An AI Title Tells You What the Last Employer Called the Job

Who Wrote AI Into That Title, and Why?

The last employer wrote it, and the reason may have had nothing to do with the work under it. A title gets rewritten to make a requisition competitive, to hold on to somebody without moving their pay band, or because a function renamed itself in one memo. The person under it may simply have been told what their new title was. Read it as a fact about that company.

In the first quarter of 2022, Indeed Hiring Lab counted 264 US job titles where at least five postings carried AI or a related term in the employer's own raw wording. By the first quarter of 2026 the count was 822 ¹.

Those are distinct titles, each counted once however many postings sit behind it, so the number measures how far the word travelled through role names. The line is not straight either: the count fell to 159 in early 2023 before it climbed ¹.

And 63% of those AI-touched titles now sit outside tech occupations ¹.

Each of those figures counts words, and none can say how far the words sit from the work, because a count of titles has no people in it.

Degree requirements are a hiring phrase where that gap has been measured against the people already in the job.

Harvard Business School's Dismissed by Degrees read more than 26 million postings. In 2015, 67% of production supervisor postings asked for a college degree, while 16% of the people actually employed as production supervisors held one ². That pair is one occupation chosen to make the point, not a market average, and part of any such gap can be real change in what the role asks for.

A degree requirement and a job name are not the same object, and the size of that gap forecasts nothing about AI titles. What carries over is the mechanism: an employer writes the words, and the words go out into a market whether or not anybody has checked them against the people doing the work.

But a posting is not a practice, and the degree case is also where somebody went and checked what followed the words. The Burning Glass Institute and Harvard Business School took 11,332 roles at large firms that had dropped a degree requirement and read them against the career histories of roughly 65 million US workers ³.

Inside the roles that changed, hires without a bachelor's degree rose about 3.5 percentage points. Spread across all US hiring the effect came to 0.14 points, because only 3.6% of the large-firm roles studied dropped a requirement ³.

The text moved and the hiring barely did.

So the degree research suggests one thing to carry to the AI titles: hiring language can move on a schedule of its own.

What it cannot show is how far any one AI title sits from the job under it. On the resume, only the bullets can.

The word senior has the same weakness for the same reason. Working out what level a candidate really is turns on the decisions somebody made under ambiguity, and a title carrying AI asks for the same reading.

What Separates a Bullet About the Work From One About the Tools?

Three things separate a bullet written about AI work from a bullet written about AI access: what the model produced, what the candidate did with it, and what would have gone wrong if nobody had looked. A real AI role leaves at least one bullet carrying a judgment made under uncertainty. A renamed role leaves bullets that name software.

The same job, AI Operations Specialist, written up two ways:

• Used AI tools to streamline vendor onboarding and improve turnaround times.
• Cut vendor onboarding from nine days to four using a model-drafted intake summary, and kept a manual check on payment terms after two drafts got the net-30 clause wrong.

The second bullet is checkable and the first is not. Ask the second candidate what the intake summary got wrong and an answer arrives with a shape. Ask the first what streamlining meant and the bullet gives you nothing to hold the answer against.

None of that requires knowing the tool. What actually counts as an AI skill is the decision, and the decision shows through whatever product the bullet happens to name.

A renamed role is not a disqualification.

Somebody whose title changed to AI Content Editor without their work changing has still spent two years next to whatever that company was doing with models. Exposure counts for something at a screen. It counts for less than a decision, and it should not clear the same bar.

Which Mistake Costs More, the Renamed Role or the Missed One?

Dropping a good candidate costs more. Screening on the word in the title finds the group whose titles carry it and loses everyone doing substantial model work under an ordinary name. A title changes when an employer decides to change it. What a job actually involves changes whenever somebody opens a new tab.

Indeed's GenAI Skill Transformation Index rated almost 2,900 skills found in US job postings and mapped them onto more than 53.5 million postings between May 2024 and April 2025. In a typical posting, 46% of the listed skills fall into the two bands the index calls hybrid or full transformation ⁴.

Hybrid means a model could do the bulk of the routine work while a person still has to oversee it, so the figure is not a count of jobs going away ⁴.

The ratings also come from models scoring skill descriptions, with nobody watching work get done, which makes 46% a ceiling on what could change rather than a record of what has ⁴.

Resumes moved too.

Handshake compared what graduating seniors uploaded for the Class of 2026 against the Class of 2022. The Class of 2026 named AI more than nine times as often, and two-thirds of those mentions came from students outside computer science ⁵.

Handshake counted the word appearing on a resume, on one platform, among US four-year students. The vocabulary arrived. Whether the work did is what the bullets are for.

The resume line was never strong evidence in the first place. A meta-analysis of 81 independent samples put the correlation between prehire work experience and later job performance at .06, and experience with tasks, jobs or occupations relevant to the role did no better, at about .07 ⁶.

Those figures are corrected, so the near-zero is not an artefact of under-correction. The measures they cover are the crude ones a screen reads off a resume ⁶, and a job title is read off the same page.

Sometimes the title is the fact, and that is the case against all of this. Where an employer wrote a new job description, hired against it, and the person has held the job two years, the name carries what the organisation asked of them, and a three-bullet resume may offer nothing else to read.

The posting behind the title is the cheapest check while it is still up. What job postings mean when they ask for AI skills varies enough that the description is worth more than the name sitting on top of it.

Ask One Question When the Bullets Won't Settle It

Ask one question, and ask the same one of every candidate in the stack: name a time a model gave you something that looked right and was not, and tell me how you caught it. The answer gives you something to follow up, and it costs four minutes of a screen you were already running.

Ask it of everybody, including the candidates whose titles already say AI. A question put only to the ambiguous ones is a different test given to different people, and it will read that way to anyone who examines the round later.

How informal the step felt is not a question federal selection law asks. Under the Uniform Guidelines, a selection procedure is any measure, combination of measures, or procedure used as a basis for any employment decision ⁷. The definition reaches the full range of techniques, from paper-and-pencil tests through informal or casual interviews and unscored application forms ⁷.

So a standing rule that AI has to appear in the title is a screening criterion, whatever it is written on.

The Guidelines put a validation burden on a procedure only where adverse impact turns up, and they are a 1978 regulation that names no software. What they settle is whether the step counts as a selection procedure at all.

Confirm the compliance question with employment counsel.

Write down what the answer was measured against while the screen is still open. A note saying which claim the candidate checked, and how they caught the error, can be re-read in six months. A note saying strong on AI cannot.

The stack is still sitting there. Whatever share of those titles says AI, that share is a fact about the companies that wrote them.

The bullets underneath are the only place the work shows up. Sort on the bullets and the pile reorders.

Some of the AI titles drop. Some of the analysts and coordinators come up.

See a sample report >