Chasing Cars brings OpenAI technology to car review platform

OpenAI’s technology has been integrated with car review platform Chasing Cars, harvesting data from its own “walled garden of information”.

Jonathan Kerr, chief growth officer, Auto & General

Auto & General-owned insurance company Budget Direct rolled out a new generative AI feature on its car content review platform Chasing Cars last Tuesday, joining a wave of brands integrating the tech tool into their customer-facing operations.

Budget Direct purchased Chasing Cars, an independent car review platform, in 2020 as part of a strategic move to get the Budget Direct brand into the consumer buying cycle of new cars.

The new feature, called ChasingCarsGPT, uses OpenAI’s technology, a content-generating tool which has captured the attention of the general public as well as executives the world over since it was made ­available for mainstream use in November last year.

It comes as part of a rising number of brands applying generative AI tools to their marketing, creative and advertising endeavours globally, from Coca-Cola to HSBC.

The Chasing Cars website houses more than 5000 videos and articles and can also attract up to 20,000 unique visitors in a day. Its video content is particularly popular with audiences, with some ­videos attracting up to 130,000 individual views.

Jonathan Kerr, chief growth ­officer at Auto & General told ­The Growth Agenda: “If I had one frustration with Chasing Cars, it was because there was just such a wealth of content, but it was difficult to leverage everything in a very easy interface to bring it back up.”

Since the advertising-free Chasing Cars platform was acquired, it has retained its editorial independence. Mr Kerr said the thinking behind introducing ChasingCarsGPT was to build on its audience’s trust in the website by generating responses derived entirely from its own content.

In this instance, users can prompt ChasingCarsGPT with questions about car makes and models, which the tool then responds to, harvesting data from a “walled garden of information” – all of which is owned and/or housed within Chasing Cars’ website.

ChasingCarsGPT uses OpenAI's technology to answer specific customer questions

The tool generates answers exclusively based on Chasing Cars articles and transcriptions of its own video review content. As part of the answers, it also provides links to the primary source of the answers and, where applicable, recommends Chasing Cars videos.

Drawing on Chasing Cars’ owned content helps overcome some challenges around the responsible use of generative AI, as the answers generated on ChasingCarsGPT platform come from content the brand already owns, rather than the internet at large. This can also help instil trust in the information it serves audiences, according to Mr Kerr.

The information it draws on – and content it produces – does not contribute to OpenAI’s pool, or other data lakes.

It’s also not just a sophisticated chatbot. ChasingCarsGPT continually scores and improves data that it thinks is “the most logically and statistically ­correct answer to a question it has never seen before”, Mr Kerr said.

The key difference between a chatbot and is that this type of AI, is that ChasingCarsGPT understands the meaning and intent of the questions asked, rather than serving the more “binary” pre-approved responses that chatbots produce. The potential to respond to audience questions is exponentially larger than those that can be done manually.

Within the first two days after the tool was launched, ChasingCarsGPT clocked 10,000 question and answer interactions. In contrast, Mr Kerr said the capacity of its team could typically handle approximately 50 manual replies to similarly phrased questions on a single YouTube video in a day.

This is not to say ChasingCarsGPT will replace human-led interactions with customers – but as those initial numbers suggest, the tool meets an existing demand and an opportunity to assist more Australians research their next car purchase at a greater scale.

Mr Kerr said many of the answers to customers questions were likely to exist within its published news coverage.

“Human answers are going to have more nuance, but typically, as most of these things become, they are 80-20 rules,” Mr Kerr said.

“Yes, we will continue to answer those questions on our video content. Yes, you’re going to have very interesting conversations with people that are quite nuanced. And they require new original opinions. But boy, I reckon you can get through 80 per cent with what’s ­already available and just making it available via this interface.”

It will also help improve the searchability of articles housed within the Chasing Cars content ecosystem.

“If the user doesn’t know it’s there, they don’t know how to search for that specific article. Whereas if you ask a question, it just comes up with an amazing answer that then gives you both an answer and then the source of the answer.”

With all of these user-generated questions – and personalised responses – will naturally come a wealth of data about what audiences are genuinely interested in. Mr Kerr cautioned: “You better get your data and your sources in great shape before you let it loose”.

The tool could also offer customer insights that could help guide Chasing Car’s content strategy. While the new generative AI tool recommends articles and offers answers based on content that already exists, “The thing that’s cool about this is you can ask questions that we haven’t thought of,” Mr Kerr added.

Because the answers ChasingCarsGPT serves up also provide relevant links from its own content library with more information about a specific answer or car recommendation, from a traffic-driving standpoint there is a strong commercial imperative.

“We want to create a flywheel of activity,” Mr Kerr said. This includes serving content to audiences from its archives, such as its YouTube videos, and directing traffic back to content that might ordinarily be discovered only via other search tools.

“It is a self -feeding system,” he added.

The appetite of that system appears to be a voracious one indeed.