One day in November, a product strategist who will be called Michelle (not her real name), logged into her LinkedIn account and changed her gender to male. He also changed his name to Michael, he told Techcrunch.
They participated in an experiment called #wearthepants where Women are tested The hypothesis that Linkin’s new algorithm is considered with women.
For several months, Some Linkin Heavy Users complain about seeing a drop in engagement and impressions on career-oriented social networks. This is after the company’s vice president of preparation, Tim Jurka, said in August that the platform it’s “newer” that llms has done to help surface content that’s useful to users.
Michelle (whose identity is known to TechCrunch) is suspicious of the change because she has more than 10,000 followers and ghantuwrites posts for her husband, who only has about 2,000. But she and her husband tend to hang around the same post number, she said, albeit a larger one.
“The more significant variable is gender,” he said.
Marilynn Joyner, the founder, also changed the gender of her picture. She has been posting on LinkedIn continuously for two years and noticed a few months ago that her posts’ visibility declined. “I changed the gender on my profile from female to male, and my impressions jumped 238% in one day,” she told techcrunch.
Megan Cornish reported similar results, as did Rosie Taylor, Jessica Doyle Mekkes, Abby Nydam, Felicity Menzies, and so onSee rank-.
TechCrunch events
San Francisco
I’m fat
October 13-15, 2026
LinkedIn said so The system “Algorithms and AI do not use demographic information such as age, race, or breed as a marker to determine your own content feed that cannot be resolved, does not equal treatment or automatically biases” in the feed.
Social algorithmists agree that explicit sexism may not cause it, despite the possible bias.
The platform is “a complex symphony of algorithms that pull certain mathematical and social levers, and continuously,” Brandeis Marshalldata ethics consultancy, told TechCrunch.
“Changing one’s profile photo is one such lever,” he said, adding that the algorithm is also influenced by, for example, users and current users with other content.
“What they don’t understand is that all the other levers that make this algorithm prioritize it are prioritizing other people’s content. It’s a more complicated problem than people think,” said Marshall.
Bro-coded
The #Wearthepants The experiment started with two entrepreneurs – Cindy Gallop and Jane Evans.
She asked two men to create and post the same content as her, so that they understand that gender is the reason for many women’s feelings. Gallop and Evans both have the biggest following – more than 150,000 combined compared to the two of them with around 9,400.
Gallop reported that the post office was only 801 people, as for the person who was sent Similar content to 10,408 people, more than 100% of their followers. Other women then joined. Some, like Byyarner, who uses LinkedIn to market her business, are concerned.
“I’m very happy to see LinkedIn take responsibility for possible biases in their algorithms,” Joy said.
But Linkedin, like the relevant media platform and other LLM media platforms, offers scant details about the models it trains.
Marshall said that most of the platforms “have been embedded in a white, male, western perspective” because those who train the models. The researchers found Evidence of human biases like sexism and racism in popular llm models because the models are trained on human-generated content, and Humans often engage in training immediately after training or learn to learn.
Still, how individual companies implement these AI systems is shrouded in algorithmic black-box secrecy.
LinkedIn says the #wearthepants experiment cannot show gender bias against women. Jurka’s August statement said – and LinkedIn’s head of Milk AI and Government AI and Change, Sakshi, reaffirmed in another post In November – if the system does not use demographic information as a signal for visibility.
However, Linkedin told TechCrunch that it tests millions of posts to connect users for opportunities. It says demographic data is only used for such tests, like seeing if the posts “from different shows are similar to the creators,” said the audience, “said the audience,” said the audience, “said the company that told Techcrunch.
LinkedIn has been noted for research and adjust that algorithm to try to provide less bias Experience for users.
It’s an unknown variable, Marshall said, that may explain why some women have improved impressions after changing their gender to male. PartAKK in the Virus trend, for example, can generate a lot of engagement; Some accounts post for the first time, and the algorithm may reward them for doing so.
Tone and writing style may also play a part. Michelle, for example, said the week she posted as “Michael,” she adjusted the tone a little, writing in a direct style, as she did for her husband. That’s when he said impression jumped 200% and eragement went up 27%.
He concludes the system is not “explicitly sexist,” but seems to assume the communication style commonly associated with women is “a proxy for inferiority.”
Stereotypical male The writing style is believed to be more concisewhereas Stereotypes Writing style for women imagined to be softer and more emotional. If LLMs are trained to promote writing that conforms to male stereotypes, that is a subtle, implicit BIAS. And As previously reportedResearchers have determined that most LLMS are stacked with them.
Sarah Dean, an assistant professor of computer science at Cornell, said that platforms like LinkedIn often use their entire profile, in addition to emphasizing content to promote it. That includes projects on user profiles and the types of content they typically engage with.
“People’s demographics can affect the algorithm ‘both ways – what they see and who sees what they post,'” says Dean.
LinkedIn told Techcrunch that its AI system looks at hundreds of signals to determine what attracts users, including insights from people’s profiles, networks, and activities.
“We’re running ongoing tests to understand what helps people find the most relevant, timely content for their careers,” a spokesperson said. “Member behavior also creates feeds, what people click on, save, and participate in changes every day, and what formats they like or dislike. What behavior is shown in real time from us.”
Chad Johnson, an active sales expert at LinkedIn, explained Change is the abuse of likes, comments, and reposts. The LLM system “no longer cares how you deliver or what time of day it is,” Johnson wrote in a post. “Really what your writing shows is understanding, clarity, and value.”
All this makes it difficult to determine the true cause of the #wearthepants result.
People just don’t like algos
However, it seems that many people, among the flags, do not like or do not understand LinkedIn’s new algorithm – whatever.
Shailvi Wakulu, a data scientist, told Techcrunch that he paid at least one a day for five years and used it to see thousands of impressions. Now he and his wife are lucky to see a few hundred. “It’s demotivating to create content with a loyal following,” he said.
One person told Techcrunch they’ve seen about a 50% drop in bookings over the past few months. Still, other people say that they seem to post impressions and reach more than 100% in the same time frame. “This is largely due to writing specific topics for a specific audience, which is what the new algorithm is,” he told TechMe, adding that clients are seeing similar improvements.
But in Marshall’s experience, he, who is black, believed posts about his experience did worse than posts related to race. “If black women only interact when talking about black women but not when talking about their specific expertise, then that’s biased,” she said.
The researcher, Dean, believes the algorithm will likely amplify “whatever signal is already there.” You can reward them for giving certain leads, not because of the author’s demographics, but because of the history of responding to them on the platform. While Marshall may have stumbled into other areas of bias, his anecdotal evidence is insufficient to determine that with certainty.
LinkedIn offers some insight into what they are doing today. The company says that its user base has grown, and as a result, shipments are up 15% year-on-year while they are up 24% yoy. “This means more competition in the feed,” the company said. Posts about professional insights and career lessons, industry news and analysis, and educational or informative content on work, business, and the economy are all done, so say.
If anything, people are just confused. “I want transparency,” Michelle said.
However, as the algorithm that chooses the content is always kept secret by the company, and transparency can affect the game, it is a big Ask. This is one of those things that you can’t get enough of.