The human rater rates you on language use (grammar and vocabulary) and topic development (the content of your answer) including the details, and whether or not you were accurate in your information.
THE AI SOFTWARE:
The software will check your words per minute. You should speak quickly, and avoid pauses, because that will affect your score in a negative way. It also checks if you pause OFTEN. In other words, if you only pause once, it may not affect your score. But if you have many pauses, then it will negatively affect your score.
The software program also checks for repetition. AVOID REPETITION. If you say, “I enjoy softball because of the exercise I get… softball is a good form of exercise because it exercises all my muscles….” Then the repetition of “exercise” and “softball” may negatively impact your score.
Furthermore, the software checks the LENGTH of your response. According to Josh MacPherson of TST Prep, a good word count goal is 115 words for the independent task, and 130 words for the integrated tasks.
It also is able to check your INTONATION and your RHYTHM, so practice imitating a native speaker.
What does this mean overall? It is hard to say. According to the ETS, the speaking scores should remain the same as they were before they introduced artificial intelligence into their ratings.
However, keep this in mind. There is only ONE person grading your response. That means there is more “luck” involved, in my view. In others words, it was always good to have two, because if one rater was extremely strict, the other rater balanced it out. This may be a reason to get a re-score, especially if you are only one point below your minimum speaking requirement. However-- THIS IS JUST AN OPINION. Feel free to ignore it!
I must thank Michael Goodine from TOEFL RESOURCES for publicizing this information. Without him this blog post would not have been possible. His wonderful website is:
You may see his video version of this at: