed lead action and reports it back to OpenAI for attribution. It is one of a fixed set of standard events OpenAI defines, and it carries the customer_action data shape; purchases use the contents shape instead.
To work, the event needs three things in place: a data source connected to your Pixel ID, a conversion event setting that names lead_created as the action to measure, and that setting attached to the campaigns you want to report on. You can fire it from the browser pixel, from the server-side Conversions API, or from both with a shared event ID so the same lead is not counted twice.
Within the OpenAI event taxonomy, lead_created sits alongside registration_completed and appointment_scheduled as customer_action events, each describing a different mid-funnel commitment. For marketers tracking how AI search turns into pipeline, AirOps helps connect that ChatGPT Ads signal to the organic answer-engine visibility work happening on the same accounts.
Resources: See how to set up ChatGPT Ads conversion tracking the right way
Here is the path a single lead takes from form submission to a number you can optimize against in ChatGPT Ads.
Create a source: Spin up a data source in Ads Manager and grab its Pixel ID, which identifies where your events come from.
Add the event: Fire lead_created with the customer_action shape when the form submits, through the browser pixel, the Conversions API, or both.
Define the setting: Create a conversion event setting that names lead_created and set an attribution window that fits your sales cycle.
Attach to campaigns: Link that setting to the campaigns you want to measure, or make it the optimization goal of a conversion-optimized campaign.
Deduplicate and attribute: OpenAI matches events by Pixel ID, event name, and shared event ID, then credits leads inside your attribution window.
The output tells you how many lead submissions each campaign drove and lets the system bid toward more of them. It does not tell you whether those leads became revenue; that still lives in your CRM.
Whether you keep funding ChatGPT Ads next quarter often comes down to whether you can prove the clicks produced real leads, and lead_created is the event that makes that case to your finance team with hard numbers.
Optimization depends on it: A conversion-optimized campaign bids toward the lead_created event you attach, so without it the system has no goal to learn from and spends evenly across weak and strong clicks alike.
Reporting gets honest: It gives you a lead count tied to actual ad clicks, so you can compare ChatGPT Ads cost per lead against Google and Meta on the same terms, instead of guessing from traffic alone.
Failure mode without it: Fire the wrong event, like registration_completed on a form that is really a lead, and your conversion counts, attribution, and bidding all train on the wrong action until the quarter's budget is already spent.
Demand gen leads use lead_created to tell ChatGPT Ads which form fills came from paid clicks, so campaigns can optimize toward booked pipeline.
Growth marketers use lead_created to compare cost per lead across ChatGPT, Google, and Meta before shifting budget between channels.
SEO managers use lead_created to connect AI-search ad conversions with the organic answer-engine visibility they already track for the same brand.
The Pixel ID names the data source every lead_created event reports to, and the same ID ties browser and server events to one campaign so OpenAI can match and deduplicate them across both paths.
Lead_created always carries the customer_action shape, a lighter payload than the contents shape, so you send the lead action without the amount and currency fields a purchase requires, so a basic lead form needs almost no extra data.
The attribution window is the span after an ad click during which a later lead_created event still gets credited to that click, and it decides how much delayed pipeline your campaign appears to earn, and too short a window hides leads that close weeks later.
Measures which ChatGPT Ads clicks become real leads instead of counting raw traffic.
Feeds conversion-optimized bidding so campaigns learn to chase qualified leads.
Standardizes cost per lead across ChatGPT, Google, and Meta for clean channel comparisons.
Separates lead_created from registration_completed so each funnel stage reports on its own.
Deduplicates browser and server events through a shared event ID so no lead counts twice.
Map each form to one event so a lead form fires lead_created and nothing else, keeping your conversion counts clean.
Send a stable event ID on both pixel and server calls, because that shared ID is what prevents the same lead being counted twice.
Set the attribution window to your sales cycle, since a window shorter than your typical lead-to-close time will undercount delayed conversions.
Use the Conversions API for server-confirmed leads, so ad blockers and lost browser sessions do not erase leads the pixel missed.
Verify events in monitoring before launch, because a misfiring tag caught on day one is cheaper than one found after a month of spend.
Attach the setting to every relevant campaign, so conversion reporting and optimization both see the same lead data.
Avoid making lead_created your optimization goal before it has collected enough clean conversions; bidding toward a sparse, noisy signal teaches the system the wrong pattern and wastes early budget.
AirOps: Connects your ChatGPT Ads lead_created results with organic answer-engine visibility, so you see paid and AI-search performance for the same brand in one workflow.
OpenAI Ads Manager: Creates the data source, defines the lead_created conversion event setting, and attaches it to the campaigns you want to optimize.
Google Tag Manager: Fires the lead_created pixel call on form submission and manages the shared event ID without hard-coding tags on every page.
Map your lead moments: List the forms and actions that count as a lead this week, and decide which single event each one should fire. No budget or tools needed, just a shared definition of a lead.
Create the data source: In OpenAI Ads Manager, create a data source and copy its Pixel ID, since every later step references that ID when sending or reporting events.
Fire the event: Add the lead_created call with the customer_action shape to your form's success state, through the pixel, the Conversions API, or both with a shared event ID.
Create and attach the setting: Define a conversion event setting for lead_created, set an attribution window, and attach it to the campaigns you want to measure.
Verify, then optimize: Confirm events arrive in monitoring, let a few genuine conversions accumulate, and only then switch a campaign to bid toward lead_created.
Lead Created (lead_created) is the standard ChatGPT Ads event that records a lead-form submission after an ad click.
You send it from the browser pixel, the Conversions API, or both, matched by a shared event ID.
Its accuracy is capped by your attribution window and by browser events that ad blockers and privacy settings drop.
Fire the wrong event and your bidding and reporting train on the wrong action for the whole campaign.
The biggest gains come from using it as a conversion-optimized goal once it has enough clean data to learn from.
They mark different funnel moments, so you should fire whichever one matches the action a visitor actually completed. Lead_created represents a prospect handing over contact details through a lead form, a demo request, or a sales inquiry, and it uses the customer_action data shape. Registration_completed represents someone finishing a sign-up, such as creating an account or registering for an event, and it also uses customer_action. The practical test is intent: a lead form that routes to your sales team is a lead_created, while a self-serve account creation is a registration_completed. Mixing them corrupts both counts, because a conversion-optimized campaign can only bid toward one event at a time and will learn from whatever you attach. If a single form genuinely does both, pick the event that reflects the outcome you care about most for that campaign, document the choice, and keep every similar form consistent so your reporting stays comparable over time.
It depends on which monitoring view you check and your attribution window, so treat the two timelines separately. For a quick sanity check, OpenAI's recent-events monitoring surfaces a sample of events from roughly the last several minutes, which lets you confirm a lead_created tag is firing right after you test the form. Attributed conversion reporting is slower and more conservative, because the system waits to tie events to clicks and to deduplicate browser and server copies that share an event ID. Delayed and offline leads can also arrive later through the Conversions API, so a count you read on day one may rise as the attribution window stays open. For practical planning, verify firing within minutes during setup, then judge real campaign numbers only after the window you configured has had time to close for most clicks. Avoid reacting to the first hour of data as if it were final.
Several things pull the two numbers apart, and most are expected gaps instead of errors. ChatGPT Ads only sees leads it can attribute to an ad click inside your attribution window, while your CRM captures every lead from every source, so organic, referral, and direct leads never appear in the ad count. Browser-only tracking loses events to ad blockers and privacy settings, which pushes the ad number lower unless you back it up with the server-side Conversions API. Deduplication by shared event ID also collapses a browser and server copy of the same lead into one, which is correct but can surprise you if you expected two. Timing adds more drift, since a lead your CRM logs today may land in the ad report later, or just outside the window entirely. Reconcile by matching on a shared ID and comparing only the ad-attributed subset against your CRM, and the gap usually makes sense.
You can influence it, but only indirectly, because lead_created records an action your visitors take and not one you set. What you control is everything upstream of the submission: the audiences and keywords your ChatGPT Ads reach, the promise in the ad, the landing page, and how short and trustworthy your lead form is. Tightening that path lifts the share of clicks that end in a lead, which is the rate you actually move. You also control measurement quality, which changes the number you see even when true demand is flat: adding the Conversions API recovers leads the browser pixel drops, and a correct attribution window stops you undercounting delayed conversions. What you cannot do is fire the event yourself to inflate it without a real submission behind it, since that poisons the bidding model you depend on. Treat the event as a scoreboard, and spend your effort on the funnel that feeds it.
There is no single benchmark, so the honest answer is that good means better than your own baseline and cheaper than your other channels. ChatGPT Ads is new enough that public conversion-rate norms are thin, and rates swing hard by offer, industry, price point, and how qualified a lead you demand, so a borrowed benchmark will mislead you. Set your own instead: measure cost per lead_created and the click-to-lead rate for your first few weeks, then treat that as the line to beat. Compare it against the same lead definition on Google and Meta, because a fair read is relative to your own costs, never an absolute figure. Watch the trend as the conversion-optimized bidding gathers data, since the rate should improve as the model learns which clicks convert. If cost per lead sits at or below your blended target from other paid channels and holds steady, that is good; if it drifts up while volume stays flat, revisit the form and the audience before the budget.