I taught an ai robot to write blog posts like me. Here is what I learned.

I

As a copywriter and content strategist, I have two choices. I can either be quick to dismiss the new tools that claim they can replace writers or, I dive in headfirst to learn them. Such is the case with AI writing tools.

However, tweets, like this one, only add to my skepticism.

AI Writing tools use freelance writers

But, I put aside my skepticism and embraced learning the ins and outs of a few of these AI tools.

To date, I’ve tried a variety of AI writing tools, such as Jarvis.ai (now Jasper) and Copy.ai, to test if they could be a serious replacement for writers in general.

In this experiment, I stacked an article from one of my best writers that I collaborate with against an AI-generated Jasper post based on the same content brief.

Here’s the truth about whether or not AI tools will replace writers. 

What is Jasper?

Jasper is a copywriting tool that leverages machine learning and artificial intelligence (AI) to produce blog posts, sales pages, social media copy, and more via pattern matching.

According to their website, its AI software can generate up to five times more content in the same or less amount of time than it would take an actual writer.

Jasper uses a combination of Templates and Recipes to help users create content. Templates are building blocks for copy that consist of long-form frameworks and shortcuts for short-form posts. Recipes are pre-built workflows that utilize a series of commands to create content in a repeatable process. 

copywriting frameworks in Jarvis

Some of their most unique features tie into Templates and Recipes. For instance, it offers robust AIDA (Attention-Interest-Desire-Action) and PAS (Problem-Agitate-Solution) frameworks that many seasoned copywriters, like myself, are deeply familiar with. 

Other unique features include Explain It To a Child, which makes copy easier to understand, and Blog Post Intro, which can help marketers avoid writer’s block at the start of a piece.

explain like a kid template

Jasper can also be helpful for more short-form copy, like email subject lines, social media copy, and even company bios, which can be time-consuming or tricky to produce. 

What are AI writing tools? 

As the name might suggest, AI writing tools use artificial intelligence to speed up or enhance the writing process — to the point that some argue these tools can replace real writers. All users have to do is input relevant information, like a target keyword or content topic, and an AI writing tool will scan the web to pull related content and flip the copy to make it unique. 

AI writing tools often go hand-in-hand with other AI tools for SEO optimization, such as Clearscope, Frase, and SurferSEO.

In fact, Jasper has direct integration to Surfer SEO to easily optimize content as its created. Other AI SEO tools can help writers generate outlines that are grounded in keyword research and pattern matched to the structure of high-ranking posts.

About the AI writing experiment 

I started this AI writing experiment by crafting a content brief for my writer. The content brief asked for an article of at least 1500 words that listed out underrated marketing thought leaders to follow on either blogs, newsletters, or social media.

I asked the writer to gather a minimum of 15 thought leaders spread across email marketers, content marketers, PPC/paid media marketers, and marketing executives. I specified that the industry was marketing and set the primary keyword as marketing thought leaders. The tone was conversational but professional. 

The finished piece linked out to each marketers’ websites, social media sites, and Substacks. You can find the finished blog, 15 Underrated Marketing Thought Leaders to Follow in 2021, live on my site. 

Once she finished the blog, I input the same brief details — including the primary keyword, topic, and tone — into Jasper. Since the writer followed a more listicle-style approach for the finished blog, I had to use multiple templates to replicate the same approach.

I’ll discuss the “writing” process in more detail after I’ve shared the results of the experiment. 

Analyzing the results 

So, did the AI-generated blog post blow the human-created blog post out of the water? I used nine metrics to compare the two.

Blog Post Headline 

Did it capture the blog sentiment and immediately engage a reader? 

Blog Post Introduction

Did it match the intent and tone as per the provided brief? Was it to-the-point and engaging or fluff-filled and generic?

Blog Post Conclusion 

Did it sum up the high points of the blog? Was there a clear call to action? 

Originality 

Did the content offer a unique angle or was it copycat content?

Expertise

Did the content read as well-researched or did it come across as regurgitated copy from page one of Google? Were there any links to credible sources? 

Grammar and Spelling 

Did it appear structurally sound or was it rampant with grammar or spelling mistakes? 

Plagiarism 

Did it withstand a pass through plagiarism-flagging apps, like Copyscape, Grammarly, or Hemingway? 

Ease of Use 

Did it truly take a fraction of the time to create this content? Was the Jasper platform simple to navigate? 

Value for Money

Did the finished product match the price tag for a Jasper subscription? Was the AI content even passable?

For full transparency, I’ve included a view-only link to the finished, AI-generated version of the blog so you can judge the quality for yourself.

[ 15 Underrated Marketing Thought Leaders to Follow in 2021 Part II (Jarvis AI)]

1. Blog post headline 

My writer titled the original blog post “15 Underrated Marketing Thought Leaders to Follow in 2021.” Jasper titled the AI-generated blog post “Incredible Marketing Thought Leaders Who Have Smaller Social Media Presence (But Great Content).” 

Do both headlines get the point across? Yes. However, it’s clear that the human-written headline is more succinct to immediately capture the theme of the blog and draw in readers. Compared to the human headline, the Jasper one is simply too wordy.

Not to mention, there are no Recipes for blog post headlines, only Templates. Unfortunately, the Perfect Headline Template feels more geared toward product descriptions, not blogs. The SEO – Blog Posts- Title and Meta Description Template required a blog title to generate a new headline, which defeated the purpose of the experiment.

The best results came from the Video Titles Template meant for YouTube, which certainly was not the most obvious choice for platform users.

For the blog post headline, I give Jasper a 2/5.

2. Blog post introduction

The best way to describe the blog post introduction generated by Jasper is like a high school essay. The introduction does nothing to relate to readers or draw them in with unique insight, nor does it clearly capture the intent of the post. Compared to the writer’s post, the AI intro is lackluster.

I used the Blog Post Intro Paragraph Template to generate the best introduction possible. With this Template, Jasper produces a standard three outputs for your noted inputs, which include the blog title and tone of voice plus the audience. 

It took about four iterations before the provided output seemed to match the blog intent, even when I fed the platform more details than I gave my actual writer.

For the blog post introduction, I give Jasper a 2/5.

3. Blog post conclusion 

The blog post conclusion generated by Jasper was more effective than the introduction; however, given the quality of the introduction copy, that’s not saying much. I used the Blog Post Conclusion Paragraph Template on Jasper and input the main points of the blog post, my desired call-to-action, and the tone of voice.

Like the introduction copy, it took roughly four outputs before I was pleased with the conclusion copy (note: I use the term “pleased” lightly). The second sentence of the conclusion was perhaps the most bizarre:

“It’s always good to take a step back from the fray every once in a while and consider what other people are thinking about your industry.”

It’s important to also note that the conclusion did point to a CTA, which was, “Tell us why below or share them on social media!” Unfortunately, I fed the platform a CTA to follow the marketing leaders on social media, which it apparently ignored completely. 

For the blog post introduction, I give Jasper a 1.5/5.

4. Originality 

It’s difficult to judge a tool like Jasper on originality since the primary focus of the platform is to use machine learning for the purpose of pattern matching. Jasper uses existing content as a blueprint, so the output will always be some form of copycat content unless you have a distinct POV and are giving the tool clear inputs.

If you want something totally unique but can’t be bothered to write it yourself, you can let Jasper paraphrase off what you feed it. For instance, you can feed the platform phases like “email marketing thought leader” and let the tool do the rest. I did something similar for the bulk of this AI post, which used a collection of Jasper Templates. 

Yet, for as much as you can give Jasper to work with, you can’t guarantee a positive result. Take a look at the section created for Ginny Martin to know what I mean. Could this last sentence use the word “marketing” any more? (I really hope not.)

“Her marketing tweets will be especially useful marketing managers looking to optimize their PPC marketing campaigns.”

Marketing is mentioned three times in one non-sensical sentence.

For originality, I rate Jasper a 1.5/5.

5. Expertise

Let me make this crystal clear: AI writing tools are only as good as the input you provide. 

By this, I mean that Jasper’s expertise will only compare to the expertise and context you feed it. You can either insert informational prompts that include key facts and statistics or hope the brief you’ve provided will spur the platform to collect relevant facts or statistics. 

Perhaps it was that the subject matter of this particular blog post centered around individuals rather than blanket topics (like email marketing tips), but Jasper failed to come up with much of any accurate information. 

The tool continued to pull “previous work experience” for the names of the marketers I provided — many for which I even included their current role and company. However, the majority of what the platform provided was totally inaccurate:

  • Continued to name random companies for Amanda Natividad, even after four separate attempts
  • Incorrectly classified Joel Klettke despite me feeding in a link to his company and Linkedin title.
  • Could not locate the correct Kath Pay, even after I wrote in the information myself for it to paraphrase
  • Failed to embellish on Jon-Stephen Stansel beyond “marketing on Facebook marketing tips” (what?)

Additionally, Jasper could not pull relevant inbound or outbound links into the content. If I wanted to hyperlink to any of the marketing thought leaders listed in the post or to their social media, I would need to do so manually, adding more time into the process. 

For the purpose of this experiment, Jasper gets a measly 1/5. (Note: This score may be higher for a more generic blog post, like a list of marketing tips or the definition of certain strategies.)

6. Grammar and spelling

Though the grammar and spelling capabilities of Jasper weren’t terrible, they definitely were not up to par with a skilled copywriter. The more I used Jasper, the more it continued to spit out the word “marketing,” in what I assume was an effort to be on target with my keyword. That effort fell flat.

For perspective, Jasper used the word “marketing” 198 times in a 1500-word blog post. An actual writer used the word “marketing” 73 times in a 2500-word post. The keyword density for Jasper reached 13% (way too much) whereas the density for the writer was nearly 3% (just right). 

Besides the clear lack of synonyms, what’s worse is that Jasper ended every other sentence with “…etc.” Even after light editing to piece each output together, I was still left with 14 different instances of etc. — 14 too many. 

While there was no obvious grammar and spelling faux pas, Jasper gets a 3/5 for this category for sentence structure and odd quirks alone. 

7. Plagiarism 

A pro search on a plagiarism checker like Copyscape did not flag the Jasperpost for direct plagiarism.

I will rate Jasper a 5/5 for direct plagiarism since I could technically publish this piece without ripping off another writer’s work; however, I hold myself and all writers to an equally strict plagiarism rule, so this score does not impress me much. 

8. Ease of use 

At face value, the layout and options on Jasper make it appear quite simple to use. If you plan on using it for email subject lines or social copy, it’s likely to be as straightforward as it seems. If you plan on using Jasper for long-form content as I did in this experiment, it likely won’t be as easy.

There are several weird quirks of Jasper in particular that will incur a learning curve. It takes personal trial and error to learn which Templates work better than others, or when a Recipe works better than a Template could. (This is not intuitive for beginners!)

There are even Templates that Jasper notes are for specific copy tasks, but ultimately don’t spit out what you need. For instance, I was able to have more success with body paragraphs using the Blog Post Outline Template, and the Video Title Template was more effective for a headline than the actual Perfect Headline Template. 

It’s important to note here that most Templates and Recipes each have an explainer video that goes into more detail about the type of pattern matching or machine learning used. These can be helpful if you want to use Jasper long-term, but could also waste time if you were choosing Jasper for its speed and efficiency. 

For ease of use, I rate Jasper a 2.5/5. 

9. Value for money

Jasper is a paid software available at three different price points: Starter, which is $29/month; Boss Mode, which is $59/month; and Boss Mode Pro, which is $119/month. For the purpose of this experiment, I opted into Boss Mode Pro to access all Jasper resources possible.

However, prior to this experiment, I had been on Boss Mode for two months and was not super impressed. So, I personally downgraded to the cheaper Starter plan, which was essentially the same thing. The only big difference between the two plans is that Starter generates 30,000 words per month, whereas Boss Mode is unlimited.

Personally, I do not think the Boss Mode plan is worth the money. If you want to experiment with AI writing tools like I did, opt for the Starter plan to save nearly $100 per month. 

Bear in mind, these once-monthly prices are significantly cheaper than the cost to put an experienced copywriter on a monthly retainer. Even the Boss Mode plan is considerably cheaper than what it would cost to have a copywriter provide multiple pieces per month. 

Yet, given the quality of the content, you definitely get what you pay for with Jasper. All things considered, I give Jasper a 3.5/5 for value for money.

In total, Jasper received an average of 2.6/5 for this experiment. 

When are AI writing tools a good idea? 

The Jasper AI writing tool left me feeling pretty underwhelmed after this particular experiment — but I won’t completely count out all AI writing tools. I see them as a way to generate ideas and beat writer’s block at the beginning of a post, as well as a method to optimize at the very end of a post. 

AI tools in general are awesome for SEO optimization since they pack a ton of research into one tool. Something I do with my clients is run the content through either Frase or Clearscope to ensure the copy is fully optimized for my target keyword.

I imagine you can also use an AI tool like Jasper to speed up the process of creating programmatic SEO landing pages based on one finished page. 

Likewise, AI writing tools can shine for short-form purposes, like meta descriptions, headline variations, and batching social media copy.

The bottom line: AI writing tools are not great for actually executing content, but they do serve as a helpful addition to a copywriter’s existing tech stack. 

Will AI replace real writers? 

In my opinion, AI writing tools will not replace real writers any time soon. AI tools are only as reliable as user inputs, and AI software can only produce content on topics that have already been written about before. By their nature, they default to writing more copycat content. They offer no unique or different perspectives, which is what makes great copy truly shine.

At the end of the day, I still use Jasper on occasion for very specific reasons, like writing variations of social media copy. But I don’t think any excellent, good, or even average writer needs to worry about AI tools replacing them anytime soon. Instead, think of AI tools as a method to help you do your job faster.

About the author

Jessica Malnik

Add comment