Index
Why the product ad outperforms the generic ad
The four steps of the process
Step 1: prepare the car details page on the website
Step 2: include the car in the feed with the correct information
Step 3: create the ad group with the correct context hint for that car
Step 4: review the generated ad and adjust if necessary
How to handle high-value cars with their own campaign
When it makes sense to isolate a car in its own ad group
The text in the feed that determines if the car appears in the right conversation
What to do when the car is sold
Dealcar and managing the ad per car
Frequently asked questions
Why the product ad outperforms the generic ad
A generic ad for the dealership ("García Dealership — used cars in Valencia — visit us") can generate clicks from buyers with very different intentions. A product ad that shows a specific car (car photo, make, model, price, link to the details page) attracts buyers who are interested in that specific car or a very similar one.
The difference in the conversion rate from click to lead can be significant. The buyer who clicks on the ad for a specific car arrives at the details page with a clear expectation: to see that car, confirm the details, and decide whether to make contact. The buyer who clicks on a generic ad arrives at the dealership's homepage without knowing exactly what they are going to find.
Check the complete guide to ChatGPT Ads for dealerships.
OpenAI has indicated that ads generated from product feeds have historically been the best-performing ones on its platform. This is consistent with what happens on Google Shopping and Meta Catalog: product ads with an image and price of the specific item have higher CTRs and conversion rates than generic brand ads.
The four steps of the process
Advertising a specific car on ChatGPT Ads follows four steps that go from the vehicle details page on the website to the active ad in the Ads Manager.
The first is to prepare the car details page on the dealership's website: making sure it has its own URL, publicly accessible photos, a visible price, and a full description.
The second is to include the car in the product feed with the correct information: item_id, title, description, url, image_url, price, and availability as a minimum.
The third is to create the ad group in the Ads Manager with the correct context hint for the buyer profile that this car is targeted at.
The fourth is to review the preview of the automatically generated ad from the feed and adjust the texts if the display is not as expected.
Step 1: prepare the car details page on the website
Before including a car in the ChatGPT Ads feed, the vehicle details page on the dealership's website must meet five conditions.
It must have its own stable URL: not a dynamic URL that changes with each session or with each filter applied, but a permanent URL that always leads to the details page of that specific car. This URL is the one that goes into the url field of the feed and appears in the ad.
See also what a dealership website needs to have to generate leads.
The photos must be hosted on a publicly accessible URL, in JPG or PNG format, with good resolution. OpenAI processes the image from the URL included in the image_url field of the feed. If that URL is not accessible or the image is in an unsupported format, the ad may be generated without an image or with a fallback image.
The price must be visible on the details page without needing to click or fill out forms. A hidden price or one that requires interaction to view creates friction and reduces the conversion rate.
The description on the details page must answer the questions a buyer would ask after seeing the ad: mileage, maintenance history, condition of wear-and-tear items, included extras, and sales conditions.
The contact form or WhatsApp button must be visible on the details page without needing to scroll down. The buyer who clicks from a ChatGPT Ads ad is in the contact mindset: if the button is not immediately visible, some buyers will leave without contacting.
Step 2: include the car in the feed with the correct information
The item_id must be unique and invariable for each car. The VIN (chassis number) is the most robust identifier because it is globally unique and does not change. The number plate also works as an identifier if the dealership does not work with VINs.
The title must identify the car precisely and add the most relevant attribute for the target buyer profile. "Seat León FR 2.0 TDI 150HP DSG 2019" is better than "Seat León". If the car is from a lease with an official history, including "ex-lease" in the title can make a difference in CTR.
The description is the field that OpenAI's system uses to understand in which conversations that car is relevant. A well-written description includes mileage, fuel type, transmission, main extras, and sales conditions. "Automatic diesel family car, 67,000 km, serviced at official workshop, C-label, panoramic roof, rear camera, transfer included" gives the system much more context than "car in very good condition".
The availability field must be "in_stock" for the car to be eligible for ads. When the car is sold, it must be changed to "out_of_stock" in the next feed update.
Check the complete technical guide to the car product feed for ChatGPT Ads.
Step 3: create the ad group with the correct context hint
For a specific car, the context hint of the ad group must describe the buyer looking for that type of car, not the car itself. OpenAI's system uses the hint to determine in which conversations the ad is relevant; the car information comes from the feed.
The hint for a diesel automatic Seat León FR like the one in the previous example could be: "Driver who does mixed city and highway trips and is looking for a sporty automatic diesel compact, under 80,000 km and with a budget of between 15,000 and 22,000 euros."
Read also how to write ChatGPT Ads context hints for dealerships.
This hint describes who the buyer is (mixed-use driver), what features they prioritize (diesel, automatic, sporty trim), what mileage limit they have, and in what price range. The system can recognize that profile in conversations where the buyer describes that situation in different ways.
If the same ad group includes several cars of the same profile in the feed (filtered by ads_metadata), the context hint describes the common profile of all those cars. If the ad group is for a single high-value car, the hint can be even more specific.
Step 4: review the generated ad and adjust
OpenAI's Ads Manager generates a preview of the ad from the feed data before launching it. This preview is the view of the ad as the buyer will see it: image, title, description, and link.
There are three things to check in the preview. The first is that the image is correct and looks good in the ad format: no strange cropping, no overlapping text, and with the car centered. The second is that the title reads correctly and is not truncated. The third is that the description conveys the value of the car clearly within the available space.
If the title or description is not as expected, the correction goes to the feed: the title and description fields are updated in the source file and the feed is uploaded or synced again. The Ads Manager regenerates the ad from the new data in the next update.
The ad template in the Ads Manager only allows selecting which image field to use. The ad texts are automatically generated from the feed, which means that the quality of the title and description in the feed directly determines the quality of the ad.
How to handle high-value cars with their own campaign
For high-priced cars (above 25,000 or 30,000 euros) with margins that justify a larger advertising investment, it can make sense to create a specific campaign instead of including them in the dealership's general feed.
A dedicated campaign for a high-value car allows you to manage the budget independently, optimize the context hint specifically for the buyer profile of that car, and track the performance of that campaign in isolation.
The downside is the additional complexity: more campaigns means more management and tracking work. For a dealership with few very high-priced cars, it may be more practical to include them in the general feed with the high-end ad group context hint, rather than filtering by ads_metadata to isolate them.
The rule of thumb is to create a dedicated campaign when the additional management cost is justified by the car's margin. For cars over 35,000 euros with margins of 3,000 or 4,000 euros, a dedicated campaign with specific tracking makes sense. For cars at 18,000 euros with margins of 1,500 euros, the general feed with a segmented ad group is sufficient.
Read also how to calculate gross and net margins on used cars.
When it makes sense to isolate a car in its own ad group
In addition to high-value cars, there are other cases where isolating a car in its own ad group makes sense.
When the car has very specific characteristics that clearly differentiate it from the rest of the stock: a convertible, a large SUV, a high-range electric car. For these cars, the buyer profile is so different from that of the dealership's other cars that a generic hint does not serve them well.
When the car has been in stock for many days without generating leads from the general feed. Isolating it in its own ad group with a more specific context hint and a dedicated budget can be a way to give it more visibility without increasing the total campaign budget.
When there is a deadline to sell that car: end of the month, need to free up space in the lot, closing a purchase agreement. A dedicated ad group with more budget for a limited period can speed up turnover.
See also the stock quality control checklist for a dealership.
The text in the feed that determines if the car appears in the right conversation
The description field in the feed is the one that most influences whether the ad appears in the correct conversations. It is not the only factor (the context hint, the title, and the landing page also contribute), but it is the one that provides the most semantic information about the specific vehicle.
A description that works well for contextual matching is not a list of technical specifications: it is a description in language close to what the buyer uses. Instead of "Power: 150HP. Displacement: 1968cc. Max torque: 340Nm", it is better to have "Diesel with efficient traction, ideal for long road trips with actual fuel consumption of 4.8 l/100 km."
The terms that the buyer uses in their conversations with ChatGPT (family use, space for children, easy to park, low maintenance cost, no mechanical trouble) should appear in the description if they are relevant to that car. This alignment between the buyer's language and the feed's language improves matching relevance.
What to do when the car is sold
When the car is sold, there are two actions that must be executed immediately or in the next feed update.
The first is to change the availability field to "out_of_stock" in the feed. This stops the ad from being shown in new conversations. If the hosted URL is synchronized with the dealership's management system, this change can be automatic the moment the car is marked as sold in the DMS.
The second is to verify that the car details URL on the web no longer shows the car as available, or that it redirects to a "car sold" page or the general catalogue. If the car still appears as available on the web after being sold, buyers who click on the ad during the lag time between the sale and the feed update will reach incorrect information.
A feed update process of less than 24 hours from the sale of the car is the minimum goal to avoid generating clicks from buyers interested in cars already sold.
Dealcar and managing the ad per car
Dealcar automatically generates the product feed from the platform's inventory, with each vehicle's data already in the correct format for ChatGPT Ads. When a car is sold in Dealcar, the availability field is automatically updated in the feed without manual intervention.
You can see how it works at dealcar.io/dealcar-boost/chatgpt-para-concesionarios or request a demo at dealcar.io.
Frequently asked questions
Can I advertise a car that is still being prepared and is not yet available for sale?
Technically yes, using the value "preorder" in the availability field. But you have to manage expectations well: if the buyer clicks and the car is not ready for viewing, the lead might cool off. It is more convenient to activate the ad when the car is available for delivery.
Does the ad image have to be the same as the main photo on the details page?
Not necessarily. The image_url field in the feed can point to any publicly accessible image of the car. To maximize consistency between the ad and the landing page, the cleanest approach is to use the same main photo from the details page as the feed's image_url.
What happens if the feed has the same car with two different item_ids by mistake?
OpenAI treats them as two different products and may generate two different ads for the same car. This doubles the exposure but can also confuse the buyer if the two pages have slightly different information. You must ensure that each car has a single, invariable item_id across all feed updates.
