Bot or Not? How Good has Text Automation Become? We Put it to the Test!

How do Google and other search engines rate content generated by text automation processes? Are there differences between copy written by humans and text generated by a machine? ‘Bot or not?’ is the big question here!

Text automation vs. Google

Particularly since the most recent Google Helpful Content Update and the Google September 2022 Core Update, many content and SEO managers are concerned that automatically generated content, such as website copy or product descriptions, might be punished in search engine rankings. Google has announced that content written by humans, for humans – content that offers real added value for people – will be ranked higher. Content designed exclusively for SEO relevance, on the other hand, will be given less visibility and reach. Websites and online stores with large volumes of this latter type of content will receive poorer rankings overall.

Given the heated discussion surrounding text automation, however, it is important to understand that Google cannot distinguish between copy written by humans and automatically generated text, as long as the text automation process produces high-quality results. Conversely, this also means it makes no difference whether the copy is written by content and SEO managers or generated by an algorithm, because the only important aspect is the quality of the text. The decisive factors here are that the published copy must be of an adequate level of quality and must offer actual added value; the text also has to provide the information that users need and match the intention of the search.

Hugging Face Output Detector vs. Text Automation

At Retresco, we subjected the latest product descriptions published by one of our customers to a quality check. We used the free output detector demo tool huggingface.co. The detector is based on the RoBERTa transformer model and requires a minimum of 50 tokens for a reliable evaluation.

In the following, we will present examples of our results based on evaluations of text published by the leading distributor of workplace products and services, Lyreco (Switzerland). Our tests analysed a range of English-language product descriptions.

Lyreco (Switzerland)

Safety shoe product description (last updated: 6 December 2022):

Safety shoe Safety Jogger Ligero, black, size 45

Here is the analysis from the Hugging Face Output Detector:

241 tokens were evaluated – and the text was rated as having an impressive 99.98% probability of being real. Wikipedia text written and edited by humans received the same high rating. This proves that Lyreco’s automatically generated product descriptions are of excellent quality and meet the highest standards.

PanzerGlass screen protector product description (last updated: 6 December 2022):

Screen Protector PanzerGlass, Privacy, iPhone 12 Pro Max

Here is the analysis from the Hugging Face Output Detector:

In this example, we far exceeded the recommended minimum of 50 tokens. The automatically generated product description was rated as highly likely to be real, at 99.97%.

Text Automation: The Bottom Line

Bot or not? Text quality is everything! Verification with the Hugging Face Output Detector proves that product descriptions created using text automation processes cannot be identified as machine-generated, because the quality of the copy meets the highest standards – the text cannot be distinguished from content written by human beings.

With our content automation platform, users can transform data into high-reach content in a wide range of languages and across industries. Not only is the copy grammatically correct; the content is also of exceptional quality. And Google ranks these types of texts correspondingly high.

If you have questions about text automation or Google updates, or if you would like further information, we would be happy to help. Please feel free to get in touch with us – we’ll get back to you straight away!