SignalGigs original data
Which kinds of AI work appeared each month?
A month-by-month reconstruction of when our tracker first observed different kinds of work. The same title rules and duplicate handling apply to every month. These are observations from our coverage, not a census of hiring demand.
Collection-run history through October 29, 2026. Published October 31, 2026. November 2026 is a partial observation period.
Read the counts with their coverage
November 2026 contains 86 first-observed groups in this reconstruction. It has 0 dates with recorded successful collection through Unavailable. Collection gaps can shift the apparent arrival of a posting; a larger count does not by itself prove that demand grew.
| Signal | September 2026 | October 2026 | November 2026Partial observation |
|---|---|---|---|
| All first-observed listing groups | 1,167 | 978 | 86 |
| Physician and specialist-doctor titles | 15 | 44 | 1 |
| Search-engine evaluation titles | 5 | 40 | 1 |
| Speech and voice recording titles | 22 | 55 | 0 |
| Paid research-study titles | 3 | 22 | 2 |
| Attorney and lawyer titles | 31 | 22 | 1 |
The observation calendar
September 2026
8 dates with a successful collection run.
First: September 8, 2026
Last: September 29, 2026
October 2026
9 dates with a successful collection run.
First: October 2, 2026
Last: October 29, 2026
November 2026
0 dates with a successful collection run.
First: Unavailable
Last: Unavailable
Use a signal to decide what to investigate
A rising title count can point to a field worth checking. It cannot establish the number of projects, applicants hired or hours available. Read the actual tasks and eligibility conditions before using a field trend to plan training or income.
The original October report remains a separate dated editorial analysis. Its physician counts of 7, 20 and 56 belong to its reviewed method and cutoff. This page uses the exact title patterns and grouping rules in the download, so those figures should not be spliced into this series.
Method and limits
Count groups by earliest saved first-seen month, regardless of current status. Merge identical source URLs within a source and the saved October 30 reviewed duplicate mappings. Do not merge titles alone. Exclude salary-only groups. Match the published title regexes consistently across all three months; signals may overlap.
5 salary-only groups are excluded across these months. Closed listings remain in the historical count. Duplicate groups are assigned to their earliest saved observation even if a later copy or a new headline appeared in a different month.
Successful collection dates are a coverage indicator, not proof that every source was checked on each day. The public JSON publishes the exact title patterns, group members, observation dates and method version. Source category labels are not used for this comparison because later recategorization can alter historical totals.
- First seen is when this tracker observed a listing, not its employer publication date.
- Periodic collection and gaps can move apparent growth between days or months. No equal-exposure or market-wide demand claim is made.
- The newest month ends at the saved observation cutoff and is incomplete. Earlier calendar months also have intermittent coverage.
- These title-based signals do not measure hires, worker earnings or unique vacancies.
- This versioned reconstruction differs from the October article's manually reviewed counts; the dated article baseline is preserved and is not silently replaced.
Method version 2026-10-31.1. Downloaded JSON includes the input fingerprint for reproducibility.