An investigation by search engine optimisation skilled James Brockbank reveals that ChatGPT could also be recommending companies primarily based on content material from hacked web sites and expired domains.
The findings aren’t a complete examine however the results of private testing and observations. Brockbank, who serves as Managing Director at Digitaloft, says his report emerged from exploring how manufacturers achieve visibility in ChatGPT’s responses.
His evaluation means that some actors are efficiently gaming the system by publishing content material on compromised or repurposed domains that retain excessive authority alerts.
This content material, regardless of being irrelevant or misleading, can floor in ChatGPT-generated enterprise suggestions.
“I consider that the extra we perceive about why sure citations get surfaced, even when these are spammy and manipulative, the higher we perceive how these new platforms work.”
How Manipulated Content material Seems In ChatGPT Responses
Brockbank recognized two major techniques that seem to affect ChatGPT’s enterprise suggestions:
1. Hacked Web sites
In a number of examples, ChatGPT surfaced playing suggestions that traced again to respectable web sites that had been compromised.
One case concerned a California-based home violence legal professional whose web site was discovered internet hosting a listicle about on-line slots.
Different examples included a United Nations youth coalition web site and a U.S. summer time camp web site. They have been each seemingly hijacked to host gambling-related content material, together with pages utilizing white textual content on a white background to evade detection.
2. Expired Domains
The second tactic entails buying expired domains with robust backlink profiles and rebuilding them to advertise unrelated content material.
In a single case, Brockbank found a web site with over 9,000 referring domains from sources like BBC, CNN, and Bloomberg. The area, as soon as owned by a UK arts charity, had been repurposed to advertise playing.
Brockbank defined:
“There’s no query that it’s the positioning’s authority that’s inflicting it for use as a supply. The problem is that the area modified arms and the positioning completely switched up.”
He additionally discovered domains that beforehand belonged to charities and retailers now getting used to publish on line casino suggestions.
Why This Content material Is Surfacing
Brockbank means that ChatGPT favors domains with perceived authority and up to date publication dates.
Moreover, he finds ChatGPT’s advice system might not sufficiently consider whether or not the content material aligns with the unique web site’s function.
Brockbank noticed:
“ChatGPT prefers current sources, and the truth that these listicles aren’t topically related to what the area is (or ought to be) about doesn’t appear to matter.”
Brockbank acknowledges that being featured in genuine “better of” listicles or media placements might help companies achieve visibility in AI-generated outcomes.
Nevertheless, leveraging hacked or expired domains to govern supply credibility crosses an moral line.
Brockbank writes:
“Injecting your model or content material right into a hacked web site or rebuilding an expired area solely to idiot a language mannequin into citing it? That’s manipulation, and it undermines the credibility of the platform.”
What This Means
Whereas Brockbank’s findings are primarily based on particular person testing reasonably than a proper examine, they floor an actual concern: ChatGPT could also be citing manipulated sources with out absolutely understanding their origins or context.
The takeaway isn’t nearly danger, it’s additionally about accountability. As platforms like ChatGPT turn into extra influential in how customers uncover companies, constructing respectable authority by way of reliable content material and earned media will matter greater than ever.
On the identical time, the investigation highlights an pressing want for corporations to enhance how these methods detect and filter misleading content material. Till that occurs, each customers and companies ought to method AI-generated suggestions with a dose of skepticism.
Brockbank concluded:
“We’re not but on the stage the place we will belief ChatGPT suggestions with out contemplating the place it’s sourced these from.”
For extra insights, see the authentic report at Digitaloft.
Featured Picture: Mijansk786/Shutterstock
