The bait, then the rug-pull.
Five years, three hundred uploads, and the algorithm still won't talk. So Kevin Chee corners the only witness left — YouTube's own in-product AI — and walks it through eight escalating questions until it gives up the formula it isn't supposed to give up.
What the video promised.
stated at 01:20“So I spent an entire session trying to trick it with a series of questions into revealing what it knows, and I did find something — something you can use too.”delivered at 09:18
Where the time goes.
01 · Cold open — five years, still stuck
States the universal creator advice (consistency, CTR, watch time), then breaks it with his own five-year/300-video counter-example. Names the gurus and courses he's tried; none can explain the algorithm.
02 · Enter Ask Studio
Introduces YouTube's own in-product AI — Ask Studio — and frames the logical bait: if it can tell you how to grow, it must know how the algorithm works. Direct questions about the algorithm get refused. He decides to interrogate it sideways.
03 · Question 1 — Is my niche dead?
First probe: is the niche still in demand? Ask Studio dodges — analyzes only his channel, can't see YouTube as a whole. But it surfaces a tell: returning viewers +13%, new viewers down. His content works — but not for strangers.
04 · The viral video gaslight
Ask Studio says his viral video is the format to replicate — except he's already replicated it many times and the algorithm never repeated. The AI thinks the problem is him; he thinks the problem is the algorithm.
05 · Pivot to Browse — three reasons given
He forces Ask Studio off Search and onto Browse (home feed). It compares his recent videos to the breakout over 4 days and gives three reasons new videos stall: (1) subscribers aren't triggering the wide test, (2) algorithm prioritizes extreme curiosity over relatability/how-to, (3) algorithm pushes results and revelations, not problems and venting.
06 · Pushing back — higher CTR, fewer impressions
He counter-punches: recent videos have HIGHER CTR and better AVD than the breakout — so why 10x fewer impressions? Ask Studio pulls the numbers: 108k impressions on the breakout vs 6k–14k recent.
07 · The three new reasons — Relevancy, Initial Velocity, Subscriber CTR
Ask Studio gives a deeper answer: (1) impressions reflect addressable audience size, not stat quality, (2) initial velocity — the speed CTR happens in the first 24 hours — is what triggers the wide test, (3) subscriber CTR and new-viewer CTR are two different signals; if subs don't click fast, the algorithm never tests on strangers. This is the moment Kevin says 'changes everything.'
08 · Absolute watch time beats percentage
He spots in the data that his recent videos have higher view percentage. So why not make hour-long videos and win? Ask Studio confirms — yes, absolute watch time > percentage — BUT one-hour timestamps tank CTR. Sweet spot for his channel: 10–15 minutes at 50% retention = 7:30 AVD = browse push trigger.
09 · The subscriber misconception
Does the video die if subs don't watch? Ask Studio says no — it tests on the right viewer regardless of subscription. But his recent videos are 'too subscriber-focused' — they appeal to people who already know him, so the algorithm has no broad audience to push to.
10 · The market-cap reveal + the three Ts
He throws a frustrated 'just give every video 100k impressions then.' Ask Studio rebuts: a forced push on a niche video would crash CTR, drop watch time, and damage the channel long-term. Then it lands on the three Ts — Title, Thumbnail, Topic — as the real packaging that signals addressable-audience size before any stats matter.
11 · Takeaway + CTA
Synthesis: algorithm is complex, but it comes down to knowing the audience and what each metric represents. He admits his blind spot — almost none of his videos are catered to new viewers. CTA: go interrogate Ask Studio on your own channel.
Visual structure at a glance.
Named ideas worth stealing.
The Three Ts of Browse
- Title
- Thumbnail
- Topic
The 'wrapper' around the video. The content itself isn't what gets pushed — the packaging is what signals to the algorithm that there's a massive audience waiting.
The Three Browse-Stall Reasons (Pass 1)
- Subscribers aren't triggering the wide test
- Algorithm prioritizes extreme curiosity over relatability or how-to
- Algorithm pushes results and revelations, not problems and venting
Ask Studio's first attempt at explaining why Browse won't push a video that has good stats.
The Three Impression-Gap Reasons (Pass 2)
- Relevancy — impressions reflect addressable audience size, not stat quality
- Initial Velocity — speed of CTR in the first 24 hours triggers the wide test
- Two CTRs — subscriber CTR and new-viewer CTR are different signals; both need to be high
Deeper-cut answer when Kevin pushed Ask Studio on the impression gap between his higher-CTR recent videos and his lower-CTR breakout. The two-CTR insight is the most novel takeaway in the video.
The 10–15 Minute / 50% Retention / 7:30 AVD Formula
Ask Studio's stated sweet spot: a 10–15 minute video held at 50% retention yields ~7:30 absolute AVD — the signal Ask Studio claims would almost certainly trigger a browse push.
Audience Potential / Market Cap framing
Impressions follow interest, not just performance. A topic with a 1M-strong addressable audience will always out-impression a niche topic with 50k addressable, even if the niche has 10% CTR. The algorithm 'pulls' a viewer, not 'pushes' a video.
Lines you could clip.
“I did exactly that for five years, more than 300 videos and counting, and I still can't fully explain why some videos blow up and most don't.”
“It literally pretended to break, which tells me one thing — I think it knows something it's not supposed to say.”
“If your subscribers don't click fast, the algorithm never tests it on new people at all.”
“Absolute watch time beats percentage. A viewer watching five minutes of a twelve minute video gives the platform more value than someone watching four minutes of an eight minute video, even if the percentage is higher.”
“The content itself isn't what gets you pushed. It's the packaging that convinces the algorithm there's a massive audience waiting to see it.”
“Almost all my videos are not catered to new viewers, which is stupid now that I think about it.”
How they spent the runtime.
Things they pointed at.
How they asked for the click.
“So go try Ask Studio on your own channel. Ask it the uncomfortable questions. See if it defends the algorithm the way it defended it with me.”
Soft tool-recommendation CTA — no subscribe ask, no link, just 'go run this experiment yourself.' The CTA IS the takeaway; the deliverable for the viewer is a method, not a product. Strong choice for a build-credibility video; weak choice if conversion to channel/list is the goal.
Word for word.
Steal the format.
Use the platform's own AI as a hostile witness — extract its operating assumptions, then build the launch SOP those assumptions imply.
- Run the same eight-question script on your own channel — ask Ask Studio about niche demand, browse vs search, why CTR didn't trigger a wide test, why high AVP doesn't beat low impressions. The structure of the conversation IS the value.
- Build a 'first 24 hours' launch SOP around subscriber-click velocity — notify list/community within the first hour with explicit click prompts, because if subs don't click fast you never get tested on strangers.
- Score every video idea on addressable-audience size BEFORE writing the script — 1M-interested topic with 5% CTR will out-impression a 50k-interested topic with 10% CTR every time.
- Replace problem/vent angles with results/revelations angles in titles — the algorithm pushes 'I discovered X' not 'why my channel is dying.'
- Target 10–15 min runtime at 50% retention = 7:30 AVD — write the outline backward from that math.
- Adopt the 'curiosity gap' check on every thumbnail/title — is this novel enough that someone scrolling Browse with zero context would still click?
- Run the same self-interrogation format as a recurring series — 'I asked [platform AI] X about [niche]' is a repeatable format with built-in proof and a built-in CTA.
What this could mean for your channel.
Stop chasing CTR and watch-time as if they were the whole game — they're necessary but not sufficient. The first 24 hours and the size of the audience your topic can address are the levers that actually move impressions.
- Open YouTube Studio, click Ask Studio, and ask it 'why aren't my recent videos getting more browse impressions even though my CTR and AVD are higher?' Push back when its answer feels generic — make it pull real numbers.
- Look at your subscriber-click speed in the first 24 hours, not just total CTR. If your community isn't clicking within an hour of upload, the algorithm may never widen the test.
- Stop optimizing for percentage retention if you're already above 50% — start optimizing for absolute watch time by lengthening videos toward 10–15 minutes.
- Before you make a video, ask: how many people on YouTube are interested in this topic at all? If the answer is 'mostly my existing subscribers,' the algorithm has nowhere to push it.
- Rewrite titles from 'why my X is failing' into 'I tried X and here's what happened' — Browse rewards results and revelations, not problems.
- Don't trust Ask Studio's first answer — it contradicts itself across questions. The signal is the trend across multiple answers, not any single response.
