Google's new AI-generated search results feature is suffering from the same problem that its regular results have had of late: Spammy, if not outright malicious links are rising to the top of the SERP stack.

Google reportedly began testing its Search Generative Experience (SGE), previously an opt-in only test, on a small subset of US users at the end of last week without them needing to opt in. The quality and safety of the results being displayed for users, however, is debatable based on findings by Lily Ray, VP of SEO strategy and research at marketing firm Amsive. 

Ray posted to X over the weekend to talk about an ongoing problem with Google search results she first discovered back in December of last year. Those results are being gamed by spammers and scammers, leaving the top results for long-tail searches flooded with junk. 

Now, Ray said, the same garbage she's seen on regular Google results is showing up on SGE results, which appear at the top of SERPs, as shown in an image Ray provided to The Register. 

"The keywords are pretty random but are generally long-tail keywords with relatively low volume and/or limited SEO competition," Ray told us, noting that the spam results often show up on branded search results, like pages for Craigslist, particular locations and other specific results.

According to a Search Engine Journal story that cites Ray's findings from late last year, the SEO spam she spotted (outside of SGE) involves manipulating long tail, local results served by new domains with no reputation score to boost themselves in ranking. Now those same malicious links, found via the same queries, are showing up in SGE results.

"This time… I was able to use SGE (via Google Labs) to generate a response, which pulled in content from these spammy cloaked domains," Ray told us. 

The fact the results came via Google Labs SGE results is crucial to note, Ray said. 

"The results I saw were generated using the opt-in SGE feature via Google Labs, which is not the same as the new public test Google announced last week," Ray told us. "Google has indicated that there will be much more limited results shown for the users in the public experiment, and they won't be able to force generate the responses the way users currently can in Google Labs." 

The Register was able to generate malicious SGE results using some of the keywords highlighted by Ray, but we had to opt in via the Google Labs page as well.

"We continue to update our advanced spam-fighting systems to keep spam out of Search, and we utilize these anti-spam protections to safeguard SGE. We've taken action under our policies to remove the examples shared, which were showing up for uncommon queries," a Google spokesperson told us.

Around 99 percent of Google search results are spam free, we're told, but Google noted spammers are constantly evolving their method to avoid filters, leaving the company playing catch up. Google also reiterated that results from Google Labs aren't indicative of what test users are seeing.

We're unable to confirm what public SGE users are seeing, however, and Google provided no evidence that those users are safe beyond its claims they are.

Until we can get some sort of confirmation that SGE is safe for the general public, which has been unwittingly subjected to the latest questionable AI experiment from Google, it might be a good idea to at least closely examine URLs before opening. 

As always in the age of an ever-worsening internet, caveat usor. ®

https://www.theregister.com//2024/03/25/googles_aipowered_search_results_are/