What the data shows
- 1
Of the 195 robot training roles listed from 14 Sep 2026 to 31 Oct 2026, half were taken down within 36.4 days of posting (Kaplan-Meier median); after four weeks, 68.7% were still up.
- 2
A median built from removed roles alone comes to 22.3 days; add back the 75.4% that are still up and it is 36.4 days, 63.0% longer.
- 3
The quickest turnover is in Robot Fleet Operations (half gone within 34.2 days, 71.9% still up at four weeks). AV & Drone Operators lasts longest: 88.9% of its roles were still up four weeks after posting.
- 4
Among employers, micro1 pulls roles soonest (half gone within 12 days, 8.3% still up at four weeks), while Aurora Innovation keeps them up longest (100.0% still up at four weeks).
- 5
Some apparent turnover is just reposting: 8.3% of removed roles were followed by a fresh listing with the same title from the same employer at a later date.
- 6
On 1 Nov 2026, 720 of 1,150 open roles (62.6%) had been up for 47 days or longer, past the point where even the longest-lasting listings usually come down. Egocentric & Wearable Capture had the highest share, at 88.8% of its open roles.
Figure 1
Week by week: the share of listings still up
Of the robot training roles posted, what share is still up one, two, four or eight weeks later?
| Age of the listing | Still listed | ||
|---|---|---|---|
| After 1 week | 92.8% | 148 | 137 |
| After 2 weeks | 88.7% | 91 | 77 |
| After 3 weeks | 79.6% | 90 | 71 |
| After 4 weeks | 68.7% | 59 | 45 |
| After 6 weeks | 28.7% | 11 | 4 |
| After 8 weeksPartial week | n/a | 0 | 0 |
Base: 195 roles with a known listing date, listed from 14 Sep 2026 to 31 Oct 2026; 48 had come down by 1 Nov 2026. Back catalogues are excluded.
Method: Kaplan-Meier. A removed role's life ends on the last day the feed showed it. A role that is still up stays in the count for every day it has been observed, then drops out without counting as a removal.
Ages are measured in hours from the posting date on the job board, or from the role's first day in the feed when that came earlier or no date is given. When the first sighting comes more than 2 days after the posting date, the feed was slow to show the role, so it only enters the count at the age it was first seen. A gap of 2 days or less is just the daily read catching up, and the role counts from the day it was posted.
Each share rests on the roles posted at least that long before 1 Nov 2026, and the count of those roles sits next to it. An age that no role in the group has reached yet shows n/a.
Credit: RobotTraining.jobs Research, Robot Training Job Lifespans: How Long Listings Last, October 2026 edition.
Figure 2
Listing lifetimes by kind of work
Which kinds of robot training work come down quickest, and which stay up the longest?
| Kind of work | Still listed after 2 weeks | After 4 weeks | Median lifetime (days) | ||
|---|---|---|---|---|---|
| All kinds of work | 195 | 88.7% | 68.7% | n/a | 36.4 |
| AV & Drone Operators | 80 | 92.9% | 88.9% | n/a | n/a |
| Robot Fleet Operations | 75 | 98.5% | 71.9% | n/a | 34.2 |
| Data & Demonstration Collection | 13 | 88.9% | 66.7% | n/a | n/a |
| Robot Task Evaluation & QASmall sample | 2 | 50.0% | 50.0% | n/a | n/a |
| Robot TeleoperationSmall sample | 7 | 44.4% | 44.4% | n/a | 6.6 |
| Embodied-AI Data AnnotationSmall sample | 8 | 87.5% | 29.2% | n/a | 26.0 |
| Egocentric & Wearable CaptureSmall sample | 10 | 0.0% | 0.0% | n/a | 12.0 |
Base: 195 roles with a known listing date, listed from 14 Sep 2026 to 31 Oct 2026, grouped by their classified kind of work.
Both the shares and the median lifetime come from Kaplan-Meier estimates: a role that is still up counts for every day it has been observed.
An n/a median lifetime has one of two causes. Either over half the group's roles were still up at the oldest age the data reaches, so the median lies beyond what has been observed, or fewer than 3 of its roles were posted early enough to read it (no median here rests on fewer). A share shows n/a when none of the group's roles was posted that long ago.
A kind of work whose roles add up to fewer than 176 days of observed listing time gets a small-sample flag and no rank. Below that, the removal pace shared by all kinds of work tells more about it than its own removals do; how much weight that shared pace carries is fitted from the removals themselves.
Credit: RobotTraining.jobs Research, Robot Training Job Lifespans: How Long Listings Last, October 2026 edition.
Figure 3
Listing lifetimes by employer
Which employers take their robot training roles down soonest?
| Employer | Still listed after 2 weeks | After 4 weeks | Median lifetime (days) | |||
|---|---|---|---|---|---|---|
| TSMG | 35 | 100.0% | 100.0% | n/a | n/a | n/a |
| Avride | 31 | 100.0% | 79.3% | n/a | 34.2 | 0.0% |
| Tesla | 31 | n/a | n/a | n/a | n/a | n/a |
| micro1 | 18 | 49.7% | 8.3% | n/a | 12.0 | 0.0% |
| Amazon Prime AirSmall sample | 10 | 100.0% | n/a | n/a | n/a | 100.0% |
| Aurora Innovation | 8 | 100.0% | 100.0% | n/a | 31.4 | 66.7% |
| Serve Robotics | 7 | 100.0% | 75.0% | n/a | n/a | 0.0% |
| Wayve | 6 | 83.3% | 66.7% | n/a | n/a | 0.0% |
| Anduril Industries | 4 | 100.0% | 75.0% | n/a | n/a | 0.0% |
| Collaborative Robotics | 3 | 100.0% | 100.0% | n/a | n/a | n/a |
| Genesis AI | 3 | 100.0% | 100.0% | n/a | n/a | n/a |
| Kodiak AISmall sample | 3 | 66.7% | 66.7% | n/a | n/a | 0.0% |
| MotionalSmall sample | 3 | 66.7% | 66.7% | n/a | n/a | 100.0% |
| Torc RoboticsSmall sample | 3 | n/a | n/a | n/a | n/a | 0.0% |
| TransdevSmall sample | 3 | n/a | n/a | n/a | n/a | n/a |
| WaymoSmall sample | 3 | 66.7% | 66.7% | n/a | n/a | 0.0% |
| Figure AISmall sample | 2 | 0.0% | 0.0% | n/a | n/a | 0.0% |
| MonumentalSmall sample | 2 | 100.0% | n/a | n/a | n/a | n/a |
| MooveSmall sample | 2 | n/a | n/a | n/a | n/a | n/a |
| Nimble RoboticsSmall sample | 2 | 50.0% | 50.0% | n/a | n/a | 0.0% |
| Shield AISmall sample | 2 | n/a | 100.0% | n/a | n/a | n/a |
| SymboticSmall sample | 2 | n/a | n/a | n/a | n/a | n/a |
| DexorySmall sample | 1 | 100.0% | 100.0% | n/a | n/a | n/a |
| Dyna RoboticsSmall sample | 1 | n/a | n/a | n/a | n/a | n/a |
| HumanoidSmall sample | 1 | 100.0% | 100.0% | n/a | n/a | n/a |
| Keywords StudiosSmall sample | 1 | n/a | n/a | n/a | n/a | n/a |
| Mind RoboticsSmall sample | 1 | n/a | n/a | n/a | n/a | n/a |
| NuroSmall sample | 1 | 100.0% | n/a | n/a | n/a | n/a |
| Skild AISmall sample | 1 | 0.0% | 0.0% | n/a | n/a | 0.0% |
| Stack AVSmall sample | 1 | 100.0% | n/a | n/a | n/a | n/a |
| Tutor IntelligenceSmall sample | 1 | n/a | n/a | n/a | n/a | n/a |
| VaySmall sample | 1 | 100.0% | 100.0% | n/a | n/a | n/a |
| Welo DataSmall sample | 1 | n/a | n/a | n/a | n/a | n/a |
| ZooxSmall sample | 1 | n/a | n/a | n/a | n/a | 0.0% |
Base: 195 roles with a known listing date, listed from 14 Sep 2026 to 31 Oct 2026, grouped by the employer posting them. Based on listings alone.
An n/a median lifetime has one of two causes. Either over half the group's roles were still up at the oldest age the data reaches, so the median lies beyond what has been observed, or fewer than 3 of its roles were posted early enough to read it (no median here rests on fewer). A share shows n/a when none of the group's roles was posted that long ago.
Re-listed means a removed role whose title the same employer listed again at a later date (the gaps between a removal and the next listing with that title form one kind only, so every such listing counts whatever the gap). The column is filled only for employers with enough removed roles to read on their own: 0.9 or more removed roles, a weight fitted from the employers' own counts. Across the cohort, 8.3% of removed roles were re-listed this way. An employer that pulls roles and reposts them, say every week, shows short lifetimes here even if the job itself stays open for months. Listings are counted exactly as the feeds publish them, with no further merging.
Where an employer posts a separate listing for each city or country, every one of them gets its own lifetime.
A employer whose roles add up to fewer than 68 days of observed listing time gets a small-sample flag and no rank. Below that, the removal pace shared by all employers tells more about it than its own removals do; how much weight that shared pace carries is fitted from the removals themselves.
Credit: RobotTraining.jobs Research, Robot Training Job Lifespans: How Long Listings Last, October 2026 edition.
Figure 4
Counting only removed roles gives too short a median
By how much does a median built only from removed roles understate how long listings really stay up?
| Kind of work | Still listed on 1 Nov 2026 | Median days, removed roles only | Median lifetime, all roles (days) | |
|---|---|---|---|---|
| All kinds of work | 195 | 75.4% | 22.3 | 36.4 |
| AV & Drone Operators | 80 | 91.3% | 19.4 | n/a |
| Data & Demonstration Collection | 13 | 76.9% | 27.0 | n/a |
| Robot Fleet Operations | 75 | 66.7% | 29.1 | 34.2 |
Base: Same cohort: 195 roles listed from 14 Sep 2026 to 31 Oct 2026. Only kinds of work whose lifetimes can be read on their own are shown, as in the chart by kind of work.
Removed roles only: take the roles that have already come down and find the median number of days from posting to last sighting. This is the number a hiring report gives as 'median days listed'. It ignores every role still up, and those are the ones that last longest, so the figure runs short. The more roles still up, the bigger the gap.
All roles: the Kaplan-Meier median. Every role still up is included for each day it has been observed.
An n/a median lifetime has one of two causes. Either over half the group's roles were still up at the oldest age the data reaches, so the median lies beyond what has been observed, or fewer than 3 of its roles were posted early enough to read it (no median here rests on fewer). A share shows n/a when none of the group's roles was posted that long ago.
The removed-only median is shown only when a group has 3 or more removed roles, the fewest any median here rests on.
Credit: RobotTraining.jobs Research, Robot Training Job Lifespans: How Long Listings Last, October 2026 edition.
Figure 5
Evergreen roles, still up past the normal life of a listing
Of the roles open right now, how many were first posted 47 days or more ago?
| Kind of work | Listed 47+ days | Share of open roles | ||
|---|---|---|---|---|
| All kinds of work | 1,150 | 720 | 62.6% | 101 |
| Egocentric & Wearable Capture | 134 | 119 | 88.8% | 101 |
| Robot Teleoperation | 82 | 67 | 81.7% | 119 |
| Robot Task Evaluation & QA | 9 | 7 | 77.8% | 73 |
| AV & Drone Operators | 544 | 369 | 67.8% | 117 |
| Data & Demonstration Collection | 127 | 56 | 44.1% | 17 |
| Robot Fleet Operations | 237 | 96 | 40.5% | 31 |
| Embodied-AI Data Annotation | 17 | 6 | 35.3% | 29 |
Base: All 1,150 roles open on 1 Nov 2026, whatever their listing date, back catalogues counted in.
A role counts as evergreen when it is still open on 1 Nov 2026 and was posted 47 days or more earlier. Past that age even a listing of the longest-lasting kind is more likely to have come down than to still be up. The cut-off comes from fitting the lifetimes of the 195 roles listed from 14 Sep 2026 to 31 Oct 2026 as a mix of 1 kind: the longest-lasting kind holds 100.0% of them and stays up 67.3 days on average, so a role open beyond the cut-off is being kept open on purpose.
When a back-catalogue role has no posting date on its board, its age starts on the first day this site saw it. Its true age may be higher, so both that age and the evergreen count are minimums.
A row with under 5.9 open roles carries a small-sample mark and is left unranked: under that count, the average across all rows says more than the row's own open roles do (a weight fitted on the counts themselves).
Credit: RobotTraining.jobs Research, Robot Training Job Lifespans: How Long Listings Last, October 2026 edition.
Figure 6
Weekly churn: open roles that come down
Each week, what share of open robot training roles is taken down?
| Week | Share removed in the week | Open when the week began | ||
|---|---|---|---|---|
| 14 to 20 SepPartial week | 2.2% | 647 | 14 | 9 |
| 21 to 27 SepPartial week | 3.9% | 642 | 25 | 21 |
| 28 Sep to 4 OctPartial week | 3.8% | 688 | 26 | 30 |
| 5 to 11 OctPartial week | 2.0% | 693 | 14 | 47 |
| 12 to 18 OctPartial week | 2.5% | 746 | 19 | 8 |
| 19 to 25 OctPartial week | 2.5% | 1,026 | 26 | 75 |
Base: Each full ISO week (Monday through Sunday, UTC) between 14 Sep 2026 and 31 Oct 2026, counting only employers that were already tracked when the week began.
Churn takes the roles open at 00:00 UTC on Monday and asks what share were seen for the last time that week and had left the feed by the end of the observation. A role posted and pulled inside the same week is left out of the rate and counted under new listings instead.
A week only includes employers whose tracking started before it, so bringing a new employer into the dataset cannot show up as churn or growth.
When one employer pulls many roles at once, for instance because a project ended or its board was reorganised, that week stands out as a spike.
A row with under 2571 roles open at the week's start carries a small-sample mark and is left unranked: under that count, the average across all rows says more than the row's own roles open at the week's start do (a weight fitted on the counts themselves).
Credit: RobotTraining.jobs Research, Robot Training Job Lifespans: How Long Listings Last, October 2026 edition.
Figure 7
Do roles that draw applicants in their first week come down sooner?
Once a robot training role has been up a week, does it come down sooner when people tried to apply to it in that week?
| In its first week here | Roles still up at one week | Still up 2 weeks later | 4 weeks later | Median lifetime (days from posting) | |
|---|---|---|---|---|---|
| Someone clicked apply on it hereSmall sample | 25 | 7 | 72.3% | 56.4% | n/a |
| Nobody clicked apply on it hereSmall sample | 104 | 20 | 96.2% | 55.4% | 36.4 |
Base: 129 of the roles listed from 14 Sep 2026 to 31 Oct 2026 that were still up one week after posting, first seen on their posting day and posted after this site began logging apply clicks (18 Sep 2026).
Landmark method: each role's life is read from the end of its first week onward, so a role cannot be counted as attracting applicants just because it stayed up longer. Kaplan-Meier, as in the other figures.
Apply clicks are human clicks on this site's apply buttons for that role (automated traffic left out); employers see applicants from many places, so this measures attention from this site's visitors, not a role's whole demand. A role coming down is not proof that it was filled.
Every group gets a small-sample flag and none is ranked: the groups take roles down at paces no further apart than chance alone would produce.
Credit: RobotTraining.jobs Research, Robot Training Job Lifespans: How Long Listings Last, October 2026 edition.
Robotics labs ยท Fleet operators ยท Data vendors ยท Research teams
Buy the data behind this report
This report sums up records we gather ourselves, day after day:
1,432 robot training roles followed, history included
Every role on 100 employer job boards, read each day since 12 Sep 2026: its pay, its place, and the first and last day it was up.
536 apply clicks over 28 days
Which robot training roles people really apply to, by kind of work, pay and place. The clicks happen here, so no one else can count them.
25 sourced industry datasets
Funding, revenue, deals, fleets, deployments and prices across the robot training industry, each record linked to the publication it comes from.
On offer
The raw records
Every robot training role we track: title, employer, place, kind of work, pay as stated, and the days it was first and last open.
A regular feed
New and closed roles every week or month, as file drops or through the data endpoints.
A custom cut
Any figure here by country, city, kind of work, vehicle, employer or date, or your own hiring benchmarked against the market.
Need the people behind this data for your own robots? (you contract them yourself) or (we deliver the robot data).
Investing in robotics, AV or robot-data companies? The Robot hiring radar is these records role by role, every day, as files.
Robot hiring brief: free, once a month
The robot-training job market in figures, on the first weekday of each month: open roles, what opened and closed, who added and who cut, and posted pay by kind of work. To your work email.
How it was measured
Daily reads, starting 12 Sep 2026, of the public job boards run by the robot training employers and programs this site follows. A single database query produces each figure for the window shown, and the live page and every monthly edition run that identical query. The note under each figure gives its base.
- Listing date
- The posting date the employer's job board shows, or, failing that, the first day the role turned up in the feed. A role that came in with an employer's back catalogue and carries no date has no known listing date.
- Cohort
- All roles with a known listing date posted in the 26 weeks up to the end of the period, none earlier than 14 Sep 2026 (when this site's index begins). Each employer's back catalogue, meaning whatever it already had up in its first week in this dataset, is left out.
- Lifetime
- Measured in hours and shown in days, from the listing date to the role's final sighting in its employer's feed. A role still up when the period ends has lived at least that long, so it is treated as censored rather than removed.
- Still listed after N weeks
- A Kaplan-Meier estimate of what share of roles remain in their feed N weeks past their listing date. Each role contributes for as long as it has been observed, which stops roles that are still up from passing for short-lived ones. If a role was first spotted more than two days after its listing date, it surfaced late and is counted only from the age at which it was spotted.
- Median lifetime
- The estimated age at which half of all roles have been taken down. If more than half are still up at the oldest age the data reaches, the median has not been reached yet.
- Removed-only median
- The familiar 'median days listed': days from posting to last sighting, for roles already taken down. Because it ignores roles still up (the longer-lived ones), it comes out shorter than the median lifetime.
- Re-listed
- A role that came down and later reappeared under the same title at the same employer inside the reposting window. That window comes from the gaps between a removal and the next listing with the same title, split into kinds (most removed roles never return): a gap counts while it is likelier to belong to the shortest kind, an employer putting the same role up again, than to any longer kind, a fresh opening that happens to share the title.
- Evergreen role
- A role still open when the period ends that is older than the age at which even the longest-lived kind of listing is more likely gone than up. The cohort's lifetimes are fitted as a mix of kinds, each with its own daily chance of coming down, and the cut-off is the first whole day t at which q to the power t is one half or less, q being the longest-lived kind's daily chance of staying up.
- Weekly churn
- Take the roles open at 00:00 UTC on Monday at employers followed before that week; churn is the share of them taken down by the end of the ISO week.
- Kind of work
- A role's title and description place it in one of the seven kinds of robot training work this site follows, ranging from robot teleoperation and demonstration collection through egocentric capture, embodied-AI annotation, robot evaluation and fleet operations to AV and drone operation. Each kind has a page of its own.
- Window
- Live page: roles posted in the 26 weeks to yesterday (UTC), none before 14 Sep 2026. Edition: roles posted in the 26 weeks to month end, none before 14 Sep 2026, followed up to that month end.
Limits of this data
- Hiring that happens without a public job board is not captured.
- A posting does not equal a hire, and a role that comes down may have been filled, put on hold or cancelled; employers do not say which.
- A brief listing life does not always mean a brief opening: some employers pull a role and repost it, or post one copy per city. The re-listing column flags where this happens.
- Feeds are read roughly once a day. A role that goes up and comes down between reads can be missed, and each removal is dated to the final read that still showed it.
- No estimate is made past the longest age the cohort has actually been watched, so while the dataset is young the later ages read n/a.
- When an employer pulls many roles at once (a project wraps up, a board gets reorganised), it shows up as one spike in weekly churn.
Collection and classification of listings are explained on the methodology page. Freshness, completeness and consistency are tested daily in the data quality report.
Monthly editions
The live page tracks each role listed in the 26 weeks up to yesterday (none earlier than 14 Sep 2026) and updates daily. Every finished month also gets its own edition, locked two days after month end, so a number quoted from it stays fixed.
- Live pageAll roles of the last 26 weeks, rebuilt daily
- October 2026 edition (you are here)Will lock when the next daily refresh runs
Quoting this report
You may quote, chart or republish these figures under CC BY 4.0, provided you link back to this page. This edition is locked: its numbers will not change.
Citation
RobotTraining.jobs (2026). Robot Training Job Lifespans: How Long Listings Last, October 2026 edition. https://robottraining.jobs/research/how-long-robot-training-roles-stay-open/2026-10
BibTeX
@misc{robottrainingjobs_how_long_robot_training_roles_stay_open_2026_10,
author = {{RobotTraining.jobs}},
title = {Robot Training Job Lifespans: How Long Listings Last, October 2026 edition},
year = {2026},
howpublished = {\url{https://robottraining.jobs/research/how-long-robot-training-roles-stay-open/2026-10}},
}Keep reading
- Robot Training Hiring: Openings, Turnover and EligibilityHow many robot training jobs are live, which employers post them, how quickly do listings come and go, and who is eligible to apply?Open report
- Scorecard of Robot Training EmployersSide by side, how do robot training employers and programs stack up on openings, pay, who can apply and how fast roles turn over?Open report
- Open Robot Training Jobs, Counted WeeklyWeek over week, is demand for people to train robots rising or falling?Open report
- Who Dominates Each Kind of Robot Training WorkFor each kind of robot training work, how many employers are competing, and how big a slice does the largest one have?Open report
- Robot Training Job Posting Times and Apply PeaksOn which days and at what hours do robot training jobs go up, and when do job seekers click apply?Open report
- Where Robot Training Hiring Is Heading, Week by WeekWeek over week, which kinds of robot training work, professions and languages are employers hiring more of, and which are cooling?Open report
- Robot Training Employer Health: Who Is Growing, Holding or ShrinkingWhich robot training employers are adding roles, holding steady or cutting back, and are the rates in their new listings rising or falling?Open report
- Robot Training Pay Report 2026How much do robot training jobs pay, broken down by kind of work and employer?Open report