How Those Personalized Generative AI Scam Emails Are Built

One of the things I often hear from writers whose inboxes are crammed with incredibly personalized marketing or book club or editor impersonation emails–you know, the ones that make it seem the sender has first-hand knowledge of your book and your bio, and is really impressed with both–is “How are these things created? How do they get so many of the details right?”

The answer, of course, is generative AI, which makes it possible to scrape authors’ data from the internet and regurgitate it in response to prompts fed into a chatbot. Mostly the chatbots get it right. But if you’ve gotten more than a few of these solicitations, it won’t be news to you that they do make mistakes: someone else’s name, wrong or nonexistent book title, hallucinated bio details.

Occasionally, those mistakes provide a little insight into how the solicitations are built and targeted.

Email 1: Just Plug in the Details

This email is part of the incredibly prolific editor impersonation scam that is responsible for more than doubling the names on my Impersonation List over the past several months. The scammer who sent it either hit Send too soon or didn’t realize they hadn’t plugged in the chatbot-supplied details. Or maybe they’re a baby scammer who is still mastering the craft?

It’s obviously a template. You can see from the text highlighted in red how information harvested from the author’s website, retailer listings, reviews, etc, is meant to be plugged in, creating a uniquely personalized solicitation for each author.

Highlighted in blue are phrases I see often enough in AI-generated solicitations that for me they are an AI tell. They often vary slightly:

• For “I recently had the opportunity”: I recently came across / I recently spent some time with / I recently had the opportunity to review / I recently spent some time looking into
• For “I was partcularly impressed by” and “What stood out most was”: What particularly caught my attention / What particularly interested me / I was particularly drawn to / What struck me immediately

As for “Warm regards,” others have commented on how often this is a closer in AI-generated emails.

The text highlighted in yellow is just stupid mistakes the scammer somehow included or neglected to correct. I’m sure Rakia Clark, who really is a HarperCollins editor, loves her job, but not so much that she has melded the publisher’s name with her own.

Email 2: You Weren’t Meant to See That

In this one, the scammer sent the actual AI output.

You can see how potential marks are classified (“Warm / Cooperatively Engaged”), and also how psychological manipulation comes into play (the “strategic” discussion of why not to include a payment link yet, which I’ve highlighted in red). Also, this isn’t the first contact: the scam has reached Phase 3, which means that several emails have preceded this one. One of the unique characteristics of this wave of gen AI scams is their boiled-frog approach, with often very lengthy email exchanges that ease the mark gradually toward the endpoint of money handover. Writers who share these solicitations with me are often confused because the initial contacts don’t mention fees: that’s on purpose.

I’ve included the first few paragraphs of the AI-proposed reply in the screenshot so you can see the strategy in action (and also what passes for a “highly professional” tone).

Strip Away the Tech, and It’s the Same Old Scam

The use of gen AI in this current wave of scams enables scammers to create seductive pitches replete with authentic details and head-turning flattery. It also makes possible a volume of solicitation that honestly is like nothing I’ve ever seen (and the Philippine publishing scams that also rely heavily on email solicitation are no slouch in the volume department). New technology aside, though, the scam boils down to what all writing scams boil down to. Pick a mark. Pitch a good or a service or the promise of a desired outcome, using lies and appeals to vanity. Take money. The end.

With the Philippine and South Asian scams, we have a decent idea how they work, thanks to the Philippine scammers’ fondness for Facebook and the South Asian scammers’ traceable webs of related businesses and shell companies, as well as lawsuits that have revealed details of both kinds of operations. We still don’t know how the AI scams are structured or who runs them. The payment and other data I’ve collected suggests that they mostly hail from West Africa, and there’s obviously some degree of centralization–the way the solicitations sort into distinct “types” or categories suggests this (i.e., marketing, book club, impersonations of well-known authors, editors, agents, etc.). But are there relatively few overlords providing leads to battalions of underlings? Or is the setup more decentralized, with individuals operating on their own, kind of like multiple sellers of a single line of products? I wish I knew.

The one bright spot amid this scam tsunami is that familiarity breeds suspicion. The aggressive volume of the solicitations, along with the recognizable categories, undercuts their effectiveness. By the tenth gushing book club invitation, or the sixth cordial expression of interest from [editor X] of [publisher Y], most writers will have become pretty skeptical. Although I’m getting more reports of these scams than of any others ever, I’m also hearing from fewer writers who’ve lost money to them.

The scammers are still making money, though. The emails wouldn’t be flooding our inboxes if they weren’t. Be careful out there.

Tags: Nigerian AI scams, Solicitation

The post How Those Personalized Generative AI Scam Emails Are Built appeared first on Writer Beware.