Artificial Intelligence Writes Texts

November 21, 2019

Artificial Intelligence

Artificial intelligence has long been a reality in copywriting. On the basis of structured data, machines create product descriptions, stock market information, and weather forecasts. So well in fact that the results can hardly be distinguished from compositions produced by humans. But AI can do even more: the creation of fairy tales and stories is also possible at the push of a button.

Machines Write Poetry

The members of the Brentano Society were astonished. They had just accepted a new poem for inclusion in their annual journal when they found out that a machine was the author of the piece. The lyrical work “Glimpses of Sun on the Run”, which was accepted into the Frankfurt Library, was attributable solely to an artificial intelligence.

The human work on this piece was limited only to the programming and the data input: numerous poems provided study material for the algorithm. What came out of that could absolutely be measured against poetry produced by man. Even experts at the renowned cultural institution in Frankfurt am Main couldn’t tell the difference.

So, artificial intelligence can write poetry. But AI cannot only write verse, but also prose. One example of this is Botnik – a project by authors, artists and programmers. Together with a computer program, Botnik developed a fairy tale. This is how the cooperation between human and machine worked:

  • The self-learning algorithm received Grimm’s Fairy Tales as the data pool
  • On that basis, the program suggested individual new sentences in the style of the well-known works
  • Based on these machine suggestions, the authors assembled a new self-contained story.

Out of that came the fairy tale “The Princess and the Fox” – the newest work by the Brothers Grimm, written at the beginning of the 21st century. In style and content, a familiar fairy tale, but in the end only a literary clone.

AI Copy for Product Descriptions

Poems, fairy tales and a chapter for a Harry Potter book or a new Rembrandt painting are, however, primarily gimmicks for the time being, which challenge the unique position of humans regarding artistic creativity. But the field of application for artificial intelligence in copywriting is not primarily the entertainment industry. It lends itself particularly for business use. Some examples are:

It’s evident that the more structured and clear the content to be conveyed is, the better the copy can be machine created. Conventional product descriptions also frequently exhibit similarities because they are modeled on specific templates. A machine can be taught these templates quickly, or and this is where AI comes into play, the program recognizes the pattern itself and applies it.

How a Text Robot Works

How do text passage creating programs work for online stores? The production of product descriptions functions according to a specific system: the text to be created is broken down into individual units. The program receives an overview of the product features in list form, which are then to be communicated in a text format. An example for bicycles could be:

  • Bicycle type (city bike, e-bike, trekking bike, racing bike, etc.)
  • Number of gears
  • Weight.

Depending on the type of bicycle, the algorithm choses the appropriate sentences for the intended purpose (“A perfect everyday bike”, “A bike for a rider with sporty ambitions” , etc.). The number of gears is also suitable for use in various text passages (“… has x gears”, “… impresses with the number of gear options”, “Only 3 speeds makes shifting simple.”, etc.) The same is true for the weight (“weighs in at only 10 kg”, “a stout 17 kg makes this bike especially robust.”).

In this way descriptions of varying length for the same product are also possible, from the detailed introduction on the product page to the meta tag description, which, with fewer than 155 characters, is ideal as the short entry on a Google results list.

Based on grammatical rules, the algorithm churns out new combinations non-stop: creativity at the push of a button. The only task a human assumes here, if at all, is choosing the best results. In an ideal case the artificial intelligence requires only a large amount of similar copies to autonomously recognize structure, linguistic style and other characteristics.

Individual AI training data for the development of AI systems can be easily gained via the clickworker crowd.

Automatic text generation based on structured data is especially beneficial to online store owners. Formulating countless product descriptions that should each appear unique despite similar content is not only laborious, but also error prone. But for each detail a computer program chooses from a variety of different wordings und in doing so creates unique content in seconds, which is what search engines love to see.

AI Allows Machine Copy to Sound Human

Of course, automatic text generation is possible without artificial intelligence. But the advantage of self-learning programs is that their algorithms recognize linguistic features in the source material. Learning analysis is therefore the basis of authenticity, which is a big plus that customers appreciate. And this advantage benefits every website operator.

Thus, programs construct any desired amount of text reasonably, quickly and reliably, with lucrative benefits for the user:

  • Potential consumers are personally addressed because the program learns the way the target audience speaks
  • Text passages created this way increase the conversion rate
  • The entries are classified by Google as unique content and rise in ranking.

Particularly the avoidance of duplicate content is a valuable factor in online marketing. An intelligent copy creating program ensures text uniqueness much more reliably than a person. This is true because the algorithm can automatically compare its results even during creation with already existing copies on the net.


What has long been reality in online marketing seems to have become interesting for the entertainment industry, too: machine produced poems, novels and other types of writing. The poetic and prosaic results of computer texts today must usually be revised by a person before they can be published, but over the long-term the question will arise whether only humans have a copyright on artistic creativity.


Dieser Artikel wurde am 21.November 2019 von Jan Knupper geschrieben.


Jan Knupper