How to stop Google from AI-summarising your website

Part of the reason for writing is to cultivate an audience, to bring like-minded people together.

Letting a middleman wedge itself between you and your reader damages the ability and does NOT benefit the writer. If the writer wanted an LLM summary, they always have the option to generate it themselves. But y'know what? Most writers don't. Because they don't want LLM summaries.

---

Also, LLMs have been known to introduce biases into their output. Just yesterday somebody said they used an LLM for translation and it silently removed entire paragraphs because they triggered some filters. I for one don't want a machine which pretends to be impartial to pretend to "summarize" my opinions when in fact it's presenting a weaker version.

The best way to discredit an idea is not to argue against it, but to argue for it poorly.

Honestly, publishers should just allow it. If the concern is lost traffic, it could be worse — the “source” link in the summary is still above all the other results on the page. If the concern is misinformation, that’s another issue but could hopefully be solved by rewriting content, submitting accuracy reports, etc.

I do think Google needs to allow publishers to opt out of AI summary without also opting out of all “snippets” (although those have the same problem of cannibalizing clicks, so presumably if you’re worried about it for the AI summary then you should be worried about it for any other snippet too).

Or asking if you want to pay to remove false information that they generate which makes you look bad.

In all honestly, if you're hosting it on the internet, why is this a problem? If you didn't want it to backed up, why is it publicly accessible at all? I'm glad the internet archive will keep hosting this content even when the original is long gone.

Let's say I'd read your website and wanted to look it up one day in the far future, only to find many years later the domain had expired, I'd be damn glad at least one organization had kept it readable.

>Good! It's literally the Internet Archive and you published it on the internet. That was your choice.

>As a general rule, people shouldn't get to remove things from the historical record.

>Sometimes we make exceptions for things that were unlawful to publish in the first place -- e.g. defamation, national secrets, certain types of obscene photos -- where there's a larger harm otherwise.

>But if you make someone public, you make it public. I'm sorry you seem to at least partially regret that decision, but as a general rule, it's bad for humanity to allow people to erase things from what are now historical records we want to preserve.

But it's my content - it's not your content. I don't regret my decision, anything I really don't want public is behind a login. The website is still there, still getting crawled.

What really upsets me the MOST though is IA won't even reply to my requests to tell me "We're not going to remove it" - your reply (I am assuming from your wording you have some relationship with them, apologies if that's not the case) is the only information I've got! (Thanks)

[Note reply was from user crazygringo but I can't find it now, almost like they... removed it? It was public though and I'm SURE they won't mind me archiving it here for them.]

Look at the reason, and get mad to the correct people.

It might be the archive themselves, but just be sure.

Why would you NOT want internet archive to scrape your website? (Im Clueless - thank you)

They won an award for the paper, and the example they given was a "holiday" search, where a hotel inserted their name, and an airline company wedged themselves as the best way to go there.

If I can find it again, I'll print and stick its link all over walls to make sure everybody knows what Google is up to.

Edit: Found it!

[0]: https://research.google/blog/mechanism-design-for-large-lang...

IMO a LLM is just a superior technology to a search engine in that it can understand vague questions, collate information and translate from other languages. In a lot of cases what I want isn't to find a particular page but to obtain information, and a LLM gets closer to that ideal.

It's nowhere near perfect yet but I won't be surprised if search engines go extinct in a decade or so.

Google even put the AI snippet above their ads, so you know how bad it stings.

And technically speaking, this citation is the first link on the results page, so you “rank” higher than all the other results. But it does take two clicks to get to your page.

They should make the citations more prominent and use the page title as anchor text. And when there’s multiple citations, the side panel should be open by default or they should put all of the citations inline as prominent links with page titles as anchor text.