If it sounds too good to be true - by Bob Newman

 I am the husband of Elizabeth Kay, of this parish. Her blog post the other day was about an attempt to scam her with promises of a film adaptation of one of her books. That book was, incidentally, a fairly unlikely choice from among her oeuvre, though admittedly it had the distinction that it was the only one to have been translated into Irish!

 Elizabeth promised a follow-up post from me, focusing on the AI aspects, so here we are.

 Before we get on to the AI, a couple of other related matters. First, the email address used by the supposed bigshot film producer, who we are calling Sheldon, didn’t look right. It was an ordinary AOL address, and we would have expected a major film producer like him to have been using something more corporate-looking. And the name of the account began with “official_” – which was perhaps protesting a little too much? He also gave us the URL for the website of his production company, which was superficially impressive – convincingly swish and glossy. However, the home page didn’t contain any information at all, and there didn’t seem to be any other pages. The only way to make any progress from that site was to download its very own app, about which no information was available. This felt like a remarkably bad idea, so I didn’t do it.

At Elizabeth’s suggestion, I set out to establish whether the emails she had received really had been generated by AI - presumably a large language model such as ChatGPT, Claude, or Gemini. There are free AI tools that enable you to check whether text is AI-generated, detect plagiarism, and so on.

So you can set an AI to catch an AI.

You can also ask another AI how to go about it! This is what I did. The AI features within Google recommended Grammarly, Copyleaks and GPTZero.

Grammarly reckoned the probability that Sheldon’s first email was AI-generated was 63%, but it wouldn’t let me do any more analyses unless I signed up. So I tried Copyleaks and GPTZero.

They both took a more black-and-white view of matters. Both 

of them were 100% confident that every one of the emails

from Sheldon and Bernie had been generated by AI. (Bernie is the name we are using for the supposed screenwriting expert whose services Sheldon was pushing.) As a control, I also tried Copyleaks and GPTZero on a chunk of my own diary. Both of them were 100% confident that that had been written by a human, which was reassuring (on more than one level!) I did eventually sign up to a free account with GPTZero, so that I could try out some of its more advanced facilities. It highlighted some of the most obviously AI-generated sentences in the emails, which were:

•    That's a story about the distance between what we have and what we truly need — and the people who help us bridge that gap.

•    But she has something he's never known - purpose, resilience, and a connection to the world that runs deeper than any possession.

•    I wanted to follow up regarding my previous email and the screenplay development options I shared.

•    I completely understand that these decisions take time, and there is certainly no pressure.

Elizabeth then suggested I should try to establish whether Sheldon and Bernie were both using the same AI. The results from GPTZero were inconclusive. They might be using different AIs, or they might be using the same one but with different parameter settings. Then I asked Google’s AI the same question, and it gave a much clearer answer:

[Sheldon] reads like an AI prompted to write a "deeply moving, artistic compliment from a Hollywood producer." [Bernie] reads like an AI prompted to write a "polite, corporate follow-up email about sales packages." Because the structural DNA and rhythmic patterns are so different, they were likely generated by two separate models.

 

If you are communicating with this person, it appears they are using one AI tool to generate flattering book reviews, and a separate CRM/sales AI tool to handle their business follow-ups.

 

Would you like me to help you draft a reply to this outreach, or would you like to scan these texts for any specific red flags common in screenplay adaptation queries?

 

Well yes, tell us about those red flags, please. The results from this were interesting too, and I’ll give them here in full.

This message contains multiple severe red flags typical of "vanity press" style scams targeting self-published or indie authors. Legitimate film producers or literary agents do not operate this way. Here is a breakdown of the specific warning signs in the text:

1. The "Package" Business Model (The Biggest Red Flag)

•          The Text: "the screenplay development options I shared," "any of the packages I outlined," "please let me know which package interests you most."

•          The Reality: In Hollywood, optioning a book means the producer pays you for the rights to your story. If someone asks you to choose or buy a "package" to develop your screenplay, they are a fee-charging service, not a film producer. You should never pay to have your own book adapted.

2. Contradictory Messaging ("The Bait and Switch")

•          The Text: [Sheldon] starts with "No pitch. No pressure. Just deep respect for what you've made." [Bernie] immediately pivots to "Whenever you've had a chance to review the options, please let me know which package interests you most..."

•          The Reality: [Sheldon] uses high-flattery to lower your guard. [Bernie’s] is a high-pressure sales pitch disguised as a casual check-in.

3. Vague Praise and Lack of Specifics

•          The Text: "Since reviewing portions of your manuscript..." and "there are some very promising cinematic opportunities..."

•          The Reality: They admit they have only read "portions" of the book. The praise is incredibly generic ("the premise is engaging"). They do not mention specific scenes, specific character arcs, or unique dialogue. This is because a machine generated the summary based on your book's public blurb, and no one has actually read your book.

4. Overwhelming Corporate Buzzwords

•          The Text: Screenplay development options, adaptation process, packages outlined, narrative, development approach, next stage of development.

•          The Reality: Legitimate film professionals talk about directors, talent, distribution, comps (similar movies), and financing. Scammers talk about "processes" and "packages" because they are selling you an administrative service (a poorly written script), not making a movie.

Summary

This is a predatory solicitation. They want you to pay them thousands of dollars to write a screenplay version of your book under the false hope that they will pitch it to Hollywood. Once you pay for the "package," you will likely receive a low-quality script, and they will disappear.

Would you like help drafting a firm rejection email to get them to stop contacting you, or would you like to know how to verify their company name to see if they are on a known publishing blacklist?

There’s nothing here that any of us is likely to find very surprising, but I think it’s helpful to have it spelt out so clearly. (Actually, I confess I hadn’t realised that published authors were likely to be targeted, vanity press style.) But even though we know intellectually that scammers are out there, the natural human instinct to take things at face value is liable to get us into trouble if we’re not careful. So be careful! Particularly if you don’t have an agent to handle this stuff for you.

Evidently AI has already penetrated more deeply into our lives than I had thought. We need to be aware that almost anything we encounter - text, image or even video - is liable to be AI-generated, so we must be on our guard. It’s a jungle out there.

It’s encouraging though to know that AI can also work for us. I was seriously impressed by the helpful analysis and advice that I was able to get from Google’s AI, with very little effort. And GPTZero, particularly, looks to be a useful tool.

Taking a step back from this investigation of our recent experiences with the scammer(s), and their apparent identity theft(s), it’s interesting that Elizabeth and I both felt instinctively that in proving (to our own satisfaction) that she had been talking to AIs, we had also proved that it must be a scam. Perhaps that could be seen as unwarranted prejudice on our part. I’m not saying it wasn’t a scam, but that actually it wasn’t the AI aspect that made it necessarily so. Already a lot of the communication one has with respectable commercial organisations is also via AIs. Look at those digital assistants on so many big sites – the noxious spawn of that misbegotten talking paperclip we all used to love to hate. The time is surely coming when virtually everything we ever get out of any large organisation is AI-generated.

AI has been around for decades, and I used to think of it as just another branch of computing, but in the past year or so it has come on by leaps and bounds. Now that it has taken off, it is going to get more and more powerful and worrying – and this is going to happen extremely fast. Already there are accusations flying around about published novels having been written by AI. Whether this is true or not, I’m certain it has the potential to be true pretty soon now. An AI-generated opus supposedly by Barbara Cartland, or Dan Brown, or Mills and Boon is easy to imagine, and I’m sure the men in suits would love it if they no longer had to deal with those pesky amateurs called “authors”.  The day is not far away when most book reviews are written by AIs too, I should imagine.

But the main message from this whole affair is an old and familiar one: if something seems too good to be true, then it almost certainly isn’t true at all.