The ChatGPT Phenomenon: Should You Trust AI for Marketing?
March 1, 2023
Nov. 30, 2022, may be remembered in history as the beginning of a new technological era that revolutionized how people went about their everyday lives. Or it could go down as the first step in AI’s overthrow of its human overlords.
Or it could turn out to be an insignificant blip on the tech radar the way Google Glass, MiniDisc music players, Quibi, and, increasingly, NFTs are.
On Nov. 30, 2022, OpenAI released ChatGPT to the public. The AI-driven large language model chatbot was unlike anything most people had seen. ChatGPT didn’t just give you answers to your questions—it could create long, custom responses based on just a few prompts from the user. The tool can generate poetry, create song lyrics, build computer code, write sermons, give practical (if not always wise) advice, explain complex topics, and do students’ homework.
Also, ChatGPT can be used to generate marketing content and strategy, which has marketers excited at the possibilities or fearing the apocalypse—and everywhere in between.
ChatGPT developments and news haven’t slowed since its public debut. Although some of ChatGPT’s luster as a marketing resource has worn off, the HubSpot App Marketplace now includes a third-party ChatGPT integration that allows users to access the tool from HubSpot.
So should businesses use ChatGPT to assist with their marketing efforts? The answer is complex, but the short answer is that the tool isn’t the game changer that some people have made it out to be. And even ChatGPT, when I asked it, doesn’t think it’s a great idea:
ChatGPT Content Reality Check
As the ChatGPT conversation has unfolded over the last several weeks, I’ve been intrigued by the varying opinions on whether the content the tool creates shows quality. Some business and tech journalists have been gushing about the readable content ChatGPT is able to produce. Other professional writers have been less than impressed, noting that ChatGPT’s copy sounds impersonal, simplistic, unimpactful, and boring.
The text I’ve had ChatGPT generate reminds me of the five-paragraph essay many of us were taught to write back in middle school. That exercise helped students collect what they wanted to say into a tidy format: Introduce the topic and a thesis in the first paragraph, offer three supporting points, and wrap it up with a conclusion. This approach is great if you’re a seventh-grader learning how to write competently ... but not so much if you’re trying to produce marketing content that inspires, entertains, and engages readers.
Quality marketing copy still matters—it resonates with your audience, ranks highly on Google, reduces bounce, and drives click-throughs. ChatGPT struggles to produce such content. The critics are right: The AI copy is dull and takes a long time to get to the point. It lacks transitions between sentences and paragraphs, and the narrative is barely compelling.
Moreover, ChatGPT dumbs down copy without making it feel more accessible. I’m not a huge fan of the Flesch Reading Ease score—I believe a piece of content’s readability is dependent on too many factors to simplify it into a rating—but I compared a blog article SmartBug wrote for a client with ChatGPT copy that the tool was directed to rewrite, based on the original blog, to a certain length. Besides not generating the requested word count, the AI rewrite scored significantly lower than our professionally written copy. In other words, the copy was not only less appealing but also more difficult to read.
Accuracy Not Guaranteed
As impressive as ChatGPT is on the surface, its most serious flaw might be that it gets things wrong. Even OpenAI states that the chatbot may not always be accurate. CNET found this out the hard way when it used ChatGPT to generate articles—which went through human editors—only to discover several of the articles contained fact errors. That’s not a good look, particularly with content that was giving financial advice ...
ChatGPT isn’t a depository of knowledge that its creators loaded into the application. The tool draws its responses from what it scrapes across the internet, as vast as that may be. So, when you ask ChatGPT a question, it formulates an answer from sources it chooses from the web—and those sources might be less than trustworthy.
Moreover, ChatGPT must interpret the sources it scrapes into a response and, sometimes, it doesn’t quite know how to do so. Some users have enjoyed trying to fool ChatGPT into giving inaccurate or confusing responses—but posting confusing AI-generated content, as CNET discovered, is no joke.
I asked ChatGPT what the worst Atari 2600 video game cartridge is. I’m a big classic video game fan, and I do have an opinion on this, so I was curious what AI thought. Here is the reply I received:
This isn’t necessarily a bad answer: Many people do consider E.T. the worst game (it’s not, but that’s for another blog post on another blog), and ChatGPT likely scraped a couple articles that espoused that opinion. And indeed, a bevy of unopened cartridges did end up in a landfill in New Mexico. But that’s where ChatGPT goes awry. The reputation of the game didn’t cause it to be dumped in the landfill; Atari had too much stock as the industry was crashing and dumped cartridges of other titles as well. Furthermore, the game wasn’t “sold in a landfill”—that’s an example of ChatGPT not being able to articulate a fact and, instead, producing something inaccurate.
Video game history aside, your brand’s reputation is only as strong as the accuracy of the information that brand delivers. You wouldn’t tolerate a human repeatedly getting facts wrong in your marketing content, so you shouldn’t settle for that poor standard from AI.
AI Leading to More Work?
The promise of AI—and what has many people excited about ChatGPT—is its ability to save organizations time. Why spend an hour writing a piece of content that ChatGPT can spit out in a minute?
This kind of time saving is overrated, particularly if you are putting in hours after the fact trying to clean up AI-generated content that is bland, off brand, inaccurate, and/or generally substandard. Consider these potential roadblocks with ChatGPT content:
- As already stated, accuracy is a big issue. Unless you are confident about the facts ChatGPT is generating, you’ll need to closely edit.
- Not knowing where ChatGPT is getting its information is also problematic. Are its sources varied and trustworthy? If it is generating thought leadership outside your organization’s expertise, you should be crediting and linking to other sources, which ChatGPT might not adequately provide, even if you ask. And though ChatGPT is good about not copying text verbatim, rewriting from just one source is still plagiarism. The bottom line: You may need to spend time researching ChatGPT’s “research.”
- You may need to edit ChatGPT content for keywords, brand consistency, tone and voice, target persona, and length.
- So many people have been using ChatGPT that OpenAI’s servers can’t keep up. That can translate to delays in generating content.
- Providing more detailed instructions might help ChatGPT write content more specific to your needs ... or it might not. Either way, you’ll need extra time to craft those instructions.
Responsibly using ChatGPT means plenty of due diligence—and time and resources—just to ensure copy meets your standards. You likely will be better off writing your own content (or trusting your digital marketing agency ... wink, wink) with the confidence you’ll be exceeding those standards from the start.
In Search Of ...
Last year, Google declared that AI-generated content was against its usage standards, setting up the possibility that webpages with such content might rank worse.
Google has since clarified that AI content isn’t against its guidelines and will rank as long as content is “useful, helpful, original, and satisfies” its E-E-A-T—short for experience, experience, authoritativeness, and trustworthiness—standards. However, Google emphatically states that automated content created to manipulate search rankings is a violation of its spam policies.
Google has countered past attempts to circumvent its algorithms; keyword stuffing, backlink stuffing, duplicate content, and poor-quality content are some of the strategies that the search engine giant has smacked down over the years. Tools are already available (including one from OpenAI itself) to determine if a piece of content is AI-generated, and Google will likely get good at determining if a webpage wasn’t written by a human.
Whether AI-generated content is ranking on searches is up for debate, but Google definitely seems on the defensive. It has much to lose if people start bypassing search engine results for detailed (if not always accurate) answers to queries. To counter the ChatGPT’s sudden emergence, Google is rushing its own AI tool, which—maybe not so unpredictably—produced a notable factual error during its first demo. (Microsoft’s Bing AI also made mistakes in its recent demo and is producing some truly dark responses to users’ queries.)
Content marketing depends heavily on people finding your articles and guides when they search for answers to their questions. How well AI content ranks—and will rank in the future as Google evolves its algorithms—remains a bit of an unknown. For example, would a ChatGPT-generated blog article that’s heavily rewritten still be flagged as AI content and possibly spammy by a search engine? Or, if ChatGPT significantly dents Google’s search dominance, would Google get more stringent with how it ranks AI content?
The Merits of ChatGPT
If I’m coming across as skeptical about ChatGPT, it’s perhaps because I’ve explored and vetted other AI marketing tools and can see how much ChatGPT’s content capabilities are being overhyped. As a long-time marketing writer, I see ChatGPT as the shiny new toy that loses its appeal a week after Christmas because, well, your old toys were much more reliably fun.
That said, ChatGPT is undeniably impressive. Marketing aside, its potential to help people—from coding to research to education to empowering people with disabilities—has barely been realized yet. And, yes, it’s kind of fun to use.
And ChatGPT does have marketing uses. Although it’s not especially reliable or creative with long-form content, it can save time and not totally induce groaning with shorter content. For example, you can plug in a 1,000-word blog article you wrote into ChatGPT and ask it to create a series of social media posts across platforms. Even though the ChatGPT response I shared earlier didn’t recommend this, I tried it, and the results were solid (although it took a few minutes for the AI to come up with 12 posts). Any posts that might not be so great can be easily edited. Trying to clean up a long, AI-generated blog may require a significant time commitment; doing the same for a one-sentence tweet takes only a minute, and if you’re not satisfied, you can just write a new one yourself, no sweat.
Similarly, you can use ChatGPT to write meta descriptions, article titles and subheadings, quick summaries, and any other short text that feels more tedious than creative. However, dedicated AI tools exist that also do this—and possibly do it better.
Beyond smaller bits of text, ChatGPT also can help create outlines to get you started with writing a longer piece of content. It will give you sources to link to and even a title if you ask. Again, there are dedicated applications that do this better, but for that occasional time you just can’t wrap your brain around a topic, ChatGPT provides a way to break through the mental block.
Furthermore, ChatGPT can provide a nice starting point to research a topic you aren’t familiar with. For example, a marine business in Wisconsin may want to write an article on boating in Minnesota. I used this example and plugged in “What maritime laws should I know if I'm taking my boat to a lake in Minnesota?” and got back a concise list of suggestions. I wouldn’t use the results directly as a blog article (and when I asked where ChatGPT found this information, it showed a few non-linked references but also said to “double-check with official sources”), but they do provide a great launching point (no boating pun intended) for additional, focused research on what Wisconsinites should know when they head west.
Helpful, but Not Omnipotent
Despite its limitations, ChatGPT still should be in every marketer’s and copywriter’s toolbox. It shouldn’t replace the work you do, but it can create more efficiency and allow you to devote more time toward creating outstanding work. Other dedicated marketing AI tools may achieve this efficiency even better, potentially freeing you from the boring tasks that get in the way of the skills only you can bring to your organization.
There are three things you should remember when using ChatGPT or any AI tool. First, ChatGPT isn’t all-knowing and it doesn’t have your skills. If I may coin a term, don’t fall for Google Maps Syndrome, in which people think they must make a right turn when their GPS tells them so, even though they know it’s not the best route. AI tools don’t understand context and nuances the way an expert marketer, editor, or copywriter does. If ChatGPT shows an outline you aren’t sure about or is just plain wrong, ignore it and trust your instincts and expertise instead.
Second, ChatGPT’s appeal is that its basic version is free, but it can be slow, has no tech support, and isn’t built to solely handle social posts or any other marketing-specific content. It’s currently the big-box store of AI when a little boutique might fit your marketing needs better.
Finally, ChatGPT, despite its ability to write poetry, song lyrics, and flowing prose in any style, pales in comparison to the creativity of a human being. No matter what your feelings are about AI tools, there are some things they will never replace. And if your competitors choose to overuse ChatGPT copy, your human-driven content will stand out even more—especially with your prospects, leads, and customers.
About the author
Joe Gillespie is Director of Inbound Copy for SmartBug Media. He graduated from Marquette University with a B.A. in journalism and, before coming to SmartBug, was a two-decade veteran of the newspaper industry. Read more articles by Joe Gillespie.