Pick the Title for the Level You Will Assess Against

How much does the title actually decide?

It decides less about who finds the posting than the standard advice assumes, and more about what happens after they apply. A findability mistake is visible and cheap: applications do not arrive, you notice inside a week, and the words can be changed. Level, comparison class and compensation expectation are the other half of what a title fixes, and none of the three surfaces until the offer or the debrief, by which point none of the three is recoverable.

Title vocabulary is also moving fast enough that a standard title is a moving target. Counting job titles where at least five postings in a quarter carry AI or a related term in the employer's own title text, Indeed Hiring Lab found US AI-touched titles rising from 264 in the first quarter of 2022 to 822 in the first quarter of 2026, which is 8.3% of all titles with enough postings to measure, and 63% of those US AI-touched titles now sit outside tech occupations ¹.

That counts distinct titles, so a title with five postings counts the same as one with fifty thousand, and it measures the spread of vocabulary across role names rather than the volume of demand. The count dipped to 159 in the first quarter of 2023 before it climbed, so the two endpoints are not a straight line. Nothing in it says whether any of those roles assess AI skill. Read as vocabulary, it is still the fact that matters here: the naming conventions you are picking from are being rewritten while you pick.

The older lesson about posting language is that it drifts away from the people actually doing the job. Analysing more than 26 million postings, Harvard Business School's degree-inflation study found that in 2015, 67% of production supervisor postings asked for a college degree while only 16% of people employed as production supervisors held one ⁴. That is one illustrative occupation, a decade old, predating the retreat from degree requirements that followed, and the two figures come from different data: posting text on one side, incumbent workers on the other. Part of any such gap is the role genuinely changing. What survives the caveats is the shape of the problem: the document and the job are separate things, and only one of them gets edited.

Pick the words candidates use, then pick the level

Pick the words first: the function, in the vocabulary somebody would type or say, with no internal jargon and no invented seniority. Then decide the level word on its own terms, before the posting is drafted around it. Editing the words mid-search costs a repost; editing the level costs the search, and that asymmetry is the whole argument for spending the thinking on the level.

A workable rule for the string itself:

• Function first, in plain words. Financial Analyst, not Growth Catalyst. Support Engineer, not Customer Happiness Architect.
• One level word, chosen on purpose. Senior, Staff, Lead, Head of. If you cannot say what the level word rules out, remove it.
• No internal codes. Analyst II is a code from your compensation system, and a candidate reading it has no key.
• Put the differentiator in the first line of the posting. Whether the role is remote, what team it sits in, which product it owns: all of that reads better as a sentence than as four extra words in a title string.

The level word is where founders and first-time hiring managers most often improvise, usually generously, because a bigger title feels free at the moment of writing. What it actually does is set what the interview loop has to be able to test, and a loop built for a mid-level hire will pass a senior title's worth of candidates without ever asking a senior question. If the posting also carries an AI requirement, the level decision drives that too, which is most of whether the job post should require AI experience at all.

Why an inflated title bills you at the offer

Because the candidate priced the job from the title before they ever spoke to you. A title is the first and often only compensation signal in a posting without a salary range, and it is anchored against every other posting carrying the same words. By the time the offer goes out, the argument is not about your budget: it is about a number the candidate assembled six weeks earlier from a title you chose in a hurry.

The pricing signal is measurable at the posting level. Lightcast reports that advertised salaries in postings mentioning AI skills run 28% higher than in postings that do not, roughly $18,000 more a year, and that in 2024 over half of postings requesting AI skills were outside IT and computer science ². That is a raw comparison between two groups of postings rather than a like-for-like wage estimate, AI-mentioning postings skew senior and urban, and Lightcast sells skills data. It measures what employers advertise rather than what anybody is paid.

What that supports is narrow: employers themselves attach a number to the words in a posting, so any title sits in a market of titles already carrying advertised prices. The candidate reading yours has that market and, without a range, very little else to price from. A title chosen to attract applications is therefore a commitment made before any evidence about the candidate exists.

There is a quieter version of the same bill, and it comes due at the debrief. A title sets the comparison class: interviewers assessing a Senior anything compare against the seniors they know, and a candidate hired into an inflated title spends their first quarter being measured against a peer group nobody ever assessed them against. The bill for that one arrives at the first performance review, twelve months after the title was chosen.

Check that one stage tests the level difference

Write the title, then write one sentence naming what this person will own that somebody one level down would not, then find the stage that tests that sentence. If no stage tests it, the level word is a claim the process cannot support, and the loop will hire a strong candidate at whatever level they happen to be and call it a match.

The sentence is usually about scope of judgment. A senior analyst is not a mid-level analyst who works faster; the difference is which calls they make without checking, and what they do when the data disagrees with the request. That difference is testable in about twenty minutes with a brief that leaves a real decision open, and it is exactly what a resume cannot show you. The same distinction has to be read back off a candidate later, which is how to work out what level a candidate really is when every title in the pipeline says senior.

Hiring at the junior end of AI-exposed work has contracted. Using administrative ADP payroll records covering 3.5 to 5 million employees monthly from January 2021 through June 2026, Brynjolfsson, Chandar and Chen report that employment of workers aged 22-25 in AI-exposed occupations sits 19% below where it would be had it kept pace with less-exposed peers, with no comparable gap for experienced workers, and that the gap opens through reduced hiring rather than increased separations ³.

The authors are explicit that these are descriptive indicators rather than causal estimates of AI, the 19% is the descriptive figure at the June 2026 vintage while other figures from the same paper circulate at earlier ones, and the sample overrepresents larger firms and AI-exposed work. What it supports is narrow and relevant: the junior end of some occupations is being hired into differently than it was, so a title that described this work three years ago may now describe a job with the same name and a different floor. That is the argument for re-levelling a job whose work has shifted before the title is written, and for setting what you expect of a junior, mid and senior hire in the same sitting.

Do the sentence test on the requisition currently open. If you cannot name what the title's level word rules out, take the word off and let the posting describe the work instead.

See a sample report >