Partly happeningReal, but smaller, slower or narrower than claimed.
HappeningStrong evidence that it is real now.
Several studies find fewer young people being hired into jobs where AI can do much of the work. But it is mostly slower hiring, not firing, the cause is not proven, and remote work explains part of it.
Still unknownResearchers cannot yet separate the effect of AI from other causes like remote work and a slow job market, and almost all the data come from the United States.
Early in your career, choose teams that train you on the job, because cuts hit entry-level hiring first.
Learn to use AI tools well in your field and show employers what you can do with them.
If you hire, keep some junior roles, or you will have no experienced staff later.
Evidence checked October 14, 2026.
“AI will cause mass unemployment within a few years”
HypeThe evidence contradicts it.
Not yetNo clear evidence so far.
Partly happeningReal, but smaller, slower or narrower than claimed.
HappeningStrong evidence that it is real now.
So far, researchers find no clear sign that AI is raising unemployment, even in the most exposed jobs. AI is changing tasks inside jobs, but there is no sign yet of the 10–20% unemployment some tech leaders predict.
Still unknownAI tools are improving fast, so effects could grow, and the studies so far cover only about three years, mostly in rich countries.
Treat big job-loss forecasts as guesses and look for measured data.
Find the tasks in your job that AI already does well, and spend more of your time on the ones it does poorly.
Keep learning new skills, since most jobs are expected to change rather than vanish.
Evidence checked October 14, 2026.
“AI data centres use as much electricity as whole countries”
HypeThe evidence contradicts it.
Not yetNo clear evidence so far.
Partly happeningReal, but smaller, slower or narrower than claimed.
HappeningStrong evidence that it is real now.
All data centres together already use far more electricity than a whole country like Ireland, and by 2030 may use more than Japan; AI is the fastest-growing part. Still, they use only about 1.5% of the world's electricity.
Still unknownTech companies do not publish full energy data, so the share used by AI alone is an estimate, and forecasts for 2030 and later vary widely.
When you see a country comparison, check whether it is today or a forecast, and whether it means all data centres or AI only.
Ask your AI or cloud provider for energy and emissions figures for the services you use.
Follow local planning decisions, because the strain is biggest where data centres cluster.
Evidence checked October 14, 2026.
“Using AI chatbots makes people worse at thinking”
HypeThe evidence contradicts it.
Not yetNo clear evidence so far.
Partly happeningReal, but smaller, slower or narrower than claimed.
HappeningStrong evidence that it is real now.
Studies show that letting a chatbot do the work can weaken learning, memory and skills, at least in the short term. There is no proof yet that it lowers general thinking ability, and guided use can help people learn.
Still unknownMost studies are small, short or based on self-reports, and no study has yet tracked whether years of chatbot use change how people think.
Try a task yourself first, then use AI to check or improve your work.
When learning, ask AI for hints and questions instead of full answers.
Read and question AI answers before you use them.
Evidence checked October 14, 2026.
“Deepfake scams are costing people real money”
HypeThe evidence contradicts it.
Not yetNo clear evidence so far.
Partly happeningReal, but smaller, slower or narrower than claimed.
HappeningStrong evidence that it is real now.
Victims told the FBI they lost $893 million in 2025 to scams involving AI, and one company lost HK$200 million after a single deepfake video call. The true total is likely higher, because many victims never notice the AI.