Rewriting the Résumé
A new working paper from the National Bureau of Economic Research paints a picture of a workforce quietly updating its own history. Analyzing monthly snapshots of 29.4 million US LinkedIn profiles, the paper found that nearly a fifth of accounts had retroactively changed the title or description of a job they had already left.
The practice has a distinct pattern: what gets added is AI-related. Since ChatGPT's release in late 2022, retroactive additions of terms including 'AI', 'GPT', 'LLM', and 'artificial intelligence' have risen more than sixfold. What gets deleted? Remote-work and diversity, equity, and inclusion (DEI) language.
"Workers modify their résumé to reflect what they believe employers want to hear," the authors wrote.
Behind the Stats
The paper, authored by Stanford economist Nicholas Bloom and researchers at Stanford and Revelio Labs, used data collected from LinkedIn profiles between August 2020 and January 2026. The practice of retroactively editing past roles is common enough that the researchers coined a term for it: 'time travel'. Those who do it are 'time travelers'. The median edit landed more than four years after the job had ended.
The trend isn't evenly spread across industries. In technology and information, 31.6% of workers had edited an old role, far above the study-wide rate of 19.7%. Arts and entertainment came in at 25%, and professional and scientific services at 24.1%.
The researchers are careful to note that adding AI language to past jobs is not necessarily exaggeration. People may be adding detail they had omitted, or they might be describing old work in current vocabulary. "The job is the same, but the way they talk about it has changed," they wrote.
Switching Terms: Remote Work and DEI
The study also tracked the shifting language around remote work and diversity. By the end of 2025, workers were adding remote-work terms like 'remote', 'work from home', and 'WFH' at almost the same rate as they were removing them, a shift that coincided with a wave of return-to-office mandates.
Similarly, additions of terms like 'diversity', 'equity', 'inclusion', and 'DEI' saw a sharp drop in early 2025, a change that happened alongside President Trump's executive orders in January 2025 targeting federal DEI programs. These edits appear to be a message about what workers believe employers want to see.
A Distant Measurement Problem
The paper's conclusions carry implications beyond individual job hunters. The authors estimate that a 2026 snapshot of LinkedIn would overstate how common AI-related skills were in 2022 by about 30%. That is a significant margin of error for labor-market analysis that relies on LinkedIn data.
In Europe, where AI systems used in recruitment and employment decisions moved into the AI Act's high-risk tier in August, this data creates a tricky question. The law's Article 10 requires training data that is relevant, representative, free of errors, and complete. With profiles rewritten in image of current demand, the historical accuracy becomes harder to guarantee.
Perspective: Workers and Employers
Workers see themselves as adapting to a changing job market. As the paper suggests, they revise their resumes to align with what they think employers want, which has led to a boom in AI skill claims and a drop in remote-work or DEI emphasis.
Employers and researchers face a challenge in measuring public trends. If profiles are rewritten with hindsight, they no longer reflect the actual skills that existed at a point in time, distorting both labor-market data and any systematic sorting that uses those profiles, a concern amplified by new regulatory requirements in the EU.