Artificial intelligence tackles the written word

If you go to this page on my blog, you’ll find an introductory post about the use of artificial intelligence in generating written content, marketing messages, and personalized email. It won’t win any Pulitzers, but it’s easy enough to understand.

What’s remarkable about this post is that I didn’t write it.

Instead, it was created by LongShot, an AI-powered software-as-a-service application that generates written content – including long-form articles – entirely from a few keywords. Similar products include Jasper, Kafkai, Copysmith, Writesonic, and ContentBot.

Using robots to write basic articles and blog posts is nothing new. The Associated Press has been doing it since 2014. What’s new is that these powerful tools are increasingly available to ordinary people at impulse purchase prices. For example, LongShot’s $29.90 monthly plan provides about 50,000 words of written copy or the equivalent of a 200-page book.

Machines are proving adept both at writing and strengthening what we write. The free plug-in from Grammarly has become one of my most valued writing tools. I’m a mediocre proofreader, and Grammarly can spot my omissions, usage errors, and grammatical mistakes with uncanny accuracy. In addition, the paid version suggests ways to clarify, tighten, adjust tone and choose less overused words.

Marketing paydirt

It should be no surprise that marketers are among the most enthusiastic adopters of AI-generated content. Persado has a platform that combines machine learning with feedback from a human test panel to help marketers determine the best way to craft a message for a particular audience.

“We try to understand how language influences behavior and what language components evoke emotion and create action,” said Vipul Vyas, senior vice president of vertical strategy at Persado. “Instead of working from an intuition of what we think will work, our technology is about harnessing millions of people’s experiences.”

Persado’s algorithm breaks copy down into formatting, emotional appeal, and call-to-action components. First, it uses a database of words and phrases to generate all possible permutations of the message. Then, based on what it’s learned from previous campaigns, it reduces millions of options to just 16 and then tests those in email messages to a small group.

“Just a few altered words makes a substantial difference,” Vyas says, claiming that messages generated by his company’s technology work better than those created by humans 96% of the time.

Automating the pitch

Intelligent Relations is a spin-off of a public relations firm applying AI to the PR pitch letter. Its technology, which is currently available on an early-access basis, uses predictive analytics to monitor media outlets and “make intelligent assumptions about what journalists are going to cover in the future,” says Steve Marcinuk, co-founder and head of operations. “If we see a journalist has written frequent cybersecurity articles in the past, we can predict they’re likely to do so in the future.”

The technology crafts a subject line and an introduction customized to the journalist, such as “I know you’ve covered learning technologies at The Washington Post and thought we might connect on a few areas of professional overlap.” The body of the message is written by a human.

In theory, the result should be a net positive for people on both sides of the exchange since the result is fewer irrelevant messages in the journalist’s inbox. And a more eye-catching introduction for the PR pro to use to catch the target’s attention (I hate to admit it, but flattery works).

 A human can also tackle the same process, but the results are inconsistent, and people don’t scale well. “If you search for ‘facial recognition,’ you get a mixed bag of local community outlets and a sports journalist who wrote one article six months ago,” Marcinuk says. “That doesn’t indicate deep interest.”

No need for people? Not so fast.

This topic begs the question of whether these technologies will eventually make human writers obsolete. Although I have a vested interest in the issue, I doubt that will happen anytime soon.

All the technologies I’ve discussed here are essentially assistive: they make humans more productive and effective communicators. LongShot, for example, has several features aimed at unsticking a jammed thought process, such as rephrasing text and suggesting alternative headlines.

Where machines are more likely to displace human writers is in creating routine reports, memos, and summaries of long-form content, tasks I doubt many of us will miss. They’ll also generate articles and reports about events that would otherwise go uncovered, such as the 4,400 earnings stories and 5,000 NCAA basketball previews the AP publishes each year. That’s a positive for everyone.

Next, Read This:

Will AI Save Journalism – or Kill It?

AI-Generated Content: Friend or Foe for Content Creators?

Best AI writer of 2022

Why brands should use AI content generation

Pros and Cons of AI Generated Content

How to know when AI is the right solution