When it comes to AI and ads it is not anything new. There are several ways that Artificial intelligence has been used for several years in order to improve the performance of ads. While we are sure there will be many more advancements in AI and ads let’s talk about how AI is being used now. There are some basic mechanics that are happening across all ad platforms and some that are happening within specific types of ad platforms.
The ability to target using AI has been going on across many platforms for quite a while. Although the sharing of information as users crossed from one platform to another has recently been severed with Apple and their introduction of permission to track.
Each ad platform has algorithms in a place where they are able to scrap information about user behaviors on their platforms to understand what their interest and behaviors are. AI is essentially able to do audience segmentation.
The segmenting of audiences is done in several ways but one of them is using natural language processing. This is where algorithms can predict with certain predictability what the likely next set of words would be around the language you have read. Algorithms are also able to read the images and videos that you share, like, and comment on be better understand what your preferences are.
AI can also now be able to pick up on the tone and sentiment of the post that you make and be able to serve you different ads based on that type of information. This is where we have seen in our opinion perhaps being taken too far.
Facebook admitted that at one point the publicly tested serving content was angrier to see what the reaction would be of those. This happens in natural environments too where you are often a product of who you are around the most.
The twist here is it be used to sell you products and services. But even then when you break that down is it that far different than a marketer using certain words in their marketing to evoke emotions, no it’s not.
All of this is used to create targeting categories that advertisers and businesses alike can use to serve you ads that would be relevant to your interest. It is not a perfect science but for us, it’s better than being served as that is completely irrelevant to us.
Many ad platforms as using machine learning algorithms to place your ads into their many subcategories of placements of ads. When an advertiser says they are running Facebook ads they are also likely running your ads on a variety of many placements such as on reels and on the regular feed.
Platforms like Facebook can use AI to detect the best placement to place your ads based on the marketing objectives you have chosen. This allows for more efficient placement.
You can however as the marketer also choose to manually place and adjust based on the data. This sometimes can lead to inefficiencies in targeting and time used. However like with any of these AI tools in ads, you should never rely solely on the machine to make decisions. There should be a marketing strategy involved.
For instance, if we had an eCommerce client it might make more sense to isolate the budget and ensure our ads are showing on the marketplace. A placement within Facebook is more likely to have consumers in the mind of shopping.
Other platforms like Google ads have advanced far into placements by offering new types of campaigns called performance max where Google is picking placements of your ads across many of their placements available (Search, Display, Youtube, and more).
Predictive modeling in ads is used when algorithms are built to predict if someone is likely to take a specific action on an ad. For example, if you were to choose a marketing objective of traffic is Facebook ads and you target a certain audience. Facebook ads will not show your ad to everyone within the audience. They will show your ad to those who are likely to click through on the ad based on their past behavior of clicking on ads before with a greater amount than others in the same audience.
For this reason, sometimes starting off with going after say “conversion” marketing objective will limit the number of people you will reach in an audience because predictive modeling has singled them out as not ikey to take certain actions.
This is why we also see that creating funnels of people from the top of the funnel, just engaging with an ad, down to taking desired actions will help with the limitations of predictive modeling. The predictive modeling will make everyone compete for the same audiences when cheaper traffic can be found up the funnel.
You will see predictive modeling also being used to predict ads you likely would want to be served. So while targeting algorithms in AI it is mixed with predictive modeling to create the algorithm.
Predictive modeling is used in Google ads search to show your ads to potential searches based on what they likely interpret as your intention of the search. When you shoes keyword match types such as broad match predictive modeling is definitely at play on when and where your ads will show.
Automation in AI and ads is where you allow ad platforms to automate certain tasks such as removing redundant keywords, automating bid increases, and a lot more. At first, it sounds like a real time saver but there are so many ways we don’t think of before automating tasks of where they could go wrong. Take your reporting for example and automate that. you could get really subpar results.
For example, we could automate having Google create ads for us but that can very poorly its ability to create unique and interesting ads. The AI behind those ads is going to be using natural language processing. For this reason, your ads will start to look like everyone else and the creative edge you once had would be gone.
A modified version of just having them create ads for you from scratch is to use dynamic ad generation. Both Google and Facebook ads AI a little differently with their dynamic ad generation. You can input a variety of headlines and creatives and it will use various different variations to achieve the overall objective.
The podcast and streaming radio ads will use AI in a fascinating way where they take into account who is listening and can actually modify the said gender of the person reading the script to match the listener.
They have been many advancements in both audio and video to be able to change in a matter of seconds based on the demographics of the listeners. This gets into an area of not knowing what is actually real and not real.
AI in performance measurement is able to use algorithms to be able to display results in ways that allow marketing and advertisers to be able to see what is working and what is not.
Sometimes these numbers though are not exact sciences and this is where AI is coming in. Algorithms are built to determine if clicks and other metrics likely happened. We would all like to believe that they are 100% true but they are often not due to a lack of technology to actually make it happen.
Recently Apple removed apps from being able to track once someone leaves an app and made it so that more algorithms with AI needed to be built to deliver the performance measurement that was there before.
This is why we always believe in using triangulated data from multiple platforms in order to keep them all honest. After all, it would be in an ad platform’s best interest to make its numbers look better than they actually are.
Artificial intelligence is also being used in fraud protection for performance measurement. Many platforms have this built into their technologies that if they see certain behavior such as multiple clicks from the same IP in a rapid period of time it will determine it as fraud. You will likely never even know that it happened. Inside Google ads, they actually show where they did not charge you for this activity.
Smart bidding and AI is allowing the ad platforms to bid for you in order to achieve the marketing objectives that you have chosen. For most ad platforms what is happening behind the scenes is an online auction. So the cost for ads changes at different times of the day.
These ad platforms have so much data that they can place bids for you based on the likelihood that someone is to take certain actions and get you the most bang for your buck.
So if you are looking to drive more traffic smart bidding will take that into account and bid higher in those scenarios where it has determined using AI that you are more likely to do so.
It’s important to keep a watchful eye on this and make sure you fully understand what you would like to achieve when setting up a campaign to use smart bidding.
Smart bidding has really advanced Google ads in ways where you might have previously been afraid to use broad keyword matches because you weren’t sure what you would get. Now that google ads can bid effectively with smart bidding it can choose where to bid on and where not to get you the conversion objectives you have chosen.
As mentioned above how your ads are placed is using AI to determine the ad ranking of your ads and who gets placed where. This is based on rules that ad platforms have designed of what they feel is best to service their end clients.
For example, Google ads place value on your keyword choice, your ad content, and your landing page all being similar in nature. When they are similar Google assigns an ad rank higher and you end up paying less per click.
Facebook ads do something similar in that they place precedents on engagement with ads as they should since it’s a social network. The more engagement your ads have the better rank you have and the more reach and less you pay per click.
What Will We Have Next With AI and Advertising?
What is fascinating about all of this is many of the citations where are describing above are happening in a matter of seconds. Technology countries to evolve to allow faster processing in computation and machine power.
All of though is done using a machine and the machine can take the information it has been given and sort it which is the basis of AI.
AI has not evolved to a state where machines can think for themselves. This is where marketers need to be vigilant in checking and ensuring that they are using Ai to help them achieve results but with strategy in mind. AI maybe be able to make things faster but they are not evolved enough to know the strategy of who, when, where, and why.