AI SEO, tested in public
Most advice about AI search is asserted, not tested. Someone notices that ChatGPT cited a page, invents a reason why, and it hardens into a "best practice" that gets repeated for a year — never checked. HZ exists to do the opposite: run actual experiments on what makes AI engines cite content, and publish everything — the hypotheses, the data, the wins, and the failures.
This isn't a brand blog dressed up as research. Every claim here is something I've either tested myself or sourced and labelled with how confident I am in it. Where I don't know, I say so.
How I work
Every experiment on this site follows the same rules:
- A hypothesis, written first. Before I run anything, I state what I expect to happen and how I'll decide if I'm wrong — so I can't quietly move the goalposts once I see the data.
- Real data, shared openly. The numbers go on the experiments log, not just the conclusions. You can check my work.
- Confidence, labelled. I mark each claim high, medium, or low confidence instead of stating everything with false certainty. AI search is a fast-moving, under-evidenced field, and pretending otherwise would be dishonest.
- Wins and failures both. A null result is still a result. Most of what circulates online is survivorship bias — the things that happened to work, with a story bolted on. HZ publishes what didn't work too.
- Kept current. AI engines change monthly. I date things, and I update or retract when the evidence shifts.
Who's behind it
I'm Haobo Zhang. I'm the co-founder & CEO of Polarbear, an Antler-backed AI-native edtech used by 40,000+ students, and the founder of HZ Tutoring, Melbourne's #1 VCE tutoring company. I've been building businesses since I was 14.
HZ Signal is where I work out, in the open, how to actually use AI SEO and GEO to drive traffic to them — separating the facts from the folklore.
That's what makes this different from most AI-SEO writing: I have real sites on the line. These experiments aren't hypothetical — the traffic they're chasing is mine, and so is the cost of getting it wrong.
More about me: haobozhang.org · LinkedIn · GitHub
What you'll find here
- Experiments — the live log of what I'm testing and what's survived contact with the data.
- Blog — the write-ups, with the numbers.
- Glossary — plain-English definitions of the AI SEO terms worth knowing.
- Subscribe — one short email per experiment: the finding and the data, nothing else.
A standing promise
No affiliate-pumping, no "ultimate guide" filler, and no claiming certainty I haven't earned. If an experiment fails, you'll read about it here first.
Find your signal.