OpenAI: AI helps workers understand salaries | Keryc
Information about salaries guides key decisions: which offers are worth it, when to negotiate, and whether it's worth changing careers. But unlike the price of a product, the price of work is often scattered and hard to interpret — especially if you're starting out, switching industries, or moving to a new city.
Why this matters
Have you ever wondered how much you should be earning for a similar role in another city? Or whether launching your idea would cover the bills? When there aren't clear numbers, decisions get exposed: you might accept less, delay a change, or skip negotiating altogether.
AI is becoming a practical tool to close that information gap. Instead of checking multiple sites, decoding confusing tables, or asking a question that feels awkward, a model can synthesize pay data and give you a reference point in seconds. In fact, in the U.S. people already use ChatGPT for this: nearly 3 million daily messages ask about salaries, compensation, or income.
What OpenAI's research found
OpenAI studied how people use ChatGPT to understand pay and found two main types of queries: turning scattered info into a usable benchmark, and estimating what a role, company, career, or business idea might pay.
Among messages labeled as salary queries, the breakdown was:
26% are pay calculations (for example: converting pay frequency, hours, or bonuses to an annual salary).
19% are about a specific role.
18% relate to entrepreneurship.
11% about a role at a particular company.
11% about occupation or career path.
The research used automated analysis and preserved privacy: no human saw individual messages.
There's another interesting pattern: salary searches concentrate where information is least transparent or most variable. Creative fields, management, health, transportation, and computing & math roles appear more frequently. In these sectors, knowing the right pay is harder — and more important — for career mobility.
WorkerBench: measuring how AI helps
To better understand model performance on work-related tasks, OpenAI introduced WorkerBench, an effort to evaluate ChatGPT on queries useful to workers. In this first version they compared GPT-5.4 with 2024 median wages from the OEWS at national and metro levels.
The result was promising: in the observed sample the model showed wide coverage, small biases, and most numerical estimates landed close to the benchmark.
It's not that AI removes all uncertainty, but it can offer a quick, consistent reference where there used to be noise and guesswork.
What this means for you — and for companies
If you're job hunting, changing careers, or starting a business, a well-trained AI can give you a starting point to negotiate, plan training, or decide to relocate. If you lead HR or run a startup, you know better market signals make fairer pay and more efficient hiring decisions.
Some practical precautions:
Use AI as a reference, not your only source. Combine it with local data and conversations with colleagues.
Ask about ranges and assumptions: does the figure include bonuses, overtime, or benefits? That changes the story a lot.
Consider salary transparency in job offers: when there's clarity, negotiation is fairer.
Final thought
AI is normalizing a kind of access to salary information that once required time, networks, or luck. It's not a cure-all, but it's a real improvement: it helps form more reasonable expectations and empowers work decisions with less friction. Can you imagine how much changes if more people have that same benchmark before signing their next contract?