Nearly 1 in 5 US LinkedIn users are retroactively editing past jobs to add AI

Nearly a fifth of US LinkedIn users have edited the job titles or descriptions of roles they already left, and the most common addition since 2022 is "AI." A new study from Stanford economist Nicholas Bloom and the labor-data firm Revelio Labs calls the practice "time travel," and describes the résumé as a real-time thermometer for what workers think employers want to hear.

The researchers coined "time travelers" for workers who rewrite jobs they have already held, and "time travel" for the edits themselves. The dataset covers 29.4 million US LinkedIn profiles, with snapshots from August 2020 through January 2026. About 19.7% of those profiles retroactively changed a past role, and the median edit came more than four years after the job had ended, so the behavior is not just fresh graduates polishing recent titles. The paper is a working paper from the National Bureau of Economic Research, not yet peer-reviewed.

Additions of "AI," "GPT," "LLM," and "artificial intelligence" to previous roles rose more than sixfold after ChatGPT's release in late 2022, and the additions cluster in the same sectors that have hired most aggressively for AI work. The pattern was most concentrated in technology and information, where 31.6% of profiles showed at least one retroactive edit, followed by arts, entertainment, and recreation (25%) and professional, scientific, and technical services (24.1%). In the paper, "expertise in artificial intelligence becomes increasingly common to retroactively add to résumés," because the keyword is what recruiters and applicant-tracking systems are scanning for.

Two other trend lines in the data show the reverse pattern. By the end of 2025, US workers were adding remote-work terms ("remote," "work from home," "WFH") to past roles about as often as they were deleting them, with net additions effectively flat. The shift tracks the wave of return-to-office mandates that began in 2023 and accelerated through 2024 and 2025. The paper's language: "Workers modify their résumé to reflect what they believe employers want to hear."

The third trend line tracks DEI. Additions of "diversity," "equity," "inclusion," and "DEI" to past roles had been climbing through 2024, then collapsed at the start of 2025, coinciding with President Trump's January 2025 executive orders targeting federal DEI programs. The paper describes the co-location and does not assert causation. The same period saw private-sector employers pare back DEI language under legal and political pressure.

The constructive read is that workers are reading the room, not lying. The edits often happen when someone is preparing to move jobs, which is when the perceived hiring bar is highest. The paper is careful to flag that retroactive edits are a worker-side signal of perceived demand, not a measure of actual AI skill, hiring, or productivity. The authors also note that future revisions will track whether the AI-addition rate cools once the labor market does, and that the same dataset can be used to study "time travel" across other categories.

A few caveats limit what the data can show. The dataset is self-reported LinkedIn profiles, weighted toward white-collar workers and toward industries that use LinkedIn heavily, and it captures edits users chose to publish. A single profile can carry multiple "time travel" edits across different roles, and the median edit landed more than four years after the job ended, which means the signals are slow-moving, not real-time.

The next test is whether the AI additions hold up once the post-ChatGPT hiring wave slows, or whether they follow the same pattern as remote work and DEI and start to come off the page.