AI Analyzes Salaries to Guide Workers

ChatGPT receives almost 3 million messages a day in the United States about salaries, compensation, or income. OpenAI analyzed this usage to understand how workers look for pay benchmarks and what kinds of decisions they try to make with that information.

Salary influences whether you accept a job, negotiate an offer, switch industries, or invest time and money in training. The problem is that, unlike the price of a product, the value of work is usually scattered across job postings, reports, and conversations that are difficult to have.

AI can bring those pieces together and turn them into a reference point in seconds. It does not eliminate uncertainty, but it can help you determine whether an offer is within a reasonable range or how much a profession might pay in another city.

How workers use ChatGPT

According to OpenAI’s new report, the queries analyzed focus on two needs: turning a salary figure into a useful reference and estimating how much a position, company, career path, or business idea might pay.

Among messages tagged as salary benchmarking queries:

• 26% ask to calculate or translate a pay figure.
• 19% ask about a specific position.
• 18% concern entrepreneurship.
• 11% ask how much a specific position pays at a company.
• 11% ask about a profession or career path.

OpenAI says it obtained this data through an analysis designed to protect privacy. It used automated classifiers and did not have people review individual messages.

More interest where salaries are harder to read

Salary-related searches appear especially often in arts, design, entertainment, sports and media; management; healthcare; transportation; sales; and business and financial operations.

Relative to existing employment, demand is higher in skilled and less transparent occupations, such as creative work, management, healthcare, and computer and mathematical professions. These are areas where pay can vary widely, be negotiated frequently, or depend heavily on experience and location.

Questions about entrepreneurship also stand out, especially in creative work and small service businesses. In these cases, there is usually no clear salary table to use as a starting point.

The pattern repeats across industries: people seek more information when salaries are more dispersed and when getting the estimate right has greater consequences. A mistaken benchmark can lead you to stay in a lower-paying job, negotiate below a reasonable level, or reject a career change that could actually work for you.

OpenAI tests the model’s accuracy

The report also introduces WorkerBench, an initiative to evaluate whether ChatGPT can handle tasks relevant to workers. In its first test, OpenAI compared GPT-5.4 responses with 2024 median salaries from the OEWS program, an official United States database that provides data by occupation and metropolitan area.

In the sample observed, OpenAI describes the model’s performance as highly accurate: it had high coverage, low bias, and almost all numerical estimates were close to the reference used.

That does not mean a ChatGPT response guarantees what you will earn. Actual pay depends on factors such as the city, level, company, experience, benefits, and negotiating ability. The evaluation was also conducted using a specific sample and United States data.

For you, the practical change is simpler: you can use a conversational tool as a first filter before applying for a job, responding to an offer, or deciding whether training is worth the cost. But it is worth checking the result against official sources, real job postings, and data specific to the market where you want to work.

OpenAI wants to expand these benchmarks beyond the national average salary and respond better to the questions workers actually ask: how much a specific company pays, what changes by city, how professional level affects compensation, and how much benefits are worth. The important question will be whether those answers remain accurate when they move from a general figure to a specific job negotiation.