"We pulled off an SEO heist with AI and stole 3.6M impressions." | Hacker News Reader

"We pulled off an SEO heist with AI and stole 3.6M impressions."

mastodon.social

59 pointsphreeza38 comments

I personally pin SEO gaming as the #1 cause of decline in the quality of the Internet. I only use Google at this point as a rapid link to websites so I don't have to type the URL. SEO posts just trash any meaningful search and Google seems content. LLMs are great for cutting through this, as long as you take time to verify your data.

A few times I have googled a company name directly and the first few results were SEO snipes.

Though you'd think something like page-rank would not be attracted to some random pages that are only linked to from the same domain.

#1 cause in the decline of Google maybe. https://search.marginalia.nu/ seems to manage though, so maybe Google just doesn't care.

It's mostly due to incentives. The most effective way to go after search engine spam is to aim for what the search engine spammers are after, which is mostly ad impressions. If you get rid of the ads, you get rid of the search engine spam. Of course, that would be directly hitting Google's bottom line, and is never going to happen[1]

The other factor is of course that they are very big, and get directly targeted in a way I'm not.

[1] http://infolab.stanford.edu/~backrub/google.html#a

To be fair, this is probably more Google not caring rather than any improvement in SEO gaming. Kagi doesn't suffer from the SEO spam like Google does, and if a small company can do it surely Google could if they wanted to.

Kagi can do it because of the fact that they're a small company, not despite it.

I doubt it has anything to do with company size. Most likely it's due to what drives their products. In the case of Google, overall it's how many ads they sell, and if improving search quality doesn't lead to more ads, then there is little incentive to do it.

For Kagi, they get revenue from customers paying for quality, so there is direct incentive to produce high quality search results.

Kagi isn't a target for SEO tactics. If they became as popular as Google, SEO manipulators would find ways to improve their position on Kagi.

Stop blaming SEO and start blaming Google.

Ultimately they're responsible for creating the incentives that are killing the web.

This is a screenshot of https://twitter.com/jakezward/status/1728032634037567509 hosted on Mastodon.

We pulled off an SEO heist that stole 3.6M total traffic from a competitor.

We got 489,509 traffic in October alone.

Here's how we did it:

1. Exported a competitor’s sitemap

2. Turned their list of URLs into article titles

3. Created 1,800 articles from those titles at scale using AI

18 months later, we have stolen:

- 3.6M total traffic

- 490K monthly traffic

I'm gonna go against the grain because I think AI content has potential to be better than human content. If Google is promoting low-quality content, regardless of who made the content, then Google should improve their algorithms, or users should use a different tool.

I could see AI content be better if humans took part of the time saved in producing an article from scratch and re-invested it in researching and editing to make the output higher value. If the idea however is to leave the AI to write the content with zero human intervention then it’s going to be the same SEO drivel we’ve been increasingly subjected to in the last decade, except at an unfathomable scale.

Seems easy enough:

"1. Exported a competitor's sitemap

2. Turned their list of URLs into article titles

3. Created 1,800 articles from those titles at scale using AI"

Wither Google (and, alas, the Internet).

What does the "at scale" part in #3 mean/imply?

A lot of work done quickly (in parallel); possibly for lower cost than previous alternatives.

It would be prohibitively expensive to write those articles by hand.

Faster than writing all 1800 articles manually I suppose.

It's fun watching LLMs poison the internet in real time. It's going even faster than previous iterations of SEO gaming and content farm proliferation.

Wait, fun? The opposite of that.

Why is it every other day there's a thread that goes something like "We used AI to find a new way to avoid jail time completely in the Prisoners Dilemma"? Yeah, if you defect you gain benefit for yourself while making the world worse. Doing things that don't benefit you directly to make the world better, and not doing things which do benefit you and make the world worse, is WHAT MORALITY IS.

Knowingly making the world a worse place because it brings benefit to you is immoral, full stop. Knowingly screwing over other people because it brings benefit to you, tricking other people into viewing your website by lying to them about the authenticity of the information, abusing the trust people have to not be giant asshats - all of these are immoral.

I'm upvoting this post because more people need to see a very clear and obvious example of what not to do. If you have ever thought about things like this, you need to change your ways.

It is because there are big incentives to make money in the current economy we live it, competition like this can benefit a market participant, and that is how the reward system works (monetarily/money-as-a-meta-resource and culturally (money as status symbol and means to exist)) until there is a legal precedent to prevent such behavior.

For most people with this mindset money is the ultimate good. If something they do makes money, it's an unquestionable benefit. Most people with this mindset either become very rich or end up in prison.

Their competitor paid real money for at least 1800 articles to trick other people into viewing their website--aka for SEO.

I think both groups are equally culpable.

If you're in the SEO game, you're already in violation of this same "morality."

What kinds of recourse do people have for this. Is there a theory under which one can sue (TM? C?) . Can it be reported to search engines?

Did they really "steal" the impressions though? I don't think anyone stops at a content farm, it's just 'oh not this one' and then back to the results to find a better one. Okay, they have gotten an impression from me, but they did not cause my actual destination to lose anything.

Notice they're not mentioning any increase in sales or in profit. So it's nothing but worthless spam. I can go to a crowded street and sing in a megaphone until they take me away, and then brag on Twitter that hundreds of people wanted to listen to my new song.

Impressions / likes / thumbs up have exactly 0.00c business value. It's only sales that matter, nothing else.

I saw it happen today.

Someone was discussing on a forum and found a news article about the issue. They posted a link to the article.

But it was a 2010s style "synonym spin" spam site, not the original source of the article. The spam article was obviously garbage to me, but that poster was happy to share it.

The use case for AI (as its current LLM version, maybe others won't be) is spam

Is there an original source for this idea that "The use case for AI is spam"?

I've seen it in a bunch of places recently, have people all independently came to the same idea at the same time or was there some big article somewhere that I missed?

Probably the first 2 usecases that came to my mind, after I saw what AI was capable of, were "spam at scale" and "writing papers to get published". Though, the latter is also kind of spam.

I think if you were ever involved in SEO, thinking "spam" as soon as you saw an LLM in action was natural.

You don't need to pay someone 0.01$ per word anymore, AI can do it for you.

It's the byline for this article

https://amycastor.com/2023/09/12/pivot-to-ai-pay-no-attentio...

Thanks! That's also a really good collection of posts that illustrate the phrase.

A lot has been made of the "pre-AI" internet as a sort of barrier for training sets as it will flood the internet with AI content that will be difficult to parse out later.

For the bullish-AI types, is there a prevailing hope that "human content" may itself be an artificial barrier and a properly trained AI could generate better training content for future AI generations? Is this strictly in "AGI" territory?

One upside is Google can maybe hopefully use that to improve its ability to detect and remove generated garbage using his sites as training material. These people don't usually paint a target on their output.

This seems like the evolution of content on the internet. What's more curiose is if you say a LLM article may be just as qualified as a random person on the internet then it seems that the LLM according to Google was better written or more useful to their audience.

nothing to do with ai.

Scaled content generation.

You could have executed the same strategy before, but coming up with several thousands coherent articles would have cost much more time and money.

Reply on news.ycombinator.com