Source
Every cutoff date comes from the monthly Visa Bulletin published by the U.S. Department of State at travel.state.gov. We read the two employment-based tables in each bulletin: "Final Action Dates for Employment-Based Preference Cases" and "Dates for Filing of Employment-Based Visa Applications".
Because travel.state.gov blocks automated downloads, each bulletin is read from its archived copy on the Internet Archive's Wayback Machine, taken from the official URL. We use the bulletin's web page where a clean copy is archived, and otherwise the official PDF of the same bulletin (travel.state.gov/content/dam/visas/Bulletins/). A bulletin that does not yield both employment tables is rejected rather than guessed at. The source column in the download says which one each row came from.
One exception, October 2026: State refused the Wayback Machine's crawler too, so no archived copy exists. We read it from the copy of State's PDF that the American Immigration Lawyers Association attached to its October 2026 bulletin post (AILA Doc. No. 26092902, posted September 29, 2026). It is State's layout and numbering (Number 19, Volume XI) with AILA's document number stamped on each page, so it is not State's own file and we cannot check it against one. It went through the same hand check and second reading as every other month, and its EB-1, EB-2 and EB-3 India cells (both charts) match prachub.com's October 2026 figures. If State's own file later becomes readable, we will compare the two and say so here.
Check on the PDF reader: for the 12 months from October 2016 to September 2017, where both versions exist, the PDF and the web page give the same value in all 1,056 cells.
Second reading of every cell: recent bulletins exist only as PDFs, so there is no web page to compare against. Instead every PDF is read a second time by a different program (PyMuPDF instead of pypdf) using a different method: rather than taking the values in text order, it places each value by its position on the page, in the row whose label is on the same line and the column whose header is closest. Every cell of every parsed month is compared. A month is published only if all its cells agree, and the result is tied to the file's fingerprint (below). The second reading also counts every date, "C" and "U" printed in the two tables, so a row that both programs skip still shows up as a gap. On 2026-10-08: 6,011 values printed in 61 bulletins, 6,011 parsed, and all 6,011 agree.
Corrected bulletins: the State Department sometimes replaces a bulletin with a corrected copy under the same file name. Before each rebuild we check every file against the newest copy the Wayback Machine has archived. If a newer copy differs, the month is not published until it is read and checked again. So far we have seen this once: the October 2024 bulletin was corrected 14 days after its first archived copy (EB-3 Other Workers India Date for Filing, June 1, 2013 to June 8, 2013). That is one case, so we cannot say how often it happens. The site uses the corrected copy.
Hand check on every published month: for each month we publish, we also render the printed tables as an image and read two to six cells by eye (EB-2 India Final Action and EB-3 China Dates for Filing; from mid-2024 also EB-2 rest of world Final Action, and for some months EB-3 India Final Action, which moves almost every month; for June to August 2023 also EB-1 India and EB-3 rest of world Final Action, the lines the October 2023 note quotes), and record them with a fingerprint (SHA-256) of the exact file that was read. The site is only rebuilt when every recorded cell matches the parsed value. If the State Department republishes a corrected bulletin and the file changes, its fingerprint no longer matches, so that month counts as unchecked again until it is re-read. Months that have not been hand-checked yet are left out, and so is anything older than a missing month, so the table never has a hole in it.
The "Change vs. previous bulletin" column in each history table compares a month's Final Action cutoff with the bulletin before it ("+2 months", "back 10 months", "no change").
"What changed" on each category page compares the two newest bulletins in the table and names both months. If a newer bulletin is being held back because its file changed after the hand check, the line says so instead.
What is in the table
- One row per bulletin month, chart (Final Action or Filing), category and country of chargeability.
- Categories: EB-1, EB-2, EB-3, EB-3 Other Workers, EB-4, Certain Religious Workers and the EB-5 rows exactly as each bulletin labels them.
- Values: a date, C (current: every priority date qualifies) or U (unavailable: no visas in that category this month). The original text from the bulletin is kept next to each value so any row can be checked.
Download it: eb-cutoffs.csv.
What the wait figures mean
The tool measures how many months of priority dates the Final Action cutoff advanced over the last 12, 36 and 60 bulletins, then divides the gap between the current cutoff and your priority date by that pace. It is plain arithmetic on the past. It does not know how many people are ahead of you, how many visas will be available next fiscal year, or what the State Department will do. Cutoffs regularly stall, jump forward, or move backward.
A figure over a fixed window depends on where the window starts. The government's fiscal year starts in October. Near the end of fiscal year 2023 several cutoffs were pulled back (retrogressed), and the October 2023 bulletin, the first of fiscal year 2024, moved them forward again. That October move was a reset, not new movement of the queue. EB-1 India is the clearest case: Final Action was February 1, 2022 in the July 2023 bulletin, January 1, 2012 in August and September 2023, and January 1, 2017 in October 2023, so the 5-year October jump only won back part of the 10 years lost in August. EB-3 India (June 15, 2012, then January 1, 2009, then May 1, 2012), EB-3 rest of world and EB-3 Philippines (June 1, 2022, then May 1, 2020, then December 1, 2021) and EB-1 rest of world (Current, then August 1, 2023, then Current) did the same. Each category page says in its history section what October 2023 did to that line.
Now that the October 2026 bulletin is in the table, the 36-month window starts at the October 2023 bulletin, after that reset, so the reset is no longer counted as movement (until October 2026 was added, the window started at September 2023 and did count it). When the bulletin right after a window starts moved the line forward 6 months or more, the figure is labelled and shown again measured from that next bulletin, and the "your date" estimate uses the later figure. The 6-month threshold is our own choice, not a standard.
Every cutoff on this site carries the month of the bulletin it comes from ("line as of October 2026"). The newest bulletin here can be a month or more behind the current one; see "Source" above.
Below the figures, each page counts how often Final Action moved forward, stayed the same or moved back from one bulletin to the next over the last 36 bulletins, with the median change. For most categories the median is "no change": the line moves in jumps, a few times a year, not a little each month.
How Current and Unavailable months are counted: a Current month means there is no line, so it is placed at the bulletin month itself, as in the chart. An Unavailable month has no cutoff and is skipped. A movement figure needs a cutoff at both ends of its window, so it shows n/a when the newest month is Current or Unavailable. In the forward/same/back counts, Current to Current and Unavailable to Unavailable count as the same; changes into or out of Current or Unavailable are left out because they have no length, and the median uses only changes between two dated cutoffs. Each page says how many changes were left out.
Forecasts
We don't publish a forecast of where the cutoffs will go. The pace figures above are arithmetic on past bulletins, not predictions.
If you want a forecast, look for one that keeps score. visa-bulletin.us publishes a model forecast before each bulletin and, after State publishes it, puts each predicted date next to the official one, with a scorecard for every bulletin month back to at least 2021. Its own October 2026 scorecard says the forecast missed by about 190 days on average across 19 dates, which was worse than simply guessing that nothing would change (by preference: EB-1 about 130 days, EB-2 about 110, EB-3 about 148, EB-5 about 487). prachub.com shows a straight-line "if the pace holds" projection on its category pages and labels it as not a prediction. (Checked 2026-10-08; we have no connection with either site.)
Nothing on this site is legal advice. For decisions about your case, talk to a licensed immigration attorney.
PERM processing times
The PERM page is built from the Wayback Machine's daily copies of DOL's processing-times page (flag.dol.gov/processingtimes). DOL overwrites that page each time it updates, so we read one archived copy per day, keep the PERM section exactly as printed, and group days whose text is identical into one version. The page states how many of the archived days have been read so far, the longest run of days not read yet, and the days the archive has no copy for. It prints DOL's figures and nothing derived from them: no speed for the queue and no date for a case.
What comes next
Snapshots of USCIS processing times (I-140 and I-485), which the government overwrites each time it updates them, and an estimate of how many applicants stand ahead of a given priority date from USCIS pending-inventory reports.