Will AI end document review jobs?

Technology-assisted review (TAR) is the established AI layer in e-discovery.

ABA Formal Opinion 512, the American Bar Association's July 29, 2024 guidance on generative AI, calls it a well-known use: lawyers use TAR "to categorize vast quantities of documents as responsive or non-responsive and to segregate privileged documents."

In that description the lawyers are the ones using TAR; the duties the opinion goes on to set out — competence, supervision and candor among them — are written for generative AI tools.

Generative AI is the second layer.

Opinion 512 lists document review among the tasks GAI tools may assist with — alongside legal research, contract review, due diligence, regulatory compliance and drafting — and it treats the help as assistance, with the duties still attached.

It also names the failure mode: some GAI tools are prone to “hallucinations,” providing ostensibly plausible responses that have no basis in fact, so uncritical reliance can mislead clients and courts.

The duties are spelled out.

A lawyer need not become an AI expert, but must have a reasonable understanding of the capabilities and limitations of the GAI tools in use, and Opinion 512 ties GAI use to competence, confidentiality, client communication, supervision, meritorious claims, candor to the tribunal and reasonable fees.

Those duties sit on the lawyer using the GAI tool, not on the tool — which is the part of the AI question that matters for a review career.

Start with what the sourced data does and does not say.

Our research found no primary data on AI's effect on document review hiring volumes — no before-and-after count of review projects, no headcount series for review attorneys, and no verified percentage of legal work the tools can absorb.

A precise automation percentage you may have seen quoted is, for our purposes, an unverified estimate, and this page will not repeat one.

The federal numbers that do exist cover lawyers as a whole.

Lawyers (SOC 23-1011) held about 863,700 jobs in 2025, and the BLS projects lawyer employment to grow 5% from 2025 to 2035 — about 28,700 openings a year on average, many from replacing workers who retire or change occupations.

On automation specifically, the BLS says some routine legal work may be automated, but "this increased efficiency is not expected to reduce overall demand for lawyers."

Neither number is a document review number, and that is the honest limit.

The BLS does not break lawyers out by practice area, and our research found no series that counts document review attorneys on their own to trend.

What the sourced material supports is narrower: the BLS describes automating some routine legal work as an efficiency gain, the lawyer market as a whole is projected to grow, and the composition of review work — who does the first pass, who checks it, who supervises it — is where the career question actually lives.

That composition is the next section's subject.

No sourced data ranks which review-adjacent titles are growing, and this page will not invent a league table.

What the guidance does pin down is the duties that stay with people even when software takes the first-pass sort.

The seats those duties land on — checking the output, directing the process, answering for the calls — are the career-relevant ones, and three of them map directly onto document review.

• Quality control (QC). The hallucination warning has teeth exactly where machine output gets relied on at scale — plausible-but-baseless coding that goes unchecked can mislead clients and courts. The QC seat exists to catch that before it carries consequences: checking the machine's calls instead of relying on them uncritically.
• Review management and supervision. Opinion 512 writes the supervisory duty down: managerial lawyers must set clear firm policies on permissible GAI use, and supervisors must make sure lawyers and staff comply and are trained. The policies belong to the firm's managerial lawyers; on a review project, the supervising seat is the one that makes sure the team follows them and is trained — work that sits above the first-pass coding rather than beside it.
• eDiscovery counsel. The attorney role that owns the process the review sits inside, from collection through production and the legal calls in between. Our eDiscovery attorney guide covers that seat in detail; the staff-side workflow beside it belongs to eDiscovery specialists.

Read together, the pattern is consistent with the ethics guidance rather than any headcount: the further a seat sits from first-pass sorting, and the closer it sits to verification, management and the legal calls, the more of its description the sourced duties actually cover.

That is a positioning argument, not a projection — but it is the part of the question a review attorney can act on.

None of this requires predicting the technology.

The sourced guidance is concrete about what the lawyer-side half of AI-assisted review looks like, and each piece doubles as a career move:

• Know that learning time isn't billable to clients. Under Opinion 512, a lawyer may not bill a client for time spent learning a GAI tool the lawyer will use regularly. The client does not fund that learning curve, so treat it as a professional investment — the skill stays with whoever builds it.
• Aim at "reasonable understanding," not vendor-deck fluency. The bar the opinion sets is a reasonable understanding of the capabilities and limitations of the tools in use — what they do well, where they fail, and what checking their output needs. That is a learnable, demonstrable skill.
• Build a record of disciplined checking. The cautionary case is instructive: in Mata v. Avianca (S.D.N.Y., No. 22-cv-1461, June 22, 2023), the court sanctioned two lawyers and their firm for filing non-existent opinions with fake quotes and citations generated by ChatGPT, imposing a $5,000 penalty jointly and severally. The court's order also faults the lawyers for continuing to stand by the fake opinions after judicial orders called their existence into question — but the trouble started with output nobody checked, which makes verification the skill the tools cannot self-supply.
• Know the duties that attach to the tools. Opinion 512 ties GAI use to competence, confidentiality, client communication, supervision, meritorious claims, candor to the tribunal and reasonable fees. An attorney who can operationalize those duties on a review — the policies, the training, the verification steps — is useful to the firms that need the tools and the duties handled together.

The same logic holds one level up.

The ABA's guidance and the BLS outlook both treat AI as something lawyers work with, under duties, rather than a replacement for judgment; our guide to AI and lawyer jobs takes the whole-profession view, and this page stays on the review-specific slice.

Review work posts under several names — document review attorney, review attorney, and eDiscovery-adjacent titles — so search the variants rather than one string.

On LawFirmHires, document review attorney jobs are collected in one place; read each posting for its own scope, platform and arrangement rather than the headline.

New to the role, or weighing it?

The document review attorney guide covers what the job actually is — the work, the platforms, the employers — and our evaluation of whether document review is a good career takes the honest long view, including how the pay works.

Career information, not legal advice — the ethics rules that govern AI use are set by the jurisdiction that regulates your practice; confirm current guidance there.