Signals · Hiring Data
Speechify posted one AI job 245 times. Underneath, AI hiring held at 28% of new tech roles
Strip out Speechify's 245 cloned listings and AI-titled roles held between 28% and 30% of new tech postings for three straight weeks.
by Jo·3 min read·
New here? Start with the free AI Survival Kit →
Speechify posted one AI job 245 times. Underneath, AI hiring held at 28% of new tech roles
0:00 / 7:04
This voice is generated by AI.
Siftwright Data reads the public career-page feeds of about 7,300 companies and published its weekly hiring count on 26 October. Between 19 and 25 October those companies posted 52,839 new jobs. 3,640 companies posted them, and 3,799 were software, data or AI roles. The largest tech poster was Speechify, with 246 roles. 245 of them were one job, "Software Engineer, AI Data", copied once per city and listed as remote. That single listing made up about one in five of the week's AI-titled postings. Strip it out and AI roles were 28% of new tech postings, 998 of 3,553. The two previous weekly counts read 28% and 30%.
A few posters set the headline
A count of listings is the easiest AI hiring number to publish and the easiest to move. Speechify moved this one with a copy-paste.
It has happened before. In the week of 12 to 18 October, Gramian Consulting Group posted 106 AI roles, about one in nine of every AI posting Siftwright recorded. A week earlier, Gramian topped the whole tech table with 136 roles. Siftwright's own summary is that consultancies and IT services firms post most of the volume, while product companies post fewer roles each.
With the clones removed, the share holds still. AI-titled roles sat between 28% and 30% of new tech postings across three consecutive weekly counts. A steady share is something you can plan around. A spike from one employer tells you about that employer.
Siftwright states its own limits. It sorts roles by keywords in the job title, so the categories are approximate. It also does not compare totals across weeks, because filled postings drop out of the feeds and older weeks look smaller than they were.
An open listing is often a standing pool
A comment under the post adds a second caution from a narrower market. AITraining.jobs tracks the job boards of platforms that hire people to train and grade AI models, including Mercor, Handshake AI and Surge. It reports that 13.2% of roles taken down between 3 August and 28 October were followed within four days by a listing with the same title on the same platform. On SuperAnnotate's expert board, re-posts were a third of removals.
Of the 4,305 roles open on 29 October, 70.1% had been live for 27 days or more. The tracker's own read is that this looks "more like standing talent pools than openings".
That is one tracker's data from contract AI-data work, posted as a comment. The commenter ends by asking whether the same thing happens on the mainstream applicant-tracking boards. Treat it as a hypothesis and test it against the posting dates in your own field.
Two numbers that do not move
Roles explicitly marked junior, intern or graduate made up 147, 118 and 182 of new tech postings across the three weeks: 4%, 4% and 5%. Salary ranges appeared on 9% of new tech roles in each of those weeks. Where US roles did publish a yearly range, the median midpoint in the week of 19 to 25 October was $216,000 across 237 postings. Siftwright warns that this sample skews senior and towards states with pay-transparency laws, so read it as a ceiling.
The bigger opportunity is in roles with no level attached. Mid-level or unspecified roles were 49% of tech postings that week, and Siftwright notes that many of them do not ask for years of experience.
AI does the routine work. You do the thinking. A scraper can count 52,839 postings. Deciding which of them are real, funded and open to you takes judgement, and that judgement decides where you apply.
Your Next Move
-
De-duplicate before you retrain. If you are thinking about moving into an AI-titled role, pull this month's postings in your field and merge identical titles across cities. Count the separate employers. If three firms supply most of the volume, you are looking at three hiring managers, and you can research each one by name.
-
Treat a listing that has been live for more than four weeks as a pool. Check the posting date before you apply. If a role has been up for 27 days or more, the application form is the slowest way in. Find the hiring manager or a current employee and ask two questions: is the role funded, and when does it close?
-
Ask for the range in the first conversation. Only 9% of tech postings publish pay, so you will usually get the number only by asking for it. Ask before the second interview. When you compare it against published midpoints, remember those skew senior.
I build systems like the one publishing this site. → Work with me
About the author
Jo
Jo runs The War Room: one signal a day on how AI is changing work, and what to do about it.
Sources
Get the Briefing
Want more intelligence like this?
The weekly briefing, free — and the AI Survival Kit with it.
More in Signals
See allSignals · Capability
Siemens CEO Roland Busch says AI could transform society in 7 to 8 years, and AI skills will decide who gains
3 min read
Signals · Power
Windows 11 is recording which AI coding agents you run, and in one US test the upload setting was on
3 min read
Signals · Labour Market
Pay moves to the bottleneck. A careers list that is really a map of who owns the work
3 min read