All Eyes On AI: Future of the Content World

‘Alexa, play that Green Day song’ is now a reality, but she might not be able to write original songs in the near future (at least not Like Humans Do). Yes, we are back to the same debate – can AI do everything out there under the sun?

The use of cutting-edge AI technologies to attain customer success has almost become imperative. There’s Natural Language Processing, Machine Learning, and a whole lot of other new technological names that pop up every other day, but is there any tech for the content marketing world, to ensure both, quality of content and results?

In a bid to find out if they are really impactful, Keywee, a digital solutions company, decided to run an experiment. Conducting ‘sanity tests’ and tinkering with content to get better reach was not new for them.

This time around, though, they were doing it to find out if AI can write better and get the best results, compared to human work.

Premise

Their AI algorithm had been trained using many models for a long time, so they assumed that it could provide interesting findings. The core metric used to find how well it performed is called the performance predictor, which ranks a piece of content on a scale of 0 to 100.

For each article on their website, they picked up two post-texts, with a performance score difference of at least 20 points. It meant that a post that went up on their blog would have at least two iterations.

Post-Text 1: You Would Never Find What Lies Beyond These Woods.
Performance Score 1: 69

Post-Text 2: These Tales Are Really Terrifying.
Performance Score 2: 93

These scores mean that the one with the higher score is expected to do well. And the success was measured using two main marketing metrics:

CPC – Cost Per Click

It gives an idea of how much did it cost to acquire every user who viewed the article. A low CPC is an ideal scenario for marketers.

2- CTR – Click Through Rate

Of those numerous viewers who saw the post-text, how many actually went ahead and clicked on the article’s link. Marketers always try to maximize their CTR by means of their activities.

Experiment

The A/B test involved using the model and was meant to predict how well each of the post-texts. The finding was nothing but fascinating!

Headline 1’s CPC turned out to be 12% lower than Post-Text 2, whereas its CTR was almost 11% higher. In general, the ones with a greater performance score like Post-Text 2 had 18% higher CTR and 14% lower CPR.

In a bid to mix up things a little bit, Keywee’s staff members threw in their own predictions into the mix. This was one area where they couldn’t beat AI, since it did a considerably better job.

Key Findings

After pitting real-world scenarios against AI, there were a few takeaways that challenged the norms in the world of content, which are

1- These post-texts are vital to have a well-performing article or blog. With CTR going as high as 18%, no one would want to take a chance.
2- More is not merrier! Shorter post-texts can perform much better than their counterparts.
3- Surprisingly, the usual CTAs like ‘Read Now’ and ‘Learn Here’ were not as impactful as the editorial texts.

It is evident that AI still has a long way to go when it comes to the content arena. Nevertheless, we humans can continue writing and testing, while churning out some amazing results!