When Your Language Textbook Has Never Met a Native Speaker: 5 Easy Ways to Detect AI-Made Language Books

 

(All the words in this article were written by a human (me). The images, however, were artificially generated – obviously!) 

AI could certainly be a helpful assistant when creating language learning materials and I suppose it’s inevitable that it will be used to some extent in this day and age. But when a book is fully AI generated, especially by a non-native speaker (or someone who has never even learned to count to 5 in the language), this can be a dangerous thing indeed.

Unlike a fiction novel generated by a robot, non-fiction books (such as foreign language textbooks, dictionaries and slang word lists) can do serious harm when learners are taught things that are incorrect, non-existent, or just not very native-like. They may also become discouraged and simply give up when things get confusing. Even worse, when AI is used to create materials for lesser-studied and endangered languages, this can actually harm language revitalization and discourage genuine native speakers from creating real materials.

While it’s virtually impossible to always be 100% certain that your language course was written by a knowledgeable and ethical human being, here are five tips to help weed out the obvious fakes:

1) Polyglot Author… or Polyglot Scammer?

One of the easiest ways to detect possible deception is to click on the author link and see what other books that author has created. If the same author has created materials in dozens of languages, especially less popular languages, it doesn’t take a PhD in linguistics to realize that they were most likely auto-generated.

Sometimes this kind of author will include a fake biography, praising themselves for having learned 43 languages by the age of 22, or having attended some prestigious university. I’ve even seen books where the author’s photo is AI generated.

Sure, there are many people, including myself, who have studied and who speak many languages, but that doesn’t mean that we are competent to sell textbooks or word lists in Welsh, Japanese, Polish, Ancient Greek, Maltese, Sentinelese, Coptic…

In some cases, you will see that this author has also generated hundreds of books in other subjects (after all, why limit your scam to just language learning!)

2) Plot Twist 1: Polyglot Author… Polyglot Names

In a twist to the above method, in order to mask their deception, some “authors” who aren’t quite as lazy (but just as greedy) will take an extra step and generate a new author name for each title in order to make it appear that each book was written by a separate individual. Most of the time, the name that is chosen will resemble the name of a native speaker of the language in question. Like, “Spanish Grammar Made Embarrassingly Easy” by Jose Carlos Santiago; or “5000 Incredibly Awesome Icelandic Slang Words” by Agnetha Bryndísarson.

In this case, you can search for the publisher's name to see if they are all published by the same source, or search for the title without the language name (they are usually too lazy to create unique book titles) and you will see a long list of books in hundreds of languages that all look predictably similar.

All of this is designed to fool you into believing that the book was written by a native speaker.

     3) Plot Twist 2: Meet the Invisible Author

     In another iteration of this grift, the author is completely invisible. Think about it, if you took the time to write a book, especially a textbook, wouldn’t you want to take credit for it? Sometimes you will find books where there is no author at all. Instead, the book was written by a company (ex: Top 1000 Finnish Words by The International Linguistic Institute of Linguistic Institutes).

Now, of course, it is entirely plausible (but, let’s face it, not very likely) that this fictitious institute, society, university, center, foundation, guild, or “academy” of languages has paid 100s of native speakers to write textbooks or create word lists.

Even if this were true, there is practically no way for this “institute” to verify the accuracy of the content of all these books in all these languages or even determine if the paid ghostwriter didn’t just use AI to generate the book (spoiler alert – they probably did!). And why not include the author's name and biography to verify authenticity and help boost sales.

4) Don’t Judge a Book by Its Cover - Except When You Can

Ok, I’ll admit it, some AI book covers can be attractive. And who wouldn’t want their book to be visually appealing? Like this article, just because the cover image was algorithmically conjured doesn’t mean the contents weren’t painstakingly crafted (yes, Karen, that’s an AI buzzword). However, it does probably mean that the book wasn’t published by a reputable publishing company (or at least one that has enough faith in their product to invest in paid cover art).

In addition, sometimes AI cover art can contain errors (like the spelling errors in the title image to this post – in case you think I didn’t notice. But don’t worry, someone will point it out to me anyway). If I really spoke the language and could recognize these mistakes, why wouldn’t I regenerate the cover for my magnum opus? I have even seen some AI books where the name of the language itself is misspelled. Sloppy on the outside, synthetic on the inside.

    5) When the Language is Rare, Take More Care

Back when I started learning languages (45 years ago), it was extremely difficult to find resources. There was no internet, no YouTube, no Babbel. It was a linguistic wasteland. I always joke that I couldn’t have learned Japanese or Swahili in my small town even if I had wanted to. Instead, I had to make do with my one (by then 10-year old) high school Spanish textbook and my tattered copy of “1001 Pitfalls of German”.

Today, there are an estimated 345,000 books to learn Spanish, 290,000 to learn German, and those are just the numbers I completely made up! So it makes sense that people who are out to scam the public would wish to publish materials in languages without a lot of competition. In this way, when you search for books to learn Lakota, there’s will appear near the top of the list. When you find a book like “The 500 Most Used Albanian Slang Expressions”, “The Ultimate Beginner’s Guide to Silbo Gomero” or “Navajo Spoken Idioms for Beginners”, extra caution should be used to make sure that the author actually has the credentials to write this book.

 *****

Perhaps this all seems like way too much work (but honestly, if it does, maybe you should study geology instead). In an effort to help learners find quality language materials (and to earn a little extra cash to buy more books), I created a list of language learning books and podcasts on a special Amazon bookstore (here: https://www.amazon.com/shop/languagecrawler) using the above-mentioned tips and by purchasing or viewing a full copy of the books (or at least a reasonably long sample). Of course, being human (yes, some of us still do write), I acknowledge that some titles may have slipped through the cracks. But just let me know and they will be promptly removed!

 

 Please use the comment section below if you have any other ways to detect fakes!