Revisionist Employment History
There is a familiar signal that someone may be looking for a new job: the sudden professional headshot on LinkedIn. A new profile picture is rarely the only change. Workers also revise their summaries, tweak their job titles and rewrite descriptions of previous roles. Sometimes they return to jobs they left years ago and change how those experiences appear on their profiles.
In new research with Nick Bloom and Gideon Moore at Stanford, we use monthly vintages of Revelio Labs data to observe these changes over time. We focus on what we call “time travelers”: workers who change the title or description of a job more than 60 days after leaving it.
While someone is still in a position, an updated description might reflect a genuine change in responsibilities. However, once the worker leaves a position, the nature of work is fixed. A later edit changes only how the worker chooses to present that experience in the future when they are making the edits.
This behavior is common. Around one in five U.S. users in our sample retroactively edited at least one previous position between 2020 and 2026. These are not primarily small corrections made shortly after departure: the median edit occurred 4.5 years after the position ended.
A profile refresh often comes before a job switch
The timing of these edits suggests that workers are not revisiting their histories at random. Among workers who change employers, the probability that a worker edits their profile begins to rise about six months before the transition. It climbs sharply in the final few months, peaks close to the employer change, and then falls back toward its earlier level. By contrast, among workers who never change employers, editing remains much lower and comparatively flat over the same time window.
Workers are 66% more likely to change employers during the period in which they update their profiles. The timing suggests that many workers revise their profiles as part of preparing for a job search. A stronger profile may also help them attract recruiters, which can help them land a new job. Either way, retroactive edits are a strong signal that a job transition may be approaching.
A professional profile is therefore not simply an archive of what someone did. It is also a marketing document, adjusted for the opportunity the worker hopes comes next.
Workers rewrite the past for today’s labor market
Because these edits apply to completed jobs, sometimes even from many years ago, analyzing these edits offer an unusual way to observe how workers’ perceptions of employer demand change.
Normally, an increase in a keyword on professional profiles could mean that more workers are doing that kind of work, or simply that they are becoming more likely to describe existing experience using that term. Time-travel edits help distinguish between the two. The job itself is already over, so only its presentation can change.
The clearest example is artificial intelligence. Following ChatGPT’s launch, workers started adding AI-related language to descriptions of jobs they had already left. The increase is visible in both descriptions and job titles, although it is stronger and more volatile in descriptions. At the peak, the rate of adding AI terms was more than five times its earlier baseline.
DEI language follows the opposite trajectory. During the earlier part of the sample, workers generally added DEI references to previous jobs. Around early 2025, however, net additions turned sharply negative, particularly in job titles. The pattern partially rebounds afterward, but remains weaker than before.
Remote-work language shows a similar, though more gradual, reversal. References to working from home were frequently added in the period after the pandemic, but that pattern has steadily weakened, and workers have recently been removing language on remote work.
These changes do not necessarily tell us what employers objectively value. They tell us what workers believe will make them more attractive to employers. That belief matters in its own right: workers respond not only to labor demand, but also to their perception of it.
AI is also changing how profiles are written
Workers are not only changing what they emphasize in their profiles, but increasingly, how they write them as well. After ChatGPT’s launch, words strongly associated with LLM-generated writing—such as “delve,” “underscore,” and “showcase”—rose rapidly in retroactive job-description edits. Before ChatGPT, these markers appeared in roughly 0.3% of edits. By late 2023 and early 2024, their share had climbed above 3%.
The rate has since fallen substantially, settling below 1% by early 2026. Even so, it remains well above its pre-ChatGPT baseline. The pattern suggests a sharp initial wave of AI-assisted rewriting, followed by a partial normalization rather than a full return to earlier writing styles.
AI therefore appears twice in the profile-editing process. Workers increasingly present themselves as having AI skills, and many are also using AI tools to help edit their profiles.
- About 20% of users observed on online professional networking websites retroactively change their profiles. Workers often update descriptions of previous jobs while preparing to change jobs. Retroactive editing begins to rise about six months before a job transition, peaks around the move, and declines afterward.
- These retroactive edits reveal how workers think employers’ demand is changing. Workers started adding AI language to old jobs since the release of ChatGPT, while removing references to DEI since Trump’s executive order ending federal DEI programs.
- Workers who typically struggle more with writing are more likely to use LLMs for profile edits. LLM usage for edits decreases with the education level of workers - for example, workers without a college degree use LLMs more than those with a Bachelor’s degree. Within MBAs, graduates from lower ranked schools are more likely to use LLMs for writing than graduates from top ranked business schools.