Google is changing how its search engine handles explicit deepfakes, with new ranking adjustments meant to make nonconsensual synthetic images harder to find. The company says the changes are already shifting some searches away from pages claiming to host fake explicit images and toward news articles, warnings, and other nonexplicit results.
What changed in Google Search
A few weeks ago, a search for “deepfake nudes jennifer aniston” showed at least seven prominent results that purported to offer explicit AI-generated images of Jennifer Aniston. Those results have now disappeared, according to the source article.
Google product manager Emma Higham says ranking changes introduced this year have cut exposure to fake explicit images by over 70 percent on searches looking for that material about a specific person. Instead of elevating pages that claim to show nonconsensual fake explicit images, Google’s systems are trying to surface articles and other nonexplicit material.
In the Aniston example, the search results now include items such as “How Taylor Swift's Deepfake AI Porn Represents a Threat” and an Ohio attorney general warning about “deepfake celebrity-endorsement scams” aimed at consumers. That shift matters because it changes what a user is likely to encounter first: material about the social impact and risks of deepfakes, rather than pages presenting the images themselves.
Why explicit deepfakes are a search problem
The issue is not limited to celebrities. Victims’ advocates say the broader availability of AI image generators, including tools with few restrictions, has contributed to an uptick in nonconsensual explicit imagery, or NCEI. The source article describes the tools as making it easy for almost anyone to create spoofed explicit images of a target, from a middle school classmate to a mega-celebrity.
That creates a difficult challenge for a search engine. Google has already made it easier for people to request removal of unwanted explicit content, but victims and advocates have pushed for more proactive measures. Their concern is straightforward: a takedown process can help after harm appears, but search visibility can also amplify that harm while victims are trying to respond.
Google has also tried to avoid taking on too much of a regulator role for the internet or reducing access to legitimate porn. In response to earlier scrutiny, a Google spokesperson said multiple teams were working to strengthen safeguards against NCEI.
The new takedown protections
As part of the crackdown described by Higham, Google is applying three measures used for real but unwanted explicit images to synthetic and unwanted explicit images. These measures connect takedown requests to broader search behavior, so a successful removal can affect more than a single URL.
- After Google approves a takedown request for a sexualized deepfake, it will try to keep duplicate versions out of results.
- It will filter explicit images from search results for queries similar to those named in the takedown request.
- Websites with “a high volume” of successful takedown requests will be demoted in search results.
Those steps are designed to address a common weakness in removal systems. If one page comes down but copies remain easy to find, the burden stays with the person targeted. Duplicate filtering and related-query filtering are intended to reduce that cycle, though Google has acknowledged that its measures do not work perfectly.
What the data shows so far
The source article points to earlier reporting that helps show the scale of the problem. In March, a WIRED analysis found that Google had received more than 13,000 demands to remove links to a dozen of the most popular websites hosting explicit deepfakes. Google removed results in around 82 percent of those cases.
Those figures show both action and limits. Removal can work in many cases, but the number of demands also indicates that victims, advocates, and search platforms are dealing with a large and recurring problem. The new ranking and demotion measures are meant to reduce discoverability, not only respond one link at a time.
The ranking change also follows a WIRED investigation this month. That investigation found that, in recent years, Google management rejected numerous proposals from staff and outside experts aimed at addressing intimate portrayals of people spreading online without consent.
The remaining debate
Former employees and victims’ advocates say Google could go further. One example raised in the source article concerns warnings. Google prominently warns people in the US who search for naked images of children that such content is unlawful. Advocates support that kind of warning as a possible deterrent, although its effectiveness is unclear.
Similar warnings do not appear for searches seeking sexual deepfakes of adults, even though the source article notes that laws exist against sharing NCEI. A Google spokesperson confirmed that this will not change.
For now, Google’s approach is focused on ranking, filtering, takedown follow-through, and site demotion. The practical result is that some explicit deepfake searches may lead users toward reporting, news, consumer warnings, and social context instead of the images they were looking for. That does not eliminate the underlying creation of explicit deepfakes, but it can reduce how easily search helps people find and spread them.