Anywhere you turn today, you’re bombarded with promises about what generative AI is going to do to your job. Depending on what you read and believe, millions of professions are either going to become a total breeze or be eliminated outright.
The emphasis being placed on tools like ChatGPT inevitably creates pressure for digital marketing agencies to find use of them. In part because, we’re told, we’ll be left behind if we don’t. I’ve even seen some chatter in Reddit marketing circles from marketers who claim they don’t even use Excel at all anymore—they do all their data analysis through AI tools.
So it’s worth discussing how some marketing agencies are using generative AI and what that may mean for their clients. In this blog, we’ll take a look at potential applications while sharing some of our own feelings about where these tools can shine in digital marketing—and where you’re probably better off avoiding them.
The Many Potential Uses of Generative AI in Marketing Agencies
In the digital marketing community, workers have identified countless applications for these technologies—and are testing them with varying degrees of success. Some of those being discussed include:
- Identifying larger industry-level trends (ex: seasonality in search volume)
- Generating ideas for ad copy tests
- Generating ideas for landing page tests
- Keyword research and generating negative keyword lists
- Researching audience/customer profiles based on specific attributes provided by clients; designing buyer personas where applicable
- Creating talk tracks (example: how to discuss specific topics or explain certain performance trends to their clients)
- Client competitor analysis (example: ads, keywords, and audiences that the competition may be using)
- Budget recommendations
- Completing simple design tasks
- Ideating direct response tactics for email or SMS (ex: stronger email subject lines or SMS messages designed to drive more clicks)
- Quickly organizing and analyzing complex data sets
As a stubborn, increasingly grey-haired veteran of this ol’ digital marketing game, I’m weary of becoming reliant on these tools the way many marketers are. Some of these applications are perfectly logical time-savers—like surface-level research tasks—as long as the outputs are treated with the proper veracity and verified by the user.
Generating ad ideas might be one of AI’s strongest applications at the moment, but it’s important to differentiate between ideating and trying to make your final-product ad creative with it. As these tools are increasingly being integrated into the ad platforms, even those platforms are advising advertisers to primarily approach them as a brainstorming aid—knowing that the results are often too inaccurate or bizarre to fuel successful campaigns.
Data analysis is probably the one application that gives me the most pause. If a user is just plugging in datasets and taking AI analysis at face value, it’s likely they’re not actually digging in and really understanding performance. This means they’re going to have a harder time with generating strategy, or at least with recognizing and correcting weak strategic recommendations that might be made by a GPT using incomplete or ineffective data analysis, or a GPT that doesn’t have a good sense of “future proofing” or “looking ahead”.
Client Feelings About AI-Generated Content
The rise of LLMs and generative AI has driven many brands to investigate using these tools for their ad copy and creative assets. Technically, AI can own this entire process—provided a marketer is supplying adequate, informative prompts up front.
In our conversations with clients, many have expressed interest, but they have also remained cautious. There’s a lot of grounded trepidation around the risk of generative AI outputs that fail to represent the brand accurately. For example, one of our medical clients is strongly against using generative AI for visual assets because the images don’t look like the inside of its clinics. They’d prefer a true-to-life representation of their locations rather than anything an AI tool can create.
Advertisers seem to be a little more open-minded when it comes to ad copy, but the caution remains. Getting generative AI ad copy to a level that meets brand standards requires inputting some really specific info about the company. That would mean doing things like inputting the client’s own internal core tenets or having a tool review their Google Business Profile pages to find common themes from 5-star reviews and featuring those items in the ad copy. All that can turn into an iterative process nearly as exhaustive as just writing it ourselves.
The Risks of Over-Reliance on Generative AI
The danger is relying too heavily on this tech. Over-reliance on AI can lead to some really underwhelming outputs. Typically this takes the form of weak strategies and hyper-generic assets that fail to effectively communicate value propositions or competitive advantages.
On the content side, it can mean that strategies aren’t actually aligned with client goals, and it can also lead to a failure for the brand to establish its own voice or speak with its existing voice in the marketplace.
On the strategy side, the thing to keep in mind is that these tools aren’t going to have much of a sense of how to create forward-looking strategies. They’re a lot more reactive: they only know what’s been done before. For example, if a client is seeing poor results at driving awareness from their Video Reach campaign in Google Ads, a generative AI tool may spit out strategies for optimizing this under-performing campaign. However, a better result might be achieved by moving away from Video Reach entirely and using Demand Gen—a relatively new, more action-driven campaign type that’s still a part of the awareness stage in the funnel.
Because Demand Gen is newer and not yet as widely-utilized, a generative AI tool may not consider this a viable option because it doesn’t have enough crowd-sourced data to understand the potential of this tactic. Again, a smart marketer might account for this in their prompting, but even with good prompts it’s hard for an AI tool to create solid recommendations for things that don’t have proven use-cases. Even-newer tactics, like AI Max functionality for Search campaigns, have even less content to draw from—so an LLM is likely to underestimate its potential in ways a knowledgeable human marketer wouldn’t.
What Matters Most? Doing Right By Clients
At the end of the day, how marketers approach AI comes down to understanding their clients and those clients’ goals. If a strategy seems like it’s failing to truly align with those goals, or if it’s failing to capture brand attributes, then it’s not something an advertiser should use. Now, an advertiser may modify the prompts and provide additional info that helps to get the AI more closely aligned—but at that point, there’s a deeper requirement for the advertiser to insert their own understanding into the process (ie: relying less on the GPT).
As generative AI technology becomes mainstream, it’s important to ask your digital marketing agency whether they are using it and how they are using it so you can determine your level of comfort. Certainly, there’s nothing wrong with ad agencies becoming more efficient—but at the moment, you may not want to leave your entire ad budget and brand reputation in the hands of machines.
