Are you a fan of a good heist story? Do you like rooting for the anti-hero? A clever scheme to pull one final job?
Well, that’s not what we have here.
There’s no Clooney-esque charming rogue. Instead, you’ll find a story about how AI was used to create what critics called “a roadmap on how to destroy the internet.”
More specifically, how one agency founder used generative AI to pull off what he called an “SEO heist that stole 3.6M total traffic from a competitor.”
To better understand it, I spoke with two experts in SEO about how it happened and whether it really is the death rattle of the internet.
An AI-Powered SEO Heist
Last month, a small corner of social media was set on fire when tech entrepreneur Jake Ward posted the results of his self-described “SEO heist.”
If you’re not familiar with SEO, it stands for “search engine optimization.” It’s the way you make your website more attractive to search engines and improve your site’s ranking in the results pages.
For some sites, that means creating the kind of high-quality experience that people want to search for. To others, it means trying to game the system with cheap tactics and crap content.
And there’s no better description than that for what Ward’s agency did.
Using 100% AI-generated content, they were able to “steal” 3.6 million views from a competitor.
The plan was a simple one:
- Download a competitor’s sitemap. (A list of all of their web pages.)
- Turn the entire list into blog titles.
- Use those titles as prompts for an AI content writer.
Following this blueprint, it only took a few hours to crank out 1,800 “new” articles – all without the touch of a human hand.
Was the agency’s SEO heist successful?
For a surprising amount of time, it did work. Over the course of 18 months, Ward’s derivative site garnered 3,600,000 views.
Here’s a look at the “stolen” content’s traffic over two years, taken from Ahrefs, an SEO tool.
That’s the kind of growth that any marketer would salivate over. And salivate they did.
At the time of publishing, a post on Ward’s LinkedIn profile had hundreds of supportive comments.
Meanwhile, on X (formerly Twitter), an equally large number of comments called the heist “dangerous,” “cringe,” and “ethically sh***y.”
But if you’ve worked in SEO or content marketing, you know this shouldn’t have happened.
How did they pull this off?
“I was initially quite surprised this did work – and frustrated, of course!” says Aja Frost, director of Global Growth at HubSpot.
She’s made a career out of doing the exact opposite of this.
Believe it or not, Google has algorithms in place specifically to catch derivative content. That’s why you no longer see the copycat, word-salad websites that were popular in the early 2000s.
So why did this tactic work?
“My hunch is that [Ward’s] AI-spun articles were different enough at the semantic level to pass Google’s duplicate content systems,” says Frost.
Ward’s AI tool didn’t rewrite the competitor’s content; it simply used it as a template for new content.
Head of Content SEO at HubSpot, Rory Hope, adds that the overwhelming volume of content also played a part in this success.
“The site involved in the ‘heist’ was able to quickly publish content that sent signals [...] that initially appeared to Google as unique content, with depth that created broad topic coverage at lightning speed.”
Frost points to the final piece of the puzzle: the type of content.
“Ward’s company used AI to rewrite content from Exceljet, a website helping people use Excel with articles like ‘XLOOKUP vs VLOOKUP’ and ‘How to concatenate in Excel.’ This type of content is fairly commodified.”
In other words, it’s easy for AI to generate “how-to” articles because the content is straightforward with clear answers. There’s not much need for tone, interpretation, or perspective.
It would be harder for an AI tool to duplicate content that required personality or first-hand experience. So Ward’s tactic likely wouldn’t have worked on, say, an opinion blog or a news site with original reporting.
“Whether it was written by a human or not [...] it’s probably very similar to what AI would generate,” Frost says.
To be clear, when Frost says the content is similar, that doesn’t mean it's equally valuable.
In fact, there’s a crucial difference.
All of Exceljet’s content is tested and verified by an expert in Excel. (That kind of effort is worth some free advertisement.) On the other hand, Ward’s content was published without any oversight from a human expert.
That’s a big deal when you remember that generative AI models are prone to making up fake answers.
However, this similarity does mean that Exceljet’s straightforward content made it an ideal victim.
A Victimless Crime?
The words “stolen” and “heist” came from Ward’s own description, most likely calculated to generate controversy and attention. (And to his credit: Mission accomplished.)
What his agency did, however, doesn’t actually rise to the level of a crime. But that doesn’t mean there aren’t victims, though.
Those victims are Dave and Lisa Bruns, the husband and wife duo who run Exceljet.
Using their own career expertise, the Bruns built a library of content over the course of eleven years – only to have it scraped and replicated in a matter of hours.
And “scraped” is a better description than “stolen.” The AI tool didn’t actually plagiarize the Bruns’ content. Instead, it created new content from the topics scraped off of Exceljet.
So, if there was no direct plagiarism, what’s the big deal?
In a reply to Ward’s bragging on LinkedIn, Dave Bruns cuts to the heart of the matter.
“Who will answer for the AI-generated content, not reviewed or tested by anyone who cares? This kind of project is really the worst of the web. It will seriously degrade the quality of the content we humans consume.”
Did this agency create a black hat SEO roadmap?
“I think it’s a fair worry,” says Rory Hope.
AI makes content creation fast, cheap, and easy – regardless of quality. So, if that content performs just as well on the search engines, it could quickly crowd out human-authored content entirely.
“Google must be worried about this, and we’re seeing it in the three core algorithm updates this year,” Hope adds.
He’s talking about several major changes Google made to its search algorithms in 2023. These algorithms are responsible for deciding what shows up when you make a search.
The recent updates to them aim to make sure you find content that’s actually helpful to humans instead of stuff that’s written to appeal to search engines.
But those changes aren’t about eliminating AI-generated content.
“Google is in between a rock and a hard place,” points out Aja Frost. “Consumers and businesses are clearly incredibly interested in using AI to improve productivity, and Google itself is adding a ton of AI-generated content to its [search results] in the form of the AI-powered snapshot.”
In fact, they’re already rolling out the AI-enhanced version of their own search results.“But at the same time, everyone recognizes that AI trained on AI leads to a quality race to the bottom.”
New AI is trained on content from the internet, so if the internet is solely populated by AI, it becomes a photocopy of a photocopy. Each time, the flaws are magnified.
The Future of the Internet
Despite that possibility, both Frost and Hope strike an optimistic tone. They’ve seen the rise and fall of hundreds of ways people have tried to trick Google.
“I think as long as there’s a system to game – and significant financial incentives for gaming it – people will use manipulative and inherently degrading strategies,” says Frost.
She continues, “In the early days of the internet, that was keyword stuffing and irrelevant link-building. These days, it’s AI-generated content produced without a ‘human in the loop.’”
However, she points out that Google has a financial incentive of its own.
“If Google’s results become saturated with ‘AI-spun garbage,’ consumers won’t use Google anymore. I think that, rather than prohibiting AI for content generation entirely, Google will consistently increase the importance of human ‘signals’ in every algorithm update.”
Hope echoes that sentiment.
“Google wants to better understand the EEAT of content in its battle with low-quality AI spam.”
EEAT refers to the criteria that Google uses to determine the credibility of a website: Expertise, Experience, Authority, and Trustworthiness.
“EEAT is only going to become more important, to the point where it’s much harder for purely AI-generated content to rank,” Frost agrees.
“Not to mention Google updated its Search Quality Evaluator Guidelines to place a greater emphasis on factually correct content,” adds Hope. “I’m optimistic that in the end, Google will figure it out, and it will force SEOs to evolve into better content marketers.”
This follows the normal pattern between Google and bad actors, according to Frost.
“Google and other algorithms will keep evolving their systems to punish bad actors, and bad actors will keep coming up with new ways to temporarily outmaneuver those systems,” she says. “So no, I don’t think AI is the downfall of the web.”
Black Hat Techniques Work, But Not Forever
Google seems to have finally caught on to Ward’s shenanigans.
“Since Ward’s X post went viral, his site’s organic traffic has decreased by 42%,” says Frost.
Here’s an updated look at the “stolen” traffic at the time of writing.
So what do Frost and Hope have to say to those who might be tempted to try Ward’s tactic for themselves?
“Publicly crowing about manipulating Google’s results is typically a one-way road to penalization,” says Frost. “[But] I don’t believe in using this technique under the radar, either.”
Penalty aside, she points out that the value of the content is questionable.
“I imagine its conversion rate was extremely low. Humans usually convert on content they trust, and generic, personality-less, ‘AI by AI’ content is inherently untrustworthy.”
Hope takes a more forceful tone.
“I strongly advise no one to adopt this approach of scraping URLs and using AI to mass produce content,” he says. “It will be unhelpful, and your entire site will eventually be hit, either with a site-wide performance dampener or a manual penalty.”
But that doesn’t mean you have to avoid AI altogether.
“AI gives humans more time for all the work that can’t and shouldn’t be generated,” says Frost. “From interviewing experts, compiling first-party data, crafting original insights, and more.”
“Google has said it’s okay to use AI when creating content,” adds Hope. “However, all of your content needs to be helpful, reliable, people-first content, per their guidelines. If you use generative AI to ‘spin’ up content at scale, it will not be helpful, reliable, nor people-first.”