Google CEO Sundar Pichai.
Beck Diefenbach/Reuters
A recent Google Search update is causing anger across the web. Some site owners say they are being pushed down in results with huge drops in traffic.Google recently tweaked its guidelines to permit AI-generated content.
Morgan Overholt saw half of her website’s traffic disappear overnight.
“This is a joke,” another person wrote on a forum for website owners. “I’ve got long-form content, well written, well researched, filled with original image content LOSING to 500 word AI generated crap…Google is apparently forcing publishers to generate AI spam or die.”
Those are just two voices in a chorus complaining that Google’s latest “helpful content” update – which is meant to improve the quality of search results – has sent their websites tumbling down the rankings.
The new update, which has been rolling out this month, promised several positive tweaks to punish pages that are built only to rank well in search engines.
But it also came with a big red flag that currently has a lot of website owners worried: human authorship is no longer necessary.
As spotted by SEO Roundtable’s Barry Schwartz, Google recently tweaked the documentation explaining how it defines “helpful content.” No longer is it “helpful content written by people, for people,” but simply, “Helpful content written for people.”
The timing of that change, which coincided with the latest Search algorithm update, has made many website owners nervous about what’s to come.
One travel blogger said she watched 80% of her traffic disappear in the space of 48 hours, with posts “very obviously AI written” now outranking her.
“Yesterday we saw 55% drop from normal traffic, today it’s even worse,” Overholt, who runs several travel-related websites, told Insider. She watched pages that previously ranked first for certain terms tumble to fifth, sixth, or seventh place in Google search results. Web pages that are three or four years old now rank above her, she added. (Overholt has written freelance articles for Insider in the past).
Fluctuations are common when Google pushes out new updates, and content creators often complain when important things like this change. But some website owners claim they’ve never seen a shift like this before.
AI content and ‘rehashing’
“The helpful content ranking system was designed to show more content in search results that’s created to help or inform people, and less content created solely to rank better on Google Search,” a company spokesperson told Insider. “We are not targeting content produced by any particular method – AI or otherwise – we’re concerned with the quality of a given webpage and its helpfulness to readers.”
One major goal seems to be to demote sites that reuse and rejig information that’s already online for the purpose of ranking higher in Search results.
The problem is that large language models are trained on masses of content that’s been scraped from the public internet. So, in essence, these AI models are cleverly rehashing what’s already been published. That should be something Google punishes. Although, the company is also developing its own AI models that do similar things, so it’s conflicted here.
Google Search advocate John Mueller spent the day trying to calm some website owners down, while explaining the goals of the company’s recent update.
He debated with one website owner back and forth over their website, which focuses on the net worth of celebrities. Schwartz reported the details at Search Engine Roundtable on Wednesday.
Mueller’s response shows what a tricky situation Google finds itself in. One the one hand, the company is saying AI content is OK, but on the other, Mueller is criticizing a site owner for using ChatGPT.
“I think you should focus on unique, compelling, high-quality content that adds to the web,” Mueller wrote on X. “As you have it now, it looks like a compilation of ChatGPT output on topics that tons of sites have already covered.”
In August, Mueller addressed similar issues during a Reddit chat. One comment stood out as an inadvertent condemnation of Google’s own AI model efforts, which include pitching newsrooms on having its machines write stories automatically.
“By definition (I’m simplifying), if you’re using AI to write your content, it’s going to be rehashed from other sites,” Mueller said.
Spotting AI content is impossible
It’s unclear how much of the content that has been bumped up is AI-generated, or whether this is just other content written differently by humans. That underscores a problem for both users and Google: It’s currently impossible to spot the difference between machine-generated content and human work.
Google continues to say that content built entirely for the purpose or ranking high in search results will be punished. But the company no longer says it’s blocking AI content from getting through, as long as it appears to pass quality tests run by Google’s own bots.
It’s a battle that the company will be fighting on multiple fronts. Generative AI is making content creation much easier and faster. That is likely to trigger a flood of machine-generated content on the web.
Picking out which AI-powered blog or image is better than another one, amongst thousands on the same subject — all automatically published within a few hours — will be a herculean task, even for a tech company as savvy as Google.
An AI ‘tank man’
The tech publication 404 reported on Wednesday that a search for the Tiananmen Square “tank man” was recently returning an AI-generated version of the iconic image. Insider was unable to replicate it, but several other people corroborated 404’s findings, suggesting that Google made some changes after 404 published its story.
It’s not the first time Google has served up an AI version of a well-known piece of digital content. Last May, Futurism reported that the top result for “Johannes Vermeer” was an AI-generated version of The Girl With the Pearl Earring.
When it comes to images and video, there may be ways to tackle the problem: Google and others are building ways to watermark images and video that are AI-generated, which could help search engines detect them.
But the latest changes have many people nervous that the AI content flood has begun and Google is letting it flow with few guardrails.
Are you a Google employee? Got a tip? Contact reporter Hugh Langley using encrypted messaging apps Signal or Telegram at +1 (628) 228-1836. Check out Insider’s source guide for suggestions on how to share information securely.