5,540 Comments in 7 Hours. 227 of Them Were a Crypto Scam.

MKBHD posted his iPhone 18 Pro review and 5,540 comments arrived in 7 hours 50 minutes. 227 of them were a crypto scam wearing the costume of his real audience, and four of those scam comments now sit in his top 50 by likes.

We pulled every comment and reply through YouTube's official Data API and read them the way NAWA reads a connected channel. This is what a comment section that size is actually made of, and what it costs a creator to never look.

Half the comments arrive before you have finished lunch

383 comments landed in the first ten minutes. 1,484 in the first hour, 27% of the total. By 2 hours 23 minutes, half of all 5,540 were in.

Most creators know this part. The part they underestimate is the tail: the last hour before we pulled the data still brought 280 comments. The rush ends, the thread does not. If you reply for twenty minutes after publishing and then close the tab, you have answered the loudest hour and missed the other six.

The scam wave is not random, and it is not stupid

227 comments were a coordinated crypto push: 183 promoting "$Grokvoxus" and 44 promoting "Sairium", across 223 different accounts.

Here is the part worth sitting with. 71 of them copied a real comment from the same video and appended the pitch. They are not obvious spam that pattern-matching catches. They are a real viewer's words with a coin name bolted on.

“@joelconolly5574**, 7 minutes in, 221 likes: "Matte black intro. Like a businessman in a suit."”

>

“A scam account, 5 hours 10 minutes in, 26 likes: the same words, plus "I called $Grokvoxus move months ago. Now everyone in there"”

The "$Grokvoxus" comments carry 3,015 likes between them, with a median of 15 each. Real comments from those same hours have a median of zero. Four of them now sit in MKBHD's top 50 comments by likes.

They arrived in three waves, all of them after the opening rush had died down: 107 comments between 2h07 and 2h49, 100 between 5h00 and 5h54, and 19 between 7h18 and 7h47. The last one landed 7 hours 47 minutes in, which means the wave was still running when we stopped counting.

Why this matters beyond tidiness. If you ever read your own comments to find out what your audience thinks, and 227 of the loudest ones are a coordinated promotion with inflated likes, you are reading a manipulated sample. The fix is not moderation for its own sake. It is knowing which comments are people.

Ten comments hold 82% of the likes

One line, "Of course, he had it for a year," took 11,708 likes. That single comment holds 31% of every like on the video's real comments. It also drew 98 replies, making it the busiest thread on the video.

The top ten comments hold 82% of the likes between them. Here is what they have in common:

Four of those five are not about the phone. They are about the viewer: what they can afford, what they are waiting for, what they will do next. The most technically substantial comment in the list sits fifth.

This is the gap between what gets talked about and what gets liked, and it is bigger than most creators assume.

Spec talk brings volume. Short, personal lines bring the likes. 79 comments about waiting a year earned thirteen times the likes of 277 comments about the camera.

107 people are correcting you about the same thing

107 comments named earlier phones with a variable aperture, saying the review credits only Samsung. They name Huawei's Mate 50 Pro, the Xiaomi 13 and 14 Ultra, and the Nokia N86. The 1:30 mark is the third most-cited moment in the whole video.

“@Muchoki_J**, 45 likes: "Suddenly everyone has forgotten about the variable aperture in Xiaomi 14 ultra from 3 years ago"”

A hundred people raising the same factual point is not a pile-on. It is a single question asked a hundred times, and one pinned reply answers all of it. Left alone it becomes the comment section's consensus that you got it wrong.

The same shape shows up elsewhere in the data. 67 comments mention the modem, the most-liked saying the US Pro Max still uses a Qualcomm chip rather than the C2. Six viewers, the last of them 6 hours 51 minutes in, still could not find the chip video link, even after the description was updated.

What a creator would actually do on Monday

The audit ends with five things in order, which is the part that makes it worth reading rather than admiring:

• Hold the scam wave. Add "Grokvoxus" and "Sairium" to Blocked words in YouTube Studio so new ones wait for review, then remove the 227 already posted.
• Pin the chip video link. Six viewers still could not find it hours after the description was fixed.
• Settle the modem question. One pinned line answers 67 comments.
• Acknowledge the variable aperture history. A short reply turns a correction thread into goodwill.
• Answer the quick ones. Three viewers asked which game is at 8:30 and which weather app is at 8:48.

Your comments are a content brief you already paid for

The last section of the audit ranks eight videos the comments are asking for, by the likes on the comments behind each one. The top one, "Upgrade now or wait for the 20th anniversary iPhone?", carries 7,588 likes across 236 comments from 231 different people.

That is not a guess about what an audience wants. It is 231 people saying it, weighted by how many others agreed.

How this audit was made

On 16 September 2026 at 19:51 UTC, 7 hours 50 minutes after upload, we pulled every comment and reply on "iPhone 18 Pro Review: All About that Chip" through YouTube's official Data API: 4,445 comments and 1,095 replies. YouTube showed 5,541; one was not returned. The video had 2,749,435 views and 84,377 likes at the time.

Scam detection, the timeline, the cited moments and every count were computed in code over all 5,540. A scam wave counts when many accounts push the same "$name" in comments that copy real ones, or arrive in bursts with unusual likes, or flood in with presale wording within minutes. Themes are keyword-matched, so a comment can count under more than one.

NAWA's report engine read the 500 most-liked comments to propose the video ideas, with every commenter's name replaced by a label before the AI saw anything. Names are put back only on your own screen. That is not a detail we added for the audit. It is how NAWA works on every connected channel.

What this looks like on your channel

This audit took one video and one API pull. On a connected channel, NAWA does it continuously: new comments every 30 minutes, sorted so a question never sits under a joke, coordinated promotion counted and set apart rather than quoted back to you as a fan, and replies drafted in your voice for you to approve, edit or skip.

We are running these audits for creators one at a time, on a real video, before anyone signs anything.

The whole thing is in the PDF: the scam waves, the hour by hour timeline, every cited moment, and the eight videos this comment section is asking for.