Top five generative AI use cases for LinkedIn Ads
ARTICLE
LLMs, stable diffusion, ChatGPT, Dall-E… The buzz around generative AI has spiked during early 2023 with the wave of new technology being released. In this blog, we are exploring the top five concrete use cases for generative AI for LinkedIn Ads.
While generative AI has use cases for complex campaign management such as recommendations, campaign structure generation, reporting summarization etc., the easiest path to profit is to improve workflows around creatives & content. Engaging content is key to marketing success and content creation is often one of the most labor-intensive steps in LinkedIn advertising.
1. Ad copy versioning
In order to find the most engaging content and combat ad fatigue, advertisers should always run multiple creatives concurrently. Versioning is an easy way to increase the probability to hit the sweet spot of performance. Large language models (LLM) excel at understanding the key points of a given text, and they can summarize and re-write the message while keeping the core message the same. Here is an example of how to get copy text versions with ChatGPT 4.0, a LLM published by OpenAI.
This is the prompt given to ChatGPT 4.0 with standard operating parameters:
“Make three versions of the following ad copy targeted towards B2B businesses. The ad is served on LinkedIn social media. Keep the tone of voice similar. The length should be 150 to 200 characters.
With Ambine’s advanced features, B2B marketers can effortlessly handle and optimize LinkedIn campaigns like a ninja with superpowers.”
It is important to describe the desired output with enough detail for the best results. Fine-tuning the instructions given to models such as ChatGPT is known as “prompt engineering”.
And the model outputs the following:
“1. Boost your B2B marketing strategy with Ambine’s powerful LinkedIn campaign optimization tools. Unleash your inner ninja today! 💼
2. Master LinkedIn campaigns like a pro with Ambine’s cutting-edge features, designed for B2B marketers seeking superpowers. 🌟📈
3. Elevate your B2B marketing game on LinkedIn with Ambine’s innovative solutions. Optimize campaigns effortlessly & achieve superhero results! 🚀🎯”
The model accurately captured the essential content of the original copy and kept a playful and superlative tone of voice. I think these copies are ready to go!
Ambine Ad Creator includes a ready-to-use ad copy versioning tool powered by GPT 4.0 to instantly create ad versions and publish them directly on your LinkedIn Ads campaigns. We offer a free 14-day trial for all new clients to use the Ad Creator with AI. Learn more about Ad Creator.
2. Asset generation
An engaging ad should have interesting visual elements to capture attention and stand out. Finding suitable graphics online is tedious, and sourcing ad assets from designers comes at a price. While not as mature as LLMs, image-generation AI is already useful for creating building blocks for engaging ads.
In this example, I made two ads using OpenAI’s “Dall-E 2” image model and the Midjourney image model. The background images are made using the AI, and text is pasted on top using generic tools.
The prompt given to each AI was:
“Background for a business-to-business ad. The color theme is a sunny yellow. No text. Ad an interesting 3D object in a contrasting purple color to capture attention. The image should have a feeling of the future.”
And here are the responses (Dall-E, Midjourney):
Dall-E captured the sense of an “ad background” better, whereas Midjourney prioritized the style and colors.
Here are ads based on my favorite responses:
3. Translation
International business does not imply that everyone speaks English. Localizing content to the target market is essential for engagement. Translation services can be pricey and slow down the process. LLMs are sufficiently capable in multiple languages for efficient translation. In this example, ChatGPT 4.0 is used to translate the copy text from before to other languages.
Here is the prompt I used:
“Translate the following business-to-business ad copy text to Finnish, Swedish, and German. Instead of a direct translation, try to localize the message so that it fits each language. Avoid using any words that are not typical in the local languages.
Master LinkedIn campaigns like a pro with Ambine’s cutting-edge features, designed for B2B marketers seeking superpowers. 🌟📈”
And here is the response:
Finnish:
“Hallitse LinkedIn-kampanjoita kuin ammattilainen Ambinen huippuluokan ominaisuuksilla, jotka on suunniteltu B2B-markkinoijille, jotka etsivät ylivoimaisia kykyjä. 🌟📈”
Swedish:
“Behärskar LinkedIn-kampanjer som en proffs med Ambines banbrytande funktioner, utformade för B2B-marknadsförare som söker överlägsna krafter. 🌟📈”
German:
“Beherrschen Sie LinkedIn-Kampagnen wie ein Profi mit Ambines bahnbrechenden Funktionen, entwickelt für B2B-Marketer, die nach Superkräften suchen. 🌟📈”
The results make sense and are grammatically correct. Some fine-tuning of the prompt could make the language sound more natural in each translation. We are developing an automatic localization function within the Ambine Ad Creator to easily localize your ads to your target markets.
4. Copies from content
Writing ads is typically the final step to launching a new campaign. Marketers would have a landing page, a white paper, a brochure, or other materials already at hand. Generative AI can be used to summarize and write the entire ad copy based on existing material. In this example, I used a prompt to first summarize the contents of a web page and then to write a suitable ad copy out of it. Some language models are already capable of reading website metadata to extract the contents, but first, scraping the main content could be more beneficial.
“Summarize the contents of the following web page https://ambine.com/linkedin-ads-performance-review-report/ and then write a suitable ad copy to promote the service listed on the webpage. Keep the ad copy between 150 and 200 characters. Use a professional tone of voice and highlight the main points.”
And the response:
“Boost your LinkedIn ad campaign success with Ambine’s Performance Review Report! Get expert analysis, actionable insights, and increased ROI. Try it now!”
Ambine Ad Creator already has an instant tool to make ad copies from a desired website. The tool allows you to select up to three different ad versions made from the content and publish them as ads directly to your LinkedIn Ads campaigns.
5. Image versioning
We’ve seen that the ad copy versioning can be automated easily. How about the image? Here is where the models still have to mature a bit more. Dall-E has a beta of image editing, where an existing image can be edited, but the capabilities for ads are still not enough for reliable use. Especially the models struggle with generating text within an image, often producing hilarious outputs. Here is an example where an existing ad was fed to Dall-E and Midjourney, and the prompt was the same as previously:
“Background for a business-to-business ad. The color theme is a sunny yellow. No text. Ad an interesting 3D object in a contrasting purple color to capture attention. The image should have a feeling of the future.”
Dall-E is more geared towards maintaining the existing image, whereas Midjourney took only some ideas from the existing image. Once the general-purpose image AIs develop into supporting multiple use cases, we can really use them to power asset versioning.
Final words
Generative AI can already be used by anyone to make LinkedIn advertising workflows easier. Ambine Ad Creator already integrates with GPT 4.0 and provides copy generation based on written instructions or a web page URL, as well as versioning out of the box. We plan to add localization as well as more fine-tuned control over the desired output to make the Ad Creator a true power tool for all LinkedIn advertisers.
Want to try it by yourself?
We offer you a free 14-day trial to try out our product, including the Ad Creator with AI, and see for yourself how you can make your LinkedIn Ads management a breeze. Start your free trial below!
