Index
The problem of a single context hint for varied stock
How to segment stock for ChatGPT Ads
The three most useful dimensions of segmentation
How to use ads_metadata to segment without separate feeds
Recommended ad group structure for a stock of 20-50 cars
How to allocate budget between segments
Context hints by segment: examples for each profile
How to manage cars that do not fit into any segment
Budget management when stock rotates
What to learn from the first 30 days with varied stock
Dealcar and automatic feed segmentation
Frequently asked questions
The problem of a single context hint for varied stock
A dealership with heterogeneous stock using a single ad group with a generic context hint for the entire inventory is being imprecise with targeting. A hint like "people looking for second-hand cars" can appear in thousands of different conversations, but many of them will be from buyers looking for something completely different from the available stock.
The result is a low CTR overall (ads appear in many conversations but are relevant to few) and a mediocre click-to-lead conversion rate (buyers who click arrive at a car that does not match what they described in their conversation).
The solution is segmentation: dividing the stock into homogeneous groups of cars with a common buyer profile and creating a different ad group for each group, with the context hint designed for that specific profile.
Check the complete guide to ChatGPT Ads for dealerships.
How to segment stock for ChatGPT Ads
Stock segmentation for ChatGPT Ads has one main criterion: grouping cars that share the same buyer profile, not those that share the same make or model.
A Volkswagen Polo and a Renault Clio of similar price and profile can be in the same ad group because the buyer looking for one may be interested in the other. A Volkswagen Golf and a Volkswagen Touareg from the same manufacturer can be in different ad groups because the buyers represent completely different profiles.
Segmentation by buyer profile generates ad groups with more precise context hints and more relevant cars from the feed for those hints. This translates into better relevance for OpenAI's matching system and, potentially, a better cost per click by gaining relevance.
The three most useful dimensions of segmentation
For a multi-brand dealership with varied stock, there are three dimensions of segmentation that cover most cases in a practical way.
The first is price range, which is the filter most closely correlated with the buyer profile. Buyers of cars under £10,000 have different needs, expectations, and decision processes compared to buyers of £20,000 to £35,000 cars. Three price ranges (budget, mid-range, and premium) cover most medium-sized dealership stocks.
The second is body style or use, which captures specific buyer needs. A buyer who needs a large boot for trips looks for an estate or SUV, regardless of price. A city buyer looking for a first car wants a compact hatchback that is easy to park. These needs are expressed in conversations with ChatGPT in very recognisable ways and can be captured with specific context hints.
The third is the origin of the stock, which determines the most powerful selling point. Ex-leasing cars with documented history have a different selling point compared to cars from private owners. Cars with ECO or zero-emission labels have a different selling point compared to conventional diesel or petrol cars. These differences in origin and environmental features can justify separate ad groups if the volume of that segment warrants it.
How to use ads_metadata to segment without separate feeds
The ads_metadata field in the OpenAI feed allows you to add custom tags to each car in the inventory. These tags are then used as filters in the Ads Manager to define which cars enter each ad group, without the need to upload separate feeds for each segment.
The practical implementation is simple. In the feed, each car receives one or more key-value pairs in the ads_metadata field. For example: segment=budget for cars under £10,000, segment=mid-range for cars from £10,000 to £20,000, segment=premium for cars over £20,000. In the Ads Manager, when creating each ad group, the filter is applied using the corresponding ads_metadata value.
The result is that a single feed serves all the dealership's ad groups, with each displaying only the cars that correspond to that segment. When a new car enters the stock, you simply assign it the correct ads_metadata value, and it will automatically be included in the correct ad group.
Check the technical guide for the car product feed for ChatGPT Ads.
You can combine multiple tags per car: segment=mid-range|type=suv|origin=leasing allows you to segment by price, body style, and origin simultaneously, creating more specific ad groups for cars that warrant it.
Recommended ad group structure for a stock of 20-50 cars
For a dealership with between 20 and 50 cars in stock of varied profiles, a structure of three to five ad groups covers most cases without excessive complexity.
A budget ad group (cars under £10,000 or £12,000) with the context hint targeted at first-time car buyers, buyers on a tight budget, and city use. Filter the feed by segment=budget.
A mid-range ad group (cars from £12,000 to £22,000) with the context hint targeted at families, professionals with intensive use, and buyers moving from new to used cars. Filter the feed by segment=mid-range.
A high-end ad group (cars over £22,000) with the context hint targeted at premium used buyers and businesses with small fleets. Filter the feed by segment=premium.
Optionally, a specific ad group for SUVs and estates if that body style represents a significant part of the stock and has a very defined buyer profile. Filter the feed by type=suv or type=estate, combined with the corresponding price segment.
And optionally, an ad group for ex-leasing cars or cars with documented history, where the main selling point is the traceability of maintenance. Filter the feed by origin=leasing.
Read also how to advertise used car stock on ChatGPT Ads.
How to allocate budget between segments
The allocation of budget between ad groups depends on two variables: the potential profit margin per sale in each segment and the expected stock turnover in each segment.
A high-end ad group with £30,000 cars and margins of £3,000 or £4,000 justifies more investment per lead than a budget ad group with £8,000 cars and margins of £600 or £800. However, the budget ad group may generate more lead volume with the same budget.
One way to start is to distribute the budget proportionally to the number of cars in each segment, and adjust that distribution after 30 days of data based on the actual cost per lead of each ad group. If the mid-range ad group generates leads at £30 and the high-end generates leads at £80, but the high-end leads convert to a sale twice as often, the cost per sale may be similar and the distribution makes sense.
See also how to measure ChatGPT Ads leads and sales in a dealership.
What does not make sense is to allocate the entire budget to the segment with the highest profit margin per sale without first verifying if that segment has enough relevant conversation volume in ChatGPT for the dealership's geographical market.
Context hints by segment: examples for each profile
For the budget range: "Person looking for their first car with a maximum budget of £10,000, prioritising reliability and low maintenance costs over performance. Also older buyers who want something simple and easy to park for daily city use."
For the mid-range, family profile: "Family with children who need more space than a hatchback provides and are comparing used SUVs or estates between £12,000 and £22,000 with a good maintenance history."
For the mid-range, professional profile: "Sole trader or sales representative doing more than 25,000 miles a year and looking for a reliable and economical second-hand diesel or hybrid between £12,000 and £20,000 to reduce running costs."
For the high-end: "Buyer with a budget between £22,000 and £40,000 who prefers a 2- or 3-year-old premium used car over a brand new mid-range car for the same price, valuing safety and comfort features."
For SUVs and estates: "Family that uses the car for daily school runs and weekend trips, looking for a second-hand SUV or estate with space for four or five people, a large boot, and reasonable fuel consumption."
For ex-leasing cars: "Buyer who values official documented maintenance history and prefers to pay slightly more for the certainty of an ex-company or ex-leasing car rather than risking a private sale without history."
Read also the complete guide to ChatGPT Ads context hints for dealerships.
How to manage cars that do not fit into any segment
Every stock has cars that do not fit perfectly into any of the defined segments: a convertible, an MPV, a large 4x4, a classic car. For those cars, there are three options.
Read also when to trade-in, keep, or liquidate stock in a dealership.
The first is to include them in the ad group closest to their profile, even if it is not perfect. A mid-priced convertible can go in the mid-range ad group even if its buyer profile is different. If the volume of that type of car is low, the imprecision of the hint may be acceptable.
The second is to create a specific ad group just for that car or that type of car, if the margin justifies it. A premium £40,000 convertible may warrant its own ad group with a very specific context hint for buyers of that type of vehicle.
The third is to exclude it from the feed by setting the is_ads_eligible field to false. If the car does not have a clear buyer profile on ChatGPT or if the advertising investment for that specific car is not justified, it is better not to include it in the ads and leave it only on the classified portals.
Budget management when stock rotates
A multi-brand dealership's stock changes constantly: new cars arrive, others are sold, prices are adjusted. This affects the number of cars in each ad group and, potentially, the optimal budget distribution.
If the mid-range ad group goes from 15 cars to 8 because several were sold, the same budget generates a higher CPM per car. And if new cars enter the high-end range during that period, it may make sense to reallocate budget towards that ad group.
The frequency of reviewing budget distribution depends on the speed of stock turnover. For a dealership selling 20 or 30 cars a month, a monthly review of the distribution is sufficient. For one with very high or very low turnover, it could be fortnightly or quarterly.
What must be monitored in real-time is that the availability field in the feed updates correctly when a car is sold: if an ad group has 8 cars in the feed but 3 are already sold without updating the availability, 37% of the budget is spent on advertising unavailable cars.
What to learn from the first 30 days with varied stock
With 30 days of data from a campaign with several ad groups, there are three questions you can answer that will guide optimization for the following month.
The first is which ad group has the best CTR. A high CTR indicates that the ad is relevant to the conversations where it appears. If the mid-range ad group has a CTR of 2.8% and the budget ad group has a CTR of 0.9%, the mid-range cars are fitting better with the conversations.
The second is which ad group has the best click-to-lead conversion rate. A high CTR with low conversion indicates that buyers are clicking but the landing page or the car does not meet their expectations. A low CTR with high conversion indicates that the ad is highly selective: it appears rarely but converts well when it does.
The third is whether there are cars in the feed that have not generated a single click in 30 days. These cars may have an issue with the title or description in the feed that makes them less relevant to the system's matching. Reviewing those fields and comparing them with cars that do generate clicks can reveal which feed attributes improve performance.
Dealcar and automatic feed segmentation
Dealcar generates the ChatGPT Ads product feed from the platform's inventory with the segmentation fields (ads_metadata) automatically assigned based on each vehicle's profile. The dealership does not need to manually tag each car: the platform classifies them and the resulting feed allows you to create the recommended ad group structure directly in the OpenAI Ads Manager.
You can see how it works at dealcar.io/dealcar-boost/chatgpt-para-concesionarios or request a demo at dealcar.io.
Frequently asked questions
How many ad groups are too many for an independent dealership?
Above five or six ad groups, management becomes complex without performance improving proportionally, especially with a limited overall budget. With less than £300 per month in total, more than three ad groups may not have enough budget per group to generate useful data. It is better to start with two or three well-defined ad groups and add more when you have data that justifies the additional segmentation.
Can I use the same context hint for multiple ad groups with different feed filters?
Technically yes, but it does not make sense. If two ad groups have the same hint but different filtered feeds, the OpenAI system treats them as equivalent candidates for the same conversations. The goal of having different ad groups is precisely to have different hints that capture different buyer profiles.
Does it make sense to segment by make instead of by price range?
For a multi-brand dealership selling cars from many different manufacturers, segmenting by make is less efficient than segmenting by buyer profile. A buyer looking for a mid-range estate may be equally interested in a Volkswagen Passat or a Ford Mondeo. Segmenting by usage profile and price captures this breadth much better than segmenting by make.
