Can AI tell you what you should be paid?

AI can produce a salary figure in seconds. But comparing CEO pay is more complicated than the number suggests.

This article was sponsored by Google Trends.

When people are negotiating their salary — or trying to decide what they should offer a new employee — a simple Google search can seem like an obvious place to start.

But increasingly, there is another option: ask AI.

Ask what a CEO in the not-for-profit sector should be paid and a chatbot can quickly produce a salary range, explain its reasoning and provide comparisons with other roles.

The problem is that a more detailed answer does not necessarily mean a more useful benchmark.

To see what people might actually encounter, we asked ChatGPT, Gemini, Claude and Copilot the same four questions about CEO remuneration in Australia, using fresh conversations.

We asked:

• What is the average CEO salary in Australia?
• What is the average CEO salary for an Australian not-for-profit organisation?
• What is the typical salary for the CEO of an Australian charity?
• How does CEO remuneration in Australian charities and not-for-profits compare with CEO remuneration in the broader Australian market?

The systems produced different headline figures and drew on different sources and benchmarks.

Graphic 1 — Google Trends search interest in CEO salary-related queries.

Graphic 2 — The four AI outputs from the controlled experiment.

Graphic 3 — Verified salary benchmarks and why the figures differ.

Why the numbers differ

Average and median are not interchangeable. The same applies to base or fixed remuneration and total remuneration.

Corporate executive pay can look very different depending on the measure. ACSI reported median ASX100 CEO fixed pay of A$1.83 million in FY2025, compared with median total realised pay of A$4.80 million.

Who is being compared also matters. A CEO of a small charity is not necessarily comparable with the CEO of a large national organisation. Organisational scale, sector, budget, workforce, location and role complexity can all affect remuneration.

And the year matters. A perfectly accurate salary figure from 2023 may not be the most useful benchmark for a salary negotiation in 2026.

An answer can therefore be based on accurate data and still be the wrong answer for the person asking the question.

What does the NFP data tell us?

The Pro Bono Australia 2026 Salary Survey provides NFP-specific context that generic salary searches cannot necessarily provide. It reports a 9 per cent increase in median NFP CEO base salaries between 2025 and 2026, alongside an 11 per cent increase in total remuneration. It also highlights variation across sector, organisational budget, employee numbers and location.

That context matters because the useful comparison is not simply another organisation with the same job title. It is a comparable role at an organisation of similar scale and type, using the same definition of remuneration.

The Salary Survey recommends considering the closest role, the level of the position and organisations of similar size, alongside the market median and mean, individual performance, organisational capacity to pay and the need to attract the right person.

AI can start the search. It can't define the benchmark.

AI can bring together information from multiple sources quickly and present it in a way that feels tailored to the question. But when it produces a salary figure, it is worth asking: What year is the data from? What population does it represent? Is it an average or a median? Does it refer to base, fixed or total remuneration? Is it actually based on NFP organisations? Is the organisation being benchmarked comparable?

Those questions matter because some figures from our experiment could be independently verified, while others relied on secondary sources that could not be treated as established benchmarks.

That doesn't make AI useless for salary research. It makes verification more important.

The number is only useful if you understand what it means

AI systems may converge on a figure because they retrieve the same underlying source. They may also produce different figures because they prioritise different datasets. Neither situation, on its own, tells us whether the figure is an appropriate benchmark for a particular role.

For someone negotiating a salary — or an organisation deciding what it can offer — the most useful role for AI may therefore be to help find and interpret information, rather than provide the final number.

Google Trends can show us what people are looking for. Search can help us find the published evidence. AI can help us explore and compare information.

But salary benchmarking requires another step: understanding whether the data is current, comparable and relevant to the role and organisation.

Because the real question isn't simply:

“What does a CEO earn?”

It's:

“Which CEO, at what organisation, measured how?”