SSA has spent more than $300 million replacing the 1991 job data behind step five. In January 2025 it stopped requiring adjudicators to resolve conflicts with it.

In 2022, 35.8 percent of the SSDI worker claims that state agencies denied were denied at step five, on a finding that the person could still adjust to other work. On adult SSI claims it was 44.2 percent. The occupations named at that step come out of the Dictionary of Occupational Titles, which the Labor Department last updated in 1991. Most of the entries in it haven't been touched since 1977.

So a claim your office files this year can come down to whether there is still work in this country as a Cutter-and-Paster, Press Clippings.

SSA knows. Between FY2012 and FY2024 it obligated more than $300 million to the Occupational Information System, the project meant to replace the DOT, and BLS built an entire new survey to feed it. The first wave of that survey wrapped in FY2018. The second wrapped in FY2023. A third started the same year and preliminary data on part of it came out in FY2025, and each additional wave runs somewhere around $167 million. There is still no implementation date, because SSA has to rewrite its vocational policy to match the new taxonomy and then change the systems to match the policy, and it has never published a date for either.

Thirteen fiscal years of funding, and the book on the adjudicator's desk is the same book.

The rule that got rescinded

On January 6, 2025, SSR 24-3p took effect and rescinded SSR 00-4p. The language is not ambiguous. SSA said it "will no longer require our adjudicators to identify and resolve conflicts between occupational information provided by VSs and VEs and information in the DOT." That duty is gone for anything pending on or after that date.

What sits in its place is softer by design. SSA now expects the expert "to identify the sources of the data they use and, where applicable, to explain their general approach to estimating job numbers." Expect is doing a lot of work in that sentence. The agency also went out of its way in the same document to say it "continue[s] to recognize the DOT as a valid and reliable source of occupational information."

Reasonable people can read this as SSA admitting the DOT was never the only source and freeing experts to use better ones, which is more or less how the agency frames it. I don't think that reading is wrong. I do think the practical effect on your files runs the other way, because the old rule put an affirmative duty on the ALJ that survived into the record, and the new one puts an expectation on a witness you have about eight minutes with.

Where the job numbers actually come from

There are 867 SOC codes. There are over 12,000 DOT codes. Federal employment data counts people at the SOC level, so when a vocational expert testifies that 240,000 of something exist, she has taken a number that was collected for a broad group and worked her way down to a single 1977 job title, and no survey anywhere performs that step for her. It's an allocation. Average it out and you get roughly fourteen DOT titles sharing each SOC code, which of course tells you nothing about any particular one, and the ones that matter at your hearings are the sedentary unskilled titles where SSA reckons the entire universe is only about 200 occupations to begin with.

That is why the same handful of jobs turn up over and over.

The 240,000 is not a hypothetical, by the way. That was Erin O'Callaghan's testimony in Biestek, 240,000 bench assembler jobs and 120,000 sorters, and when Biestek's lawyer asked for the private market surveys behind it she declined to produce them. The Supreme Court held 6 to 3 in April 2019 that her refusal doesn't automatically knock the testimony out of the substantial evidence category. Kagan wrote it. Gorsuch dissented.

SSA has already pulled 127 occupations off the board

The agency has done real work here and it's worth knowing exactly how far the work goes.

On June 22, 2024, EM-24026 barred adjudicators from citing 114 DOT occupations to support a not-disabled finding under the grids. SSA found them by pulling OEWS employment data for 2020 through 2022 and picking out the SOC occupations with fewer than 1,000 employees in each of the nine census divisions. The same day a second message, EM-24027, took a different tack with a shorter list of jobs whose descriptions refer to materials or processes that no longer exist. Those can still be cited, but only with additional evidence from an expert explaining that the work is performed the way the DOT says. There were twelve when it came out and thirteen after the January 2025 revision. Addresser is on it. So is Tube Operator, and Nut Sorter, and Almond Blancher (Hand), and two separate ways of being a microfilm clerk.

Note the ceiling, though. That is 127 titles out of more than 12,000, and neither message touches step four. EM-24026 says so directly: adjudicators may keep relying on isolated jobs when the question is past relevant work. A job can be too rare in 2024 to prove your client could go get one, and still perfectly good as proof of what he used to do.

What I don't know

Whether any of this moved the numbers. The 35.8 and 44.2 percent figures are state agency determinations from 2022, which is before both emergency messages and two and a half years before SSR 24-3p, and they describe initial denials rather than hearing decisions, which is where your firm actually lives. I could not find published step five denial rates at the hearing level for 2025, and one year post-rescission would be thin evidence anyway. If SSA has that data it hasn't put it anywhere I can reach.

Everything above came out of four documents. CRS Report R48748, published December 4, 2025, is the survey of how SSA uses occupational information, and the denial shares and the funding figures are all from it. The two emergency messages are EM-24026 and EM-24027 REV, both on SSA's PolicyNet, effective June 22, 2024 and January 6, 2025. SSR 24-3p is at 89 FR 97158, December 6, 2024. The testimony is Biestek v. Berryhill, decided April 1, 2019. NOSSCR posted the combined table of all 126 occupations from the original messages, and I counted the rows myself rather than taking the summary's word for it.

We build the back office for disability firms, so the part of this that reaches my desk is prep time per file. If you want the other side of the outcome question, we pulled a year of ALJ decisions to see how much of it comes down to which judge picked up the case.

The thing I keep turning over is the timing. Inside of seven months the same agency issued a message saying 114 of the DOT's occupations are too rare to cite against a claimant, and a ruling saying the DOT is reliable enough that nobody has to resolve conflicts with it any more.