When Human Replies Look Like AI - Real YouTube Exchange

[BikeGremlin]

This is a lightly anonymised transcript of a real comment exchange on one of my YouTube videos. I'm posting it here as a companion to the article:
SEO, LLMs, and the Death of the Open Web?
The conversation perfectly illustrates one of the main points: how even clear, structured, human-written replies are now mistaken for robotic or AI-generated responses. It's a small, ironic case study of the broader trend.

The video:
"Cable and housing swap - and tool review announcement"

Comment discussion transcript (anonymised and slightly edited for clarity):

For context, I reflected on this kind of interaction (and others) in a blog post:
Section 2 - Nuance in Public Discourse

Viewer 1: Isn't better to put a little oil in wires? By the way, thanks for the video. I've noticed a lot of mistakes with my gears. ❤

BikeGremlin: Hi,

First of all, thank you for reaching out to a real person rather than a robot.

My name is Relja – a human being, with all the abilities, but also the limitations, of human capacity – and with a sincere willingness to help you, as well as anyone else interested in this field.

Secondly, I would like to apologise if this complicates matters for you personally. However, based on 15 years of experience (and answering several thousand questions), I have concluded that the greatest benefit for humanity is if I reply exclusively on the BikeGremlin forum.

I kindly ask you to take a little time to search the forum – and if you do not find an existing answer, to create an account and post your question. I will gladly reply to you, to the best of my knowledge, with the hope that it will be useful to humanity as a whole.

If needed, instructions on how to do this are available here:https://www.bikegremlin.net/threads...t-a-comment-question-on-bikegremlin-forum.24/

The reasons for this decision are multiple:

• Poor search functionality.
• Quality of answers and deletion of my posts.
• Destruction of knowledge.

As these issues affect me personally, I have written a more detailed explanation here:https://www.bikegremlin.net/threads/why-i-answer-questions-on-bikegremlin-forum-only.497/

Thank you for your understanding – and in these times when every second is precious, I truly appreciate the fact that you have taken the time to read this message to the end.

Viewer 2: Your comments are AI enchanted?

BikeGremlin: No. Though LLMs tend to copy good writing styles (like mine ).https://bike.bikegremlin.com/17558/bikegremlin-forum/#qa

Viewer 2: I unfollowed already.

BikeGremlin: Fair enough.The irony is that I put a lot of time and effort into writing that canned response.

Perception:"Hmm... no way a human could be this polite and clear. Must be AI-enchanted!"

LOL.Even more irony is that I'm in the middle of writing an article on that topic (how AI is affecting the Internet in general).

Relja (A.K.A. Mr Robot - apparently)

Viewer 2: So you wrote it? If you copy and paste it to every answer it doesn't look good. I follow you again.

BikeGremlin: I plan to copy-paste it to every question, for reasons explained in those links.I also wrote (first) a version in my native for the videos and website in my native.

Forum makes it possible to provide better responses and build a knowledge base.YouTube and other social networks are designed to steal time and attention if not used carefully.

Viewer 2: Wow, we're still talking about this. YouTube commenters need personalized answers or nothing at all. If you copy and paste the same answer to every question to make your forum more lively it just doesn't look good. Maybe put the forum link in the description and turn comments off. I follow a very good UK based fisherman who has turned his comments off. It really doesn't matter to me. At least I get to watch the video for free. But if he's constantly advertising his forum when I ask a question, that would make me a very unhappy person.

BikeGremlin: You are missing the point - my goal is not to make the forum more lively.It is also not to prevent people from commenting (and replying to each other - which can also be good).This is my goal:https://www.bikegremlin.net/threads/why-i-answer-questions-on-bikegremlin-forum-only.497/#post-2075

Edit:I've just realised we are going in circles. My initial canned response is my best attempt at explaining what my goal is.Not everyone will understand or agree with it - and that is OK.

Viewer 2: I hope you find the way to balance the needs of social networks where you do your work and the place where you do your downtime. If I were you and didn't have the time to answer every commenter's question, I would just turn comments off. But remember, it's the comments section which makes or breaks your video viewership.

Viewer 2: OK. I will settle in that. It was nice to talk to ya. Happy summer!

BikeGremlin: I am afraid you are still missing the point. Maybe because you didn't take the time to read my initial reply - and the provided links. They aren't short, so it's understandable, but I couldn't explain it more briefly, so it is how it is.

Briefly put: comments are good even if I don't offer full replies there - at least for many viewers and for me.

These comments for example have been quite interesting and educational: how hard work with best intentions can be interpreted as AI slacking off (ironic, unfair, but that's the world we live in today).

There must be thousands of people who see it like this (most people don't bother to comment, which is also fine).

I'll try to discuss this phenomenon in the article I'm currently working on ("SEO, LLMs, and the Death of the Open Web?" is the working title).

Relja

Viewer 2: I am exhausted. I will trust my first judgement and I will stand to every word I wrote. Maybe get a kitten or a small dog (maybe big), but I'm done with this.

Viewer 1: Tf is going on here i just asked a question

"This... is no ordinary YouTube channel..."
- Narrator, probably Morgan Freeman

Let me know what you think or if you’ve experienced something similar.

Relja

 

[BikeGremlin]

Now I am the one suspecting a text (an e-mail and a website in this case) to be "made by AI":

AI spam 3.0 – can we tell?

 

[HostingSift]

Now I am the one suspecting a text (an e-mail and a website in this case) to be "made by AI":

AI spam 3.0 – can we tell?

Hi, HostingSift here.

Not a spam bot, just a developer who's not great at writing emails and used AI to help with the wording. The outreach was real though.

The site itself doesn't mislead anyone, it's just a tool that aims to be useful and transparent when picking a hosting provider. The whole idea is built around treating every provider equally, no rankings you can buy, no special placement.


That said, you mentioned Bluehost gets "listed and recommended" somehow. Where exactly? If something gives that impression I'd want to fix it.

No hard feelings about the post.

 

[BikeGremlin]

Hi,

Thank you for taking the time to comment. I can confirm that the email address checks out so it is the same "entity" that I got the original email from.

Hi, HostingSift here.

Not a spam bot, just a developer who's not great at writing emails and used AI to help with the wording. The outreach was real though.

The reach was 100% spam - unsolicited marketing email. Cleverly done to avoid getting blocked by spam filters, but spam nonetheless.

Based on my experience from the previous decade, I am also convinced that the mail was both mass-sent and, as you admit, AI-written.

The site itself doesn't mislead anyone, it's just a tool that aims to be useful and transparent when picking a hosting provider. The whole idea is built around treating every provider equally, no rankings you can buy, no special placement.

That said, you mentioned Bluehost gets "listed and recommended" somehow. Where exactly? If something gives that impression I'd want to fix it.

No hard feelings about the post.

On the site, I browsed the "Blog" section, looked at the listed hosting offers, and clicked on the "Take The Quiz" button. When I filled in the Quiz data based on my priorities and needs, Bluehost was recommended as the top pick.

Anyone who knows anything about hosting knows that Bluehost is a scam that pays the highest affiliate rates and that is how it gets promoted to the uninformed people. It is the only corporation I openly criticise:

• Affiliate frauds and Bluehost
• Who owns Newfold Digital (incorporated EIG) - Clearlake Capital owned behemoth

That is why I would not feel comfortable recommending a friend or a relative to go to your site and look for hosting advice. My site, on the other hand, if I do say so myself, is what I am 100% comfortable with my own son using for information and being sure that it is pretty damn good and correct (though not perfect).

It is also fair to note that the site looks like it is set for including affiliate links (colour me skeptical).

It definitely is an impressive feat of AI tool use and web design. But it is harmful, not helpful. To sum it up:

• AI-scraped data to make the site.
• AI-written "blog" section ("for SEO").
• AI-scraped data to contact webmasters for marketing.
• AI-written marketing email pitch.
• Mass-sent (probably) spam emails.

This has become the norm. 90% of the newly-written articles and (marketing) emails are already AI-made or heavily leaning on it. With as much automation as possible.

We will see where that leads us all in the future. But, with the current level of AI development, I try to rely on human expert knowledge and advice when possible - relying on AI can range from misleading to outright dangerous, depending on what you ask of it.

Relja MostlyCarbonBased Novovic

 

[HostingSift]

Fair point on the quiz. It scores providers based on plan types, features and pricing data, not on reputation or service quality. So yes, a provider that checks a lot of boxes on paper can rank high even if the actual experience doesn't match. It's not perfect and I'm aware of that, it's something I want to improve over time. That said, the scoring treats every provider the same way, there's no manual boost or paid placement for anyone.

As for the spam claim, I wrote to a total of 3 people and you were one of them. Not exactly a mass campaign.

The site simply presents data in a convenient way. It doesn't push users toward any provider. No editorial "top picks", no "editor's choice", no paid rankings. Just specs, prices and renewal rates so people can make their own conclusions.

And honestly, the bigger problem in this space isn't AI, it's that most reviews and articles about hosting providers are compromised. Paid content dressed up as honest opinions, rankings that go to whoever pays the highest affiliate commission. Pick 100 random "best hosting" articles and see how many are actually objective. That's what people should be skeptical about. This tool is trying to be an alternative to that. It just shows the data and lets the user decide.

 

[BikeGremlin]

Fair point on the quiz. It scores providers based on plan types, features and pricing data, not on reputation or service quality. So yes, a provider that checks a lot of boxes on paper can rank high even if the actual experience doesn't match. It's not perfect and I'm aware of that, it's something I want to improve over time. That said, the scoring treats every provider the same way, there's no manual boost or paid placement for anyone.

As for the spam claim, I wrote to a total of 3 people and you were one of them. Not exactly a mass campaign.

If it really was only 3 people, then it is still spam. Just not on a mass level (so Gmail outbound spam filter did not get triggered, nor any recipient spam filters - apparently).

The site simply presents data in a convenient way. It doesn't push users toward any provider. No editorial "top picks", no "editor's choice", no paid rankings. Just specs, prices and renewal rates so people can make their own conclusions.

Without a highly-visible bold disclaimer that the data is basically just googling for you and has no curation whatsover, the info can be worse than useless - it can be misleading (to novices who might fall for using such a tool).

To take a step back:
even with best intentions, it is difficult to rank providers in a meaningful way (for example, not every "unlimited X" is limited in the same way, just to note that). Still, without a clear disclaimer that it basically only lists prices and the (non-discounted) renewal prices, it would make little sense, apart from an AI-use experiment.

My thoughts aloud:
As an AI experiment, the project is pretty impressive. I would have no shame presenting it like that and doing a case study (explaining the infrastructure, the whole design etc.). Though I suppose most normal people would try to keep that to themselves and try to make money from that knowledge, instead of sharing it for free (that is reasonable, no objections there).

And honestly, the bigger problem in this space isn't AI, it's that most reviews and articles about hosting providers are compromised. Paid content dressed up as honest opinions, rankings that go to whoever pays the highest affiliate commission. Pick 100 random "best hosting" articles and see how many are actually objective. That's what people should be skeptical about. This tool is trying to be an alternative to that. It just shows the data and lets the user decide.

I 100% agree with you on this.
But the definition of information is "data that reduces one's ignorance" (i.e. "reduces the unknown," or "increases one's knowledge" if you like).
That is where the site fails IMO - since it doesn't do that. It just gives data.

That is sad because based on the site and all the other automation we've already discussed, you seem like a quite competent developer. That is impressive. My only "objection" (if we can call it that) is the site itself as it is now - what it does, what it offers to the user (even though I did put emphasis on the how it was made - the AI side of things - but the core problem is deeper than that).

Relja

 

[HostingSift]

If it really was only 3 people, then it is still spam. Just not on a mass level (so Gmail outbound spam filter did not get triggered, nor any recipient spam filters - apparently).



Without a highly-visible bold disclaimer that the data is basically just googling for you and has no curation whatsover, the info can be worse than useless - it can be misleading (to novices who might fall for using such a tool).

To take a step back:
even with best intentions, it is difficult to rank providers in a meaningful way (for example, not every "unlimited X" is limited in the same way, just to note that). Still, without a clear disclaimer that it basically only lists prices and the (non-discounted) renewal prices, it would make little sense, apart from an AI-use experiment.

My thoughts aloud:
As an AI experiment, the project is pretty impressive. I would have no shame presenting it like that and doing a case study (explaining the infrastructure, the whole design etc.). Though I suppose most normal people would try to keep that to themselves and try to make money from that knowledge, instead of sharing it for free (that is reasonable, no objections there).



I 100% agree with you on this.
But the definition of information is "data that reduces one's ignorance" (i.e. "reduces the unknown," or "increases one's knowledge" if you like).
That is where the site fails IMO - since it doesn't do that. It just gives data.

That is sad because based on the site and all the other automation we've already discussed, you seem like a quite competent developer. That is impressive. My only "objection" (if we can call it that) is the site itself as it is now - what it does, what it offers to the user (even though I did put emphasis on the how it was made - the AI side of things - but the core problem is deeper than that).

Relja

Appreciate the honest feedback, you make some valid points. There's definitely room for improvement, especially around adding more context to the data so it's actually useful and not just numbers. Thanks for taking the time.