Why editing work stays with people
Ask will AI replace editors and the answer sits in the task mix, not in the job title. Editors read copy to catch errors and shape it to a style guide. They also decide what runs, what gets cut, and what the publication will stand behind in public. The first kind of work is pattern matching. The second is accountability, and it does not transfer to software.
Two tasks show the split clearly. Checking spelling, grammar and house style against a set of rules is repetitive, and language models are good at it. Conferring with writers about how a piece should develop is different work. It involves reading a person, judging what a draft is really trying to say, and deciding whether the idea is worth the space at all.
Fact verification sits in between. A model can flag a date or a figure that looks wrong. An editor still has to call the source, weigh how much to trust it, and carry the consequence if it is published and wrong. That is why coverage for this job lands at 44 out of 100 rather than near the top. You can see how that figure is built on the coverage scoring page.
What AI does, what it assists, and what editors keep
Software already handles part of the day without much supervision. Error detection in copy and routine conformance to a style guide are the clearest cases, and they are the tasks most editors describe as the mechanical layer. Across this job’s task list, the share AI can take on its own is 33%.
A larger slab of the work is assisted rather than taken. Rewriting for length, drafting headline and subheading options, and checking names, dates and numbers all go faster with a model in the loop, but an editor signs them off. That assisted share comes out at 43% of task time.
Then there is the work that still needs a person in the room: approving what goes to publication, negotiating changes with a writer who disagrees, and taking responsibility for legal and ethical risk in a piece. That group accounts for 24% of task time, and it is the part of the job that sets the headline figure of 62 out of 100 (higher is safer).
No robots are involved. This is screen work, so there is no hardware cost or deployment delay holding automation back. The tooling cost is software only, which is why adoption in editing has moved faster than in jobs that need machines on site.
What the evidence actually shows
There is no direct, published test of AI against working editors on editorial judgment, which is why the evidence grade here is D. That grade means not measured, so this page gives no parity number for editors. Error-checking benchmarks exist, but catching a comma splice is not the same test as deciding whether a story holds up.
What would settle it is specific: a blind comparison where qualified editors and a model work the same unedited drafts, and independent reviewers score the results on accuracy, structure, tone and the calls made about what to cut. Repeat it across news copy, book manuscripts and technical documentation, because those are different jobs wearing one O*NET title. Until that exists, treat confident claims in either direction with care. Our quality parity method explains how a graded result would be scored.
Good to know: the Bureau of Labor Statistics counts about 91,690 editors in the United States with median pay of $77,920, and projects employment down 1.1% over 2025 to 2035 (BLS, 2025 to 2035 projections).
When the picture could change
Most likely between 2036 and 2046 (8 in 10 of our scenarios). The replacement-year method sets out what that window measures and how the spread is produced.
Two things could pull it earlier. Publishers facing cost pressure may push more copy straight from draft to a model-assisted pass, especially for high-volume formats like product pages, listings and summaries. Second, tooling is already inside the systems editors use, so there is no new hardware to buy and little training friction.
Two things slow it down. Legal and reputational risk sits with a named person, and organizations are reluctant to hand sign-off to a system that cannot be held responsible. And house voice is tacit. Much of what an editor enforces is not written in any style guide, which makes it hard to specify and easy for a model to flatten. Hiring patterns matter here too: the strain shows up first in junior roles, which you can follow on the entry-level hiring tracker.
How editors stay needed
Lean into the tasks that stay with a person. Own the publication decision, including what gets killed. Build the writer relationship, so you are the one who can tell a contributor their lead is buried and get a better draft back. Take the risk calls on accuracy, sourcing and sensitive material, and document how you made them.
Two skills compound. First, directing models well: writing clear briefs, setting house constraints, and spotting where a clean-reading draft is confidently wrong. Second, structural editing at the level of argument and sequence, which is the work least well served by sentence-level tools. The AI skills employers want guide covers what hiring managers are asking for.
Nearby jobs worth comparing: writers and authors, technical writers and proofreaders and copy markers, which separate out the checking layer. The media and communication workers family and the information sector page show how editing sits against the rest of publishing. You can also put two titles side by side on the compare page, or read how all three scores are built in the full methodology.