‘Content marketing’ often amounts to nothing more than companies filling their blogs up with posts that try to make them seem smarter than they are, while saying nothing of any value to anyone at all.
And if you’re going to get into that game, why spend time and effort crafting a 500-word article when you can buy a computer-generated one for a dollar? That’s the idea behind Articoolo, a startup that can generate articles on a range of topics based on up to five keywords.
Ever been to a tech festival?
TNW Conference won best European Event 2016 for our festival vibe. See what's in store for 2017.
While some ‘AI journalist’ products like Quill are used to write finished articles that can be published by news websites, Articoolo doesn’t pretend that its articles will be perfect. Instead the idea is that you take the piece it’s created and tweak it with your own personal touch and clean up things like flaky grammar that may have snuck in.
“It can help you write a 500-word article in five minutes instead of two hours,” as CEO and co-founder Doron Tal explains.
So what are the results like? Tal told me that Articoolo deals best with broad topics, so I entered ‘shiba inu dogs’ and selected the ‘shiba inu secrets’ suggestion it offered (mainly because I like the idea of dogs conspiring against humans with their own secrets).
Two minutes later Articoolo had generated a 272-word article about shiba inu secrets with “100 percent uniqueness.”
Until you pay up for your article, you only get to see snippets of the piece. All but a few sentences are blurred out.
As you can see from legible sentences above, the articles aren’t perfect. In my tests I found occasional grammar errors, some missing words and, in this case, a lower-case letter starting a sentence. Still, cleaning up those mistakes is easier than writing your own piece from scratch, right?
After I’d ‘bought’ the full article with the credits Articoolo had given me to test the service, I found the finished article to be perfectly readable, if light on the ‘secrets’ I was hoping for. But was it ‘unique?’ Not quite. A quick Google search found the article I would have paid money for to be a trimmed down and lightly rewritten version of this 2009 piece posted to EzineArticles, a site often associated with low-quality content designed to rank highly in search engines.
Another article I generated via Articoolo, about ‘the rise of content marketing,’ turned out to be a rewrite of this 2008 EzineArticles piece that predates the rise of content marketing by several years.
Tal tells me that Articoolo uses more than just EzineArticles as a source. “Usually, articles our algorithm creates are structured from more than one source. It is really depends on the sort of topic you wish to write about and the initial semantic (analysis) of it.”
He said that using the ‘enhanced uniqueness’ option would result in an article that was more likely to be based on more than one source.
Trying that out on the topic of ‘shiba inu dogs,’ the article I got back used multiple sources and was certainly ‘more unique’ as a result. But the way it had rewritten information from those sources rendered the finished product nonsensical in some cases.
“Their mind is proportionate to the size of their body,” it confusingly claimed, in a sentence I discovered was taken from Dog Breed Info Center. The original sentence said, “the head is in proportion with the body,” which makes a lot more sense.
If you don’t want software to write you an article based on an old information from content farms, Articoolo can also ‘rewrite’ any article you paste in, to create original content from someone else’s work of your choosing.
I decided to see what it would make of my recent look at call center gamification startup EvaluAgent. The result? My 379-word original became an ugly 232-word regurgitation.
Content is cheap
Articoolo doesn’t claim that it generates perfect articles, it accepts that some editing will be required. And at prices ranging from $0.99 to $1.25 per article, you should hardly expect amazing quality. Still, its very existence perpetuates the idea that content is cheap, and just there to fill a hole.
Who cares if your article is a lightly reworded version of something someone wrote to rank highly in search engines eight years ago? It’s just ‘content,’ right?
Tal denies that Articoolo simply rewrites other people’s work. He says that it’s more a case of doing exactly what a human would:
We developed the algorithm to imitate a human writer(‘s) modus operandi when writing an article. A human writer would first Google the topic and look for sources on the Web, then summarize all the sources he found and after constructing it to one coherent text he will rephrase it to avoid plagiarizing. This is exactly how the algorithm operates, so we actually do not change anything, only make it quicker and easier.
In other words, people use other people’s work anyway, we just simplify the process.
I’d argue that this shows a misunderstanding of how (good) writers work. Good writers use information from multiple sources, but wouldn’t just mix it up, collage-style, and change a few words. In many cases they’d link to sources, too.
Still, if some people are going to take the quasi-plagiaristic approach (and no doubt they do), I suppose a tool to help them do it faster isn’t the end of the world. You could argue that as a starting point for an original article, Articoolo is a timesaver for laying out a framework to build upon. But without knowing the sources of the information included, how can you check that it’s accurate? You have to go back and do the research anyway.
As someone who agrees with the theory that robots and artificial intelligence will eventually swallow up white collar jobs, ‘journalists aren’t safe from automation’ is an example I often give to people.
The good news is that in its current form, Articoolo doesn’t threaten the jobs of skilled writers. But it does contribute to the assumption by many marketers and publishers that content is cheap, and quality and originality are secondary concerns. And that makes me very uncomfortable indeed.