Index
Why measurement in ChatGPT Ads is a specific challenge
The four layers of measurement
Layer 1: the OpenAI pixel and how to install it
The oppref: the click identifier that connects the conversation with the conversion
Layer 2: the Conversions API for offline and server events
Layer 3: UTMs and Google Analytics for on-site behaviour
Layer 4: the post-conversion survey for the dark funnel
Which events to set up for a dealership
How to connect ChatGPT Ads data with the CRM
The metrics that matter and the ones that don't
What to expect in the first 30 days of data
Dealcar and built-in attribution
Frequently asked questions
Why measurement in ChatGPT Ads is a specific challenge
Every digital advertising channel has its attribution gap. Google Ads has one because the buyer can see the ad on their phone and buy from their computer days later. Meta has one because the ad can influence the decision without the buyer clicking directly.
ChatGPT Ads introduces an additional challenge that doesn't exist on Google or Meta: the conversation gap. The buyer holds a conversation with the AI in which they evaluate options, get information and form their purchase intent, all within the chat and without reaching any website. Only afterwards, once they are clear about what they're looking for, do they click on the ad.
Check the complete guide to ChatGPT Ads for dealerships.
That means the click on the ChatGPT ad can come at the end of a decision process that has already happened in the conversation. The buyer who clicks has far more context than someone who clicks on a display banner. And measurement has to capture what happens after that click, because everything that happened before in the conversation is not trackable.
The four layers of measurement
A robust measurement chain for ChatGPT Ads at a dealership has four layers that complement each other.
The first is the OpenAI pixel: it records the events that happen on the dealership's website after the click and connects them with the ChatGPT Ads campaigns using the oppref as an identifier.
The second is the Conversions API: it sends conversion data from the dealership's server or from the management system, which includes events the pixel cannot capture (phone calls converted in the CRM, sales closed days after the visit).
The third is Google Analytics with UTMs: it connects ChatGPT Ads traffic with user behaviour on the website (pages viewed, time on listing, other listings viewed) and allows that behaviour to be compared with that of other channels.
The fourth is the post-conversion survey: it asks the buyer at the time of the visit or purchase how they found out about the dealership. It captures the dark funnel that no tracking tool can measure: the influence of ChatGPT Ads on buyers who saw the ad but converted through another channel.
None of the four layers gives the complete picture on its own. All four together give a reliable enough view to make investment decisions.
See also ChatGPT Ads vs Google Vehicle Ads for dealerships.
Layer 1: the OpenAI pixel and how to install it
The OpenAI pixel is a JavaScript snippet installed in the head of every page of the dealership's website. Its function is to capture events (vehicle listing visits, form submissions, WhatsApp clicks) and send them to OpenAI's Ads Manager to attribute them to the active campaigns.
The basic installation requires three steps. First, create a data source in OpenAI's Ads Manager: in the account panel, in the Measurement section, you create a new source and get the Pixel ID. Second, add the base snippet to the head of the website: the code initialises the pixel with that Pixel ID and enables automatic page view tracking. Third, add the conversion events on the relevant pages or actions: the contact form submission, the click on the WhatsApp button and the visit to the appointment confirmation page are the three minimum events for a dealership.
The pixel automatically captures the oppref from the landing URL and stores it in a first-party cookie called __oppref. That cookie connects ChatGPT ad clicks with later conversions in the same session or in future sessions on the same device.
To check that the pixel is working correctly, you can enable debug mode during installation, which shows the events being sent in the browser console.
Centralise all your ChatGPT Ads leads in the Dealcar Lead CRM.
The oppref: the click identifier that connects the conversation with the conversion
The oppref is the identifier that OpenAI adds to the destination URL when someone clicks on a ChatGPT Ads ad. The URL reaches the dealership with a parameter like this: www.ejemplo.com/ficha-coche?oppref=gAAAAAb123.
The OpenAI pixel captures that parameter automatically and stores it in the __oppref cookie. When the same visitor converts later on (submits the form, clicks on WhatsApp), the pixel sends the stored oppref along with the conversion event. OpenAI can then attribute that conversion to the specific ad that generated the original click.
There is an important technical consideration: if the dealership's website uses redirects between the landing URL and the vehicle listing, the oppref can be lost if the redirects don't preserve it. You need to check that the oppref is kept in the final URL after any redirect before the pixel can capture it.
For measurement without the pixel installed, the oppref parameter can also be used as a filter in Google Analytics to identify ChatGPT Ads traffic: any session that arrives with oppref in the URL comes from ChatGPT Ads.
Layer 2: the Conversions API for offline and server events
The OpenAI pixel is the easiest way to install tracking, but it has limitations. Browsers can block third-party cookies, ad blockers can block the pixel script, and the pixel cannot capture conversions that happen outside the browser: a phone call that ends in an appointment, a dealership visit confirmed by the salesperson in the CRM, or a sale closed days after the first click.
The Conversions API solves those limitations by sending conversion data directly from the dealership's server or from the management system to OpenAI's server, without going through the user's browser. It is more reliable than the pixel because it doesn't depend on the browser environment.
For a dealership without a technical team, the Conversions API can be implemented through Google Tag Manager in server mode (sGTM), which has specific templates for OpenAI Ads available for free. With sGTM set up, every event the client-side GTM captures (form submitted, WhatsApp click) is also sent to OpenAI's server without going through the user's browser.
For a dealership that uses Dealcar as its management system, CRM events (lead created, appointment confirmed, sale closed) can be sent to OpenAI through the Conversions API once Dealcar has the integration available, which makes it possible to measure the channel's real return without the need for additional technical infrastructure.
When pixel and Conversions API are used at the same time, you must include the same event_id in both events so that OpenAI can deduplicate and not count the same conversion twice.
Layer 3: UTMs and Google Analytics for on-site behaviour
UTMs are parameters added to the ads' destination URLs to identify the source of traffic in Google Analytics. For ChatGPT Ads, the standard setup is: utm_source=chatgpt, utm_medium=paid, utm_campaign=nombre_de_la_campaña.
With those UTMs, Google Analytics can segment the traffic arriving from ChatGPT Ads and show its behaviour: which listings they visit, how much time they spend on each, whether they visit additional listings and whether they reach the contact page.
Read also how to use web analytics in a dealership to make decisions.
That information complements the data in OpenAI's Ads Manager. The Ads Manager shows performance from the delivery side: impressions, clicks, CTR and attributed conversions. Google Analytics shows post-click behaviour: what those visitors do when they reach the website.
Comparing the on-site behaviour of ChatGPT Ads traffic with that of other channels (portals, Google VLA, direct) reveals whether ChatGPT leads are of higher or lower quality: whether they spend more time on listings, whether they view more cars per session, whether they have a lower bounce rate.
Layer 4: the post-conversion survey for the dark funnel
The dark funnel is the part of the influence process that no tracking system can measure. A buyer may have seen a ChatGPT Ads ad without clicking, had a conversation with ChatGPT in which the dealership was mentioned, and arrived at the website directly or via Google days later. No tracking tool records that path.
The most effective way to capture that influence is to ask directly: when the buyer gets in touch, visits the dealership or buys the car, ask them how they found the business. A simple question ("how did you find us?") with options such as Google, car portal, ChatGPT, someone's recommendation, social media or other captures that information without technical complexity.
For that survey to be useful, the answer has to be recorded in the CRM together with the lead or the sale. With enough accumulated answers, the percentage that mention ChatGPT as a discovery channel gives a picture of the channel's real impact, which tracking data systematically underestimates.
Which events to set up for a dealership
There are five conversion events that make the most sense to set up for a used car dealership, ordered from easiest to hardest to implement.
The vehicle listing visit is the easiest event to set up: when someone lands on the page of a specific car. It isn't a high-intent conversion on its own, but it lets you measure how many ChatGPT Ads clicks reach an actual listing versus bouncing off the home page.
The contact form submission is the channel's main conversion. When the buyer submits the form with their details, there is a qualified lead attributable to the ad. This event is set up on the form's confirmation page.
The click on the WhatsApp button is the form equivalent for buyers who prefer messaging to forms. If the website has a WhatsApp button on every listing, the click on that button should be set up as a conversion event.
The appointment or visit request, if the website has an online booking system, is a high-intent conversion: the buyer has chosen a car and proposed a time to see it.
The phone call generated from the website can be set up with Google Tag Manager if the phone number on the website is dynamic (it changes depending on the source channel) or with an external call tracking service that generates unique phone numbers per channel.
How to connect ChatGPT Ads data with the CRM
The connection between ChatGPT Ads data and the dealership's CRM is the step that completes the attribution chain. Without that connection, you know how many clicks the channel generated but not how many of those clicks ended in a lead, a visit and a sale.
The most direct method is to tag each lead in the CRM with its source channel at the moment it arrives. If the lead comes from the website form and the URL has the oppref parameter, the channel is ChatGPT Ads. If it comes via WhatsApp with the UTM utm_source=chatgpt, the channel is ChatGPT Ads. If they call by phone and the number is the ChatGPT Ads tracking number, the channel is ChatGPT Ads.
With that channel tag in the CRM, it's possible to calculate the lead-to-visit and visit-to-sale conversion rates per channel. If ChatGPT Ads leads have a visit rate of 40% and a closing rate of 25%, the channel generates one sale for every 10 leads. If the cost per lead is 50 euros, the cost per sale is 500 euros.
Read also how to manage leads in a dealership to increase sales.
That number is what lets you compare ChatGPT Ads with other channels and decide how much to invest in each one.
The metrics that matter and the ones that don't
The metrics that OpenAI's Ads Manager offers immediately (impressions, reach, CTR, CPM, CPC) are delivery metrics, not business metrics. They show that the channel is working technically, not that it is generating sales.
The metrics that matter for a dealership are: cost per qualified lead (a lead with contact details who has shown interest in a specific car), lead-to-visit rate (what percentage of leads end up visiting the dealership), visit-to-sale rate (what percentage of visits end in a purchase) and cost per sale attributed to the channel.
See also how to measure and reduce the cost per lead in a dealership.
Delivery metrics are useful for diagnosing channel problems: a very low CTR indicates that the ad or the context hint isn't relevant to the conversations where it appears; a very high cost per click may indicate that the bid is poorly calibrated; a high bounce rate on the website indicates that the landing listing doesn't match what the buyer expected to find.
But the final judgement on whether the channel deserves investment is made with business metrics, not delivery metrics.
What to expect in the first 30 days of data
The first 30 days of a ChatGPT Ads campaign should be understood as a calibration phase, not an optimisation phase. The channel is new, historical benchmarking data for automotive in Spain is scarce and the system needs time to learn which cars generate the most contextual relevance.
What you usually see in the first 30 days is high variability between ad groups and between cars in the feed: some vehicles generate many impressions and few clicks (the ad appears but doesn't convince), others generate few impressions but a good CTR (the ad appears in very relevant conversations). That variability is useful information for the next optimisation phase.
With 30 days of data and a budget of 200 to 400 euros, you can start to answer: which context hints generate the most clicks, which type of cars in the feed have the highest CTR, what the real cost per click is in that geographical area and whether the leads arriving are of sufficient quality to justify continuing.
You can't conclude whether the channel generates sales in 30 days, because the buying cycle for a used car can exceed that period. The most useful indicator in the first month is whether the leads are of sufficient quality to justify following them up.
Dealcar and built-in attribution
Dealcar is developing the ChatGPT Ads integration, which includes automatic tagging of leads with their source channel in the CRM. When a lead comes from ChatGPT Ads, Dealcar's CRM will record it as such, making it possible to calculate the lead-to-visit and visit-to-sale conversion rates per channel without additional manual work.
You can see how it works at dealcar.io/dealcar-boost/chatgpt-para-concesionarios or request a demo at dealcar.io.
Frequently asked questions
Do I need to install the OpenAI pixel to launch a ChatGPT Ads campaign?
It isn't mandatory to launch the campaign, but without the pixel conversions aren't attributable and the system can't optimise towards conversions. For campaigns with a clicks objective, you can launch without a pixel. For campaigns with a conversions objective (oCPC), the pixel or the Conversions API is required.
Can I use Google Tag Manager to install the OpenAI pixel?
Yes. OpenAI has documentation for installing the pixel through GTM, and there are community templates available for server-side GTM that make it easier to implement the Conversions API.
How do I know if a phone call came from ChatGPT Ads?
With a dynamic tracking phone number that changes depending on the visitor's source channel, or with an external call tracking service that assigns unique numbers per channel. Without that infrastructure, calls can't be attributed to any specific digital channel.
Is the ChatGPT Ads attribution window the same as Google Ads'?
Not necessarily. OpenAI documents a click-through window for attribution and a one-day view-through window when it's available for the account. The exact details can be checked in the conversion settings of OpenAI's Ads Manager.
