Google, other search engines' use of generative AI threatens $68B SEO industry | Hacker News Reader

Google, other search engines' use of generative AI threatens $68B SEO industry

theconversation.com

16 pointsrntn22 comments

This article remind me of the argument that global peace is not desirable because it will affect the weapons industry which will cause economical turmoil across the globe.

Anything can be desirable. Less much is possible.

Global peace is not very possible until there're more than one human on the planet.

This article is mad. They are ignoring some giant elephants in the room, like how SEO is obliterating the quality of search and "organic" content, and how Generative AI is completely reliant on this organic content.

Its like my neighbors talking about their landscaping problems with hurricane Ian churning out in the gulf (which actually happened).

The greatest missed opportunity of organic content ranking is a robust mechanism for human feedback. Google has attempted to rank the trustworthiness/quality of content with almost no feedback from search users. About the only crude feedback that a search user can provide to Google is to return to search results and click on another result. Even that signal is largely ambiguous in meaning. Google has a team of trusted raters, but their work has not been sufficient to match the scale of bad actors manipulating content. With the advent of LLMs, there's another chance to get human-feedback right. This is being done now with RLHF, but that human-feedback system will need to be more aggressively designed against adversarial input as that feedback signal becomes more significant (and therefore profitable) to those who would distort it. TikTok is a good example of how effective a system with simple ranking rules can be when there is a tremendous volume of active and passive feedback from users (likes, subscribes, shares, total watch time, blocks/reports, etc.)

They most certainly took human actions into account, did you click on a search results and then come back? Did you click on multiple, or stay on one for a while?

Not all feedback needs to be direct and specific, though this is a feature I've long hoped for in Google Search. While I expect it to be useful, it would also be gamed by the SEO industry.

They have a paper that talks about many of the complexities: https://research.google/pubs/pub43146/

Back in the day, google recognized that people linking to other people on the open web was in fact a form of human feedback.

Of course, these days many users don't have their own personal websites anymore. (or in % terms, probably less than when google first started). But that's a story for another day.

Exactly. The original thesis of Google assumed humans creating earnest links between related/valuable content. In the modern web that realistically only happens in closed and semi-closed social networks. The vast majority of links on public web pages must now be assumed to be for calculated SEO value. There's a slim sliver of hope that a more open/federated system will replace Twitter (and ideally Facebook/Instagram too). I wish Google would do more to encourage this via Blogger or other mechanisms.

Good insight into what feedback from (technical) search users looks like: https://kagi.com/stats > "Domain Insights"

If Google ran a Tiktok-like system for long enough, it would be gamed too.

Honestly, Google search probably needs to burn to the ground to get better. That seems to be the most reliable antidote to enshittification.

> If Google ran a Tiktok-like system for long enough, it would be gamed too.

There would certainly be huge incentive to try to manipulate/game signals to boost rankings, but walled gardens (like TikTok) benefit the huge advantage of having logged-in identities for both content creators and content viewers.

One industry I would not be shedding any tears over if it disappeared is the SEO. Can't think of any non-toxic consequences of it existance.

Google and commercial websites are like two lovers who are bound together by their mutual need for each other. Google needs the traffic that commercial websites generate, and commercial websites need Google's AdSense revenue. This symbiotic relationship makes it difficult for Google to completely replace commercial websites with AI-generated summaries.

I actually think (hope?) that AI in general will get rather good at picking out SEO spam pages in their current state (think generic cooking recipe sites) and be able to downgrade them in search rankings. Now whether the major search engines will actually do that is another story.

I’m curious what incentive will remain for people to render facts about the world into AI-consumable form. It seems to me that the same incentives that motivate this supposedly-$68bn industry are the ones that motivate people to write what we think of as high-quality content: some degree of financial compensation, some degree of prestige or social capital, some degree of influence over consumer or voter behavior. If anything, what we’re stripping out is direct compensation and authorial prestige, leaving only crass commercial/political incentive.

Why would I want to share what I know about the world, from my position of hard-earned expertise and original research, if it’s just to anonymously make AIs sound smarter? And why would I be more motivated to do this than, say, entities with products or agendas to push, who hope to motivate AIs to influence consumer behavior or political beliefs—same as it ever was?

I’m reminded of the eternal example of Wikipedia turf wars, where true subject matter experts try a few times to correct stubborn self-appointed editors then eventually give up and return to the more important things they have to do.

I can only hope smart people are thinking about ways to equip the AI-based approaches with something of the same techniques we train into would-be scholars: consider the source, check your references, cite your work, ground everything in the real world.

> Why would I want to share what I know about the world, from my position of hard-earned expertise and original research, if it’s just to anonymously make AIs sound smarter?

This is a huge and important question, but it's worth considering how radically different things could be as LLMs become even more capable and popular. As one possibility:

LLMs decimate the number of people needed in many professions (SEO, education, etc. etc.).

LLMs become a subscription service. (Perhaps ranging from $50 to $200/month per household, which is not at all unreasonable considering the value they provide and what cable TV currently costs).

Governments, companies, institutions pay millions of content creators, quality raters, researchers directly for their contributions and how impactful to the LLM their input is.

Things could go any number of ways, but I feel it's a pretty safe bet there will be a lot of change.

I feel like this is already happening to me. Outside of GitHub, I've stopped posting useful information nuggets I discover.

Why waste my time screaming into an empty void on zombie forums, Reddit, Discord and such?

I'm pretty sure it's actually gonna do the exact opposite thing. Like if you look into what AI is doing to student essays and homework, it's so good that even trained humans cannot tell human and ai generated content apart. It's not a stretch to imagine AI generated SEO spam pages will be indistinguishable from human created personal websites.

Finally some good news!

... and there will be wailing, and gnashing of teeth.

I found my nearest Googler and told them "good job" :-)

Google has stopped treating double-quotes as an "exact match", which threatens their own business. Because now only a simple queries are possible with some random results.

It's similar to google map: search for reataurant and get bike repair shop. Nice job.

Good.

SEO is nonsense anyway. It's all shysters and snake oil salesmen.

Reply on news.ycombinator.com