LinkedIn Profiles: Retroactive AI Could Inflate 2022 by 30%

An analysis of 29.4 million LinkedIn profiles in the United States describes a vast retrospective rewriting of experiences, with AI being added long after the fact. The result: a snapshot from 2026 would overestimate the presence of AI skills in 2022 by about 30%. This practice, particularly widespread in tech, is accompanied by a decline in terms related to remote work and DEI.

A 2026 Snapshot Would Overestimate AI by 30% in 2022

It is estimated that a survey conducted in 2026 would overestimate the prevalence of AI-related skills in 2022 by about 30%. They warn more broadly that current LinkedIn profiles are not a fixed record of the past. The median modification occurs more than four years after the end of employment, indicating that histories evolve over time. As the researchers summarize, “the work is the same, but the way they talk about it has changed.”

What Profiles Rewrite: Rise of AI, Decline of Remote Work and DEI

Since the release of ChatGPT in late 2022, retroactive additions of terms like “AI,” “GPT,” “LLM,” and “artificial intelligence” to past roles have increased by more than six times. Conversely, users are less inclined to add mentions related to remote work; by the end of 2025, these terms were added and removed in equal measure, coinciding with mandates for a return to the office. Additions of “diversity,” “equity,” “inclusion,” and “DEI” saw a significant decline in early 2025, coinciding with executive orders from Donald Trump in January targeting these programs. The researchers refer to these revisions as a “time travel,” noting that “expertise in artificial intelligence” is increasingly appearing in these retroactive additions. They also point out that these modifications frequently occur when individuals are preparing to change jobs, writing that workers tailor their resumes to what they believe employers expect.

Who Modifies the Most: 31.6% in Tech, 19.7% Overall

In total, 19.7% of the profiles studied displayed at least one retroactive modification of a title or job description. The practice peaks in technology and information (31.6%), followed by arts, entertainment, and recreation (25%), and then professional, scientific, and technical services (24.1%). The researchers base their findings on 29.4 million American profiles, analyzed through monthly snapshots in a working paper from the NBER. The data was collected by Revelio Labs between August 2020 and January 2026.

Method, Authors, and Cautions for Interpretation

The work is authored by Stanford economist Nicholas Bloom along with researchers from Stanford and Revelio Labs, and it has not been peer-reviewed. It is noted that AI additions do not necessarily mean an exaggeration of skills; they may also reflect details initially forgotten or a rephrasing in vocabulary deemed more attractive by candidates. For them, these revisions provide insight into positioning strategies in a labor market influenced by AI, while evolving in response to external events such as return-to-office mandates.