top of page
SubscribePopUp

Can AI-generated content work for eCommerce?

Author: Adriana Stein

An image of author Adriana Stein accompanied by search-related iconography, including a calendar, a line chart, and a globe icon

When OpenAI launched ChatGTP in November 2022, it unleashed an AI tidal wave.


It’s been all over the news—for better or worse. It’s flooded my LinkedIn feed to the point where I’ve honestly had to take a break. Everywhere you turn, AI seems to be gaining popularity.


This makes it inevitable that AI will eventually make an even bigger splash in eCommerce marketing tools, content creation, and especially SEO. So, regardless of whether you’re for or against AI, it’s here to stay because of one crucial element:


Perceived efficiency.


Why?


Many people perceive AI as improving the efficiency of tasks that humans currently perform manually (whether that’s actually true still remains to be seen).


Ironically, AI isn’t all that new. Ever wondered how Facebook seems to read your thoughts? Or maybe you use a voice powered assistant like Alexa at home? These are powered by AI and we’ve long used them to make certain tasks and processes more efficient.


Because AI involves processing how humans communicate in order to create more natural speech patterns and AI-human interactions, the wide array of use cases is hard to ignore.


All of these factors have tremendous implications for eCommerce businesses and their marketing and SEO strategies. This leaves us with some important questions:


  • Just because we can use AI, does it mean we should?

  • Is AI as efficient as some people claim?

  • Can we truly use AI to complement our eCommerce marketing efforts or will we begin to rely on it to the point where there’s no human creativity and everything sounds the same?


Here, I’ll explore both sides of the argument and consider what that might mean for the future of eCommerce marketing and SEO.


Table of contents:


The past is prologue: How AI has already changed eCommerce


One of the best frame of references to think about how AI can be used in eCommerce in the future is to examine the ways in which it’s already being used. Some of the most popular uses of AI in eCommerce today include:


  • Fraud prevention: AI can analyze transactions and uncover suspicious behavior. This is extremely difficult for people to do in real time, especially as businesses grow in size.

  • Product recommendations: AI can start to recognize patterns in shopping behavior based on demographics, which could help with more targeted product recommendations over time. It’s also able to recognize IP addresses in order to create a more personalized experience based on previous purchases and browsing history.

  • Dynamic pricing: AI can analyze market trends and look at competitors to give you a better understanding of the current price point for your products and services.

  • Virtual shopping assistants: AI-powered chatbots have been around for a while but they’re getting increasingly more intelligent/responsive. Teaching AI how to interact with customers can create a stronger user experience. This allows for 24/7 responses to consumers around the world.


With all of these existing applications, are SEOs’ new found fear of AI truly unprecedented? Jumping into the unknown always seems a little scary, but AI has already proven to be pretty helpful.


So, how do we as eCommerce marketers and SEO strategists move forward? From my perspective, it’s all about how AI is used.


AI for eCommerce: Competition or collaboration?


Widespread use of AI is just beginning to break into the world of eCommerce content creation and SEO. Although there are already several different AI tools available that can help with content creation, they haven’t been discussed to anywhere near the level of ChatGTP.


ChatGTP is said to have revolutionized AI technology. It’s being used for everything from business plans and press releases, to content creation and eCommerce marketing strategies.


So, how will this change the playing field moving forward?


In the past, content has been primarily written entirely by human content writers and marketing specialists. Now, with the introduction of various AI platforms gaining popularity in 2023, people are wondering whether AI-generated content is going to take over and replace content writers altogether.


Accuracy issues are still common in AI-generated content

However, there are some major concerns about the factual accuracy of some of these tools. Since AI generates information that’s taken from a wide variety of sources, it’s difficult to confirm their reliability. The sources they pull information from may not have been fact-checked, causing the AI to put out content that is questionable to say the least.


Twitter post from Ivan Dario Valencia showing an example of a ChatGTP output that talks about a pandemic similar to cholera in France that is false information.

In fact, ChatGPT even warns users about the risk of misinformation when you first create an account:


A screenshot of the warning users see upon signing up for ChatGTP regarding the potential risk of outputs containing incorrect or misleading information.

With that being said, AI doesn’t seem like it’s anywhere near ready to take over the role of human writers. We still need to make sure that things are accurate, reliable, and unique in order for content to carry any weight online.


The technical basics: Differentiating AI vs. Human writing


To ensure that eCommerce content and SEO strategies are held at a high standard—and to reduce the chance of penalties/issues based on AI-driven SERP manipulation—it’s helpful to understand how search engines perceive and detect AI.


Google and other major search engines want to make sure that SERPs are filled with valuable websites, not spammy content. Google has long striven to incorporate ways to remove low-quality spam content and AI is no exception.

So, how would Google do that?


Natural language processing (NLP) was created to understand and mimic human writing, but it can lack human-level originality. This is problematic for the SEO world because without differentiation, AI-generated content would quickly fill up the SERPs with repetitive, generic pages that might not even help users find what they’re looking for.


Algorithmically detecting AI content

Several tools have been created to check whether content was created by AI. These tools are based on an understanding of how language modeling works and one of the most effective ones is the Giant Language model Test Room (GLTR), created by MIT-IBM Watson AI lab and Harvard NLP.


GLTR analyzed the machine-generated demo text below based on word variety and placement. Green highlights indicate the most frequently used terms, while purple indicates less frequently used terms. GLTR defines this text as machine-generated because the words lack variety and are too predictable.


An example of a human-written demo text analysis with the Giant Language model Test Room (GLTR), created by MIT-IBM Watson AI lab and Harvard NLP

You can see another demo of a human-written excerpt here. Notice that the text contains more purple words, indicating that there is a higher level of word variety.


An example of a human-written demo text analysis with the Giant Language model Test Room (GLTR), created by MIT-IBM Watson AI lab and Harvard NLP

Comparing these two analyses, the differences are slight and nearly indiscernible for the human reader.


Although GLTR has access to the GTP-2 language model and ChatGTP has now progressed to GTP-4, the foundations for AI detection are likely to remain similar.


Google doesn’t care whether you use human or AI writers

Currently, Google is not against AI-generated content, nor does it technically violate guidelines. As long as AI content isn’t used for spam or to manipulate rankings and provides quality, helpful information, Google seems largely on board with it.


A screenshot of Google’s guidelines on AI-generated content, stating, “As explained, however content is produced, those seeking success in Google Search should be looking to produce original, high-quality, people-first content demonstrating qualities E-E-A-T. Creators can learn more about the concept of E-E-A-T on our Creating helpful, reliable, people-first content help page. In addition, we've updated that page with some guidance about thinking in terms of Who, How, and Why in relation to how content is produced. Evaluating your content in this way, whether you're using AI-generated content or not, will help you stay on course with what our systems seek to reward.”
Google’s guidance emphasizes originality, quality, and E-E-A-T, regardless of whether humans or AI generated the content.

Despite the rosy approach from Google, this comes with major implications for eCommerce SEO and content generation:


Relying solely on AI-generated outputs is highly likely to get redundant. And the more businesses do that over time, the harder it will be for audiences to differentiate brands—which is incredibly harmful when it comes to providing a good customer experience.


The bottom line is that it’s incredibly important to manually review all AI-generated content and improve its uniqueness.


How to create unique, human-centric eCommerce content using AI


I strongly believe AI will become a typical part of the eCommerce marketing tech stack, whether we want it to be or not.


So, the most productive way eCommerce marketers can respond to these changes is to learn to harness its power for your advantage.


Here are my top suggestions for doing that:


01. Be as specific as possible with your prompt: Providing generic questions or requests will lead to generic outputs. Instead, include within your prompt:


  • The marketing channel the content will be written for

  • Word count

  • Target audience

  • Messaging

  • Intent


You should also consider building an outline first (which could also be done with AI), then entering the outline in subsequent prompts to ensure the generated content is consistent. This is one of the areas where input from human strategists is still heavily required.

02. Adapt AI outputs for the specific channel, best practices, purchase intent level, and general readability: A lot of the content that’s spun from AI is redundant and reads at a very low level. This is another area where input from human strategists needs to be prioritized. AI doesn’t differentiate well between an eCommerce product description, blog article, and landing page. That’s where you (as the strategist) need to ensure content is structured for success.


03. Don’t copy and paste exactly what the AI generates: Doing so will defeat the whole purpose of your content creation, as you’ll start sounding exactly like everyone else. Not to mention the content could include misinformation and incorrect product data.


04. Meticulously fact-check everything the AI generates: While this might seem like an obvious step, there are some sites that have used AI to write articles without any fact checking involved. They then published said articles as is, without seemingly reading them first, leading to PR nightmares like The Washington Post’s article about CNET entitled “A news site used AI to write articles. It was a journalistic disaster.” If you want customers to buy from you, they need to trust your reliability, so don’t be the brand spreading misinformation—take the time to thoroughly fact check.


05. Add the human element through real-life experience and real-world examples: Center your content around a technical subject matter expert, real customers, quotes, relevant product information based on how someone actually uses it, and factual sources that can help your target audience better understand how a product or service can address their pain points. ToFu-style generic content has come to an end—that, AI can certainly already produce. eCommerce marketers and SEOs now need to up their game with content personalization.



 A Twitter post from Mordy Oberstein, head of SEO branding at Wix describing his predictions about content quality changes when using AI for content production. The post says, “AS AI writers water down content quality, brand marketing will come into focus as brands that become known for having unique content will stick out like a sore thumb (relatively) and become go-to sources for topically related content.”

Examples of eCommerce content generated by ChatGTP


To give you a better understanding of how you can utilize eCommerce content generated by AI, I’ll use some examples based on the company Glowforge. I’ve chosen a rather technical eCommerce product with a unique target audience to test ChatGTP in a real and complex scenario.


A screenshot of Glowforge’s homepage, showing its 3d laser printer with text that reads “Glowforge’s powerful desktop laser cuts, engraves & scores hundreds of materials”

Glowforge is a company that sells 3D laser printers that allow users to create various items at the push of a button. It can score, engrave, and cut different materials, allowing for a variety of potential applications. Notably, its target audience is stay-at-home moms in the US who are looking for creative ways to grow their own business and work from home.


Product description

AI contains the fundamentals necessary to generate eCommerce product descriptions, though the outputs are rather basic and need to be edited by an SEO Strategist. In order to use AI for writing product descriptions, you just need to enter a prompt that directs the AI on what to do. It seems like some descriptions might be better curated with the use of an outline prompt, but in the example below, I used the following prompt:


Write a 300 word production description for the keyword “3D Laser Printer” to target moms who want to build at home businesses by making custom items with a 3D Laser Printer


An example of an eCommerce product description generated by ChatGTP that targets moms interested in buying a 3D laser printer so they can grow their own creative business by making custom items

An experienced eCommerce marketer would see that this is a terrible product description because it lacks the majority of elements eCommerce SEO product descriptions need, including keyword optimization in headings, ease of readability through the proper content structure, and generally vague messaging in relation to the brand. So, herein lies the importance of my above advice:



The big lesson here: Effectively using AI for eCommerce SEO is not a simple copy/paste job. You still need to know how SEO functions and how to use prompts to generate useful outputs for specific content types.


SEO-focused blog article

You can also enter a prompt for AI to generate an entire SEO-focused blog or article. In the example below, I used the following as a prompt (with a bit more detail than the product description example above).


Write a 1000 word blog article targeting the keyword “how to use laser engraving machine” with “how to” information to target moms who want to build at home businesses by making custom items with a 3D Laser Printer. The post should focus on the story of other similar customers who used the product and had success gaining financial independence and building a business around their passion.


An example of an SEO-focused blog article generated by ChatGTP that describes “how to use a laser engraving machine”  for moms who want to build at-home businesses by making custom items
Click to expand.

This blog article output is better because I told it to focus on “how-to” content (the intent here is informational because the reader is trying to learn and not yet ready to buy).


Nevertheless, it could be better with some help from a human SEO strategist:


  • Create an outline prompt first, then generate a full draft based on that for a stronger output

  • The big text blocks are way too formulaic to be human-written, are difficult to read, and need to be broken up (these are clear indicators of AI-generated content)

  • Typical content optimization best practices need to be applied

  • There are no real customer stories, so it massively lacks personalization

  • Internal links to conversion landing pages and external links for citations need to be added


Well, AI still has a long way to go before it replaces human SEOs!


Organic Instagram post

I also wanted to test AI’s capability for non-SEO eCommerce content. Here, I entered the following prompt into ChatGPT:


Write an organic instagram post with “how to” information that targets moms who want to build at-home businesses by making custom items with a 3D Laser Printer. The post should focus on the story of Lisa, who recently did the same.


An example of an organic instagram post generated with ChatGTP that includes “how to” information for moms who want to build at-home businesses by making custom items with a 3D Laser Printer

This is definitely the strongest output so far; however, it still needs a human review to:


  • Add the actual steps Lisa took, including what product she bought, what she created with it, and how she sold it. It’s completely lacking concrete detail.

  • Edit the formatting for Instagram.


More good news for eCommerce content marketers, AI can get close but no cigar!


Facebook ad

Let’s now test AI with a Facebook ad. I used the following prompt:


Write a Facebook ad creating curiosity around moms and how they can stop their unfulfilling full-time job and build the at-home business of their dreams by making custom items with a 3D Laser Printer. The ad should focus on the story of Lisa, who recently did the same.


An example of a Facebook ad generated by ChatGTP intended to create curiosity among moms and how they can stop their full-time job and build the at-home business of their dreams by making custom items with a 3D Laser Printer

Although this was actually the most passable content AI generated so far, it can still be better. I would improve this by:


  • Telling the story directly using Lisa’s words rather than just including one quote. Again, there just needs to be more concrete detail about Lisa’s story.

  • Adjust format slightly for readability and Facebook standards so there's only one sentence per line.


What I can tell at this point is that AI is better at short-form content rather than long-form content. So depending on how well AI outputs work for your products, there may be other ways to integrate AI than with a full SEO blog article.


Landing page

Let’s now assess how AI does with a conversion landing page. This was my prompt:


Write a 500-word conversion landing page selling the “3D Laser Printer” to target moms who want to build at-home businesses by making custom items with a 3D Laser Printer. The post should focus on the story of Lisa, who recently did the same.


An example of a landing page generated by ChatGTP intended to encourage moms to stop their full-time job and build the at-home business of their dreams by purchasing a 3D Laser Printer and making their own custom items.

As you can see, this one fell entirely short. However, I think that’s likely because the prompt was a little vague. And it’s clear that ChatGTP doesn’t understand what a conversion landing page is. This needs a heavy hand from a marketing strategist to improve elements like:


  • Using conversion copy headlines

  • Generally adjusting all of the content structure for conversion landing page best practices

  • Placing relevant CTAs throughout the text


This is a content type that needs some true testing to determine whether it’s actually efficient to use AI (as opposed to just building your own page outline and working with a knowledgeable conversion copywriter). From this low-quality example, I’d suggest sticking with a copywriter.


Nurturing email

Lastly, let’s give an eCommerce product nurturing email a try. I used the following prompt:


Write a 200-word nurturing email selling the “3D Laser Printer” to target moms who want to build at-home businesses by making custom items with a 3D Laser Printer.


An example of a nurturing email generated by ChatGTP intended to encourage moms to stop their full-time job and build the at-home business of their dreams by purchasing a 3D Laser Printer and making their own custom items.

The “heart” of good nurturing conversion copy is certainly absent here. An experienced eCommerce content strategist still needs to improve on this copy by:


  • Removing the P.S. portion (or any other repetitive parts), as they seem out of place and too salesly

  • Adjusting the formatting for email by breaking up text blocks, adding CTAs, and links.

  • Adding real customer stories and personalization


The trend I’ve seen across all these outputs is the lack of structure and personalization. AI is nowhere near ready to generate any kind of content at ready-to-publish levels.

Even with AI, you still have your work cut out for you.


eCommerce SEOs can use AI to enhance—not replace—their work


The best way to use AI is to enhance, rather than replace. Think of AI-generated content as a rough, rough draft. Instead of copy/paste, make strategic enhancements to structure and integrate real-world, user-first personalization. These elements are absolutely essential if you want to prevent your content from getting lost in the heap.



A LinkedIn post from Allie K. Miller, AI Entrepreneur, Advisor, and Investor describing her viewpoint about whether or not AI will replace copywriters.

Whether we use it or not, AI has already grown deep roots in eCommerce marketing. It seems to offer efficiency (or at least perceived efficiency from a business perspective), so we’re probably going to see a lot more of it in the future.


This means that as an eCommerce SEO, it’s time to jump on the bandwagon and adapt, or be replaced by someone who has.


 

adriana stein

Originally from the US and now living in Germany, Adriana Stein is the CEO and founder of the marketing agency AS Marketing. She leads a team of multi-language SEO experts who develop holistic international marketing strategies for global companies.


Comentários


Get the Searchlight newsletter to your inbox

* By submitting this form, you agree to the Wix Terms of Use

and acknowledge that Wix will treat your data in accordance

with Wix's Privacy Policy

Thank you for subscribing

bottom of page