23 ms·

Or rather, the end of the easy mode search engine. Google has been losing the fight against SEO for the past years. I'm not sure if they don't care (because ads

Or rather, the end of the easy mode search engine. Google has been losing the fight against SEO for the past years. I'm not sure if they don't care (because ads bring in the money, not SERPs) or if they can't. Relying on PageRank / links (like they currently do as the super-dominating factor) doesn't work if larger sites will employ AI (like BuzzFeed). Relying on their ability to tell apart AI and human generated texts? I'm not sure. I'm not an expert, but I fed in text from a German public broadcaster into one of the "identify ChatGPT" things and it said it looked like ChatGPT. I doubt these technologically conservative editors are using ChatGPT to write official reports. Instead, they'd need to understand the topic at hand to judge whether a page is useful in answering the user's question. And at that point, they have an answering machine and no longer need to show search results.

PurpleRamen 4y ago
> Or rather, the end of the easy mode search engine. Google has been losing the fight against SEO for the past years. I'm not sure if they don't care (because ads bring in the money, not SERPs) or if they can't. It's probably that they can't. It's hard to evaluate content which you don't understand and when you have the whole world gaming against you. Likewise, it will be also interesting to see if AI really can change this, because the results so far were not gamed by SEO. At the end AI has the same problem to solve if it wants to replace search engines. The question is also how fast AIs can update their knowledge. Google Search can be blazing fast in delivering the newest results even for the most random topic. > Relying on PageRank / links (like they currently do as the super-dominating factor) doesn't work if larger sites will employ AI (like BuzzFeed). Aren't they doing more than this for some time now? Though, true that links alone are not very useful against Manipulations. Maybe letting AI curate the results will be the trick? Another problem might be that Google want's to stay simple and vague. Which is of course legit, as more information will give SEO more attack-surface. But maybe add a more complex interaction-mode for people who don't just search 2-3 words. A built-in ai-prompt which switches to some ChatGPT-Like response is an obvious google-style solution to battle against the new wave of competitors. > Relying on their ability to tell apart AI and human generated texts? I'm not sure. Is this necessary? At the end, it's a matter of quality. If AI-Texts are high quality, delivering exactly what you want, why not? After all, the problem with spam today is the quality. Though, for this of course Google would need a kind of quality-check, which their own AI can deliver, maybe. > Relying on their ability to tell apart AI and human generated texts? I'm not sure. I'm not an expert, but I fed in text from a German public broadcaster into one of the "identify ChatGPT" things and it said it looked like ChatGPT. I doubt these technologically conservative editors are using ChatGPT to write official reports. But they usually use a very formalistic template-based style. Often it's even just copy & paste of paragraphs from different sources. So yes, news are already very AI-like today, even without AI involved. The human usually only brings in their skill in evaluating the content and composing them with matching parts. > Instead, they'd need to understand the topic at hand to judge whether a page is useful in answering the user's question. And at that point, they have an answering machine and no longer need to show search results. Yes, but people have not always questions to answer. And it's also a kind of trust-issue whether I want to see the original source, or some more context to the answer, or whether I blindly trust the answer from Google. I mean they already today have common Q&As in their search results, and you can see also where they are coming from. But often they don't quite satisfy my demand. Sometimes they are to short, sometimes they are plain out wrong. Sometimes they are just not what I search. Traditional search still has its place even with AI, until we reach the age of completely trustable strong AIs.
- luckylion 4y ago
  They are doing more than links, but I'd compare it to a race where you do 500 miles by car and the last 5 miles are on foot. Only when the cars are very similar in speed does the "on foot" result matter. It's why everyone in SEO is buying themselves into whitelabels on large, well-linked sites. You're right that AI-generated wouldn't matter if the AI-text was of high quality. So far, Google says that AI-generated is against the guidelines (read: you can/will be penalized if you do, like you might be for cloaking). Maybe they'll change their stance if AI provides high quality texts. But if it can, will anyone still need Google, or will they go to the AI directly? And again, to judge the quality, Google would probably need to be able to generate the result themselves, in which case they'd have the trustable AI.
pards 4y ago
> Relying on PageRank / links Search engine companies learnt long ago that parsing all that content was time-consuming and costly, so they simply use the 'meta' tags instead.[0] [0]: https://calpaterson.com/metadata.html https://calpaterson.com/metadata.html

PurpleRamen 4y ago

> Or rather, the end of the easy mode search engine. Google has been losing the fight against SEO for the past years. I'm not sure if they don't care (because ads bring in the money, not SERPs) or if they can't. It's probably that they can't. It's hard to evaluate content which you don't understand and when you have the whole world gaming against you. Likewise, it will be also interesting to see if AI really can change this, because the results so far were not gamed by SEO. At the end AI has the same problem to solve if it wants to replace search engines. The question is also how fast AIs can update their knowledge. Google Search can be blazing fast in delivering the newest results even for the most random topic. > Relying on PageRank / links (like they currently do as the super-dominating factor) doesn't work if larger sites will employ AI (like BuzzFeed). Aren't they doing more than this for some time now? Though, true that links alone are not very useful against Manipulations. Maybe letting AI curate the results will be the trick? Another problem might be that Google want's to stay simple and vague. Which is of course legit, as more information will give SEO more attack-surface. But maybe add a more complex interaction-mode for people who don't just search 2-3 words. A built-in ai-prompt which switches to some ChatGPT-Like response is an obvious google-style solution to battle against the new wave of competitors. > Relying on their ability to tell apart AI and human generated texts? I'm not sure. Is this necessary? At the end, it's a matter of quality. If AI-Texts are high quality, delivering exactly what you want, why not? After all, the problem with spam today is the quality. Though, for this of course Google would need a kind of quality-check, which their own AI can deliver, maybe. > Relying on their ability to tell apart AI and human generated texts? I'm not sure. I'm not an expert, but I fed in text from a German public broadcaster into one of the "identify ChatGPT" things and it said it looked like ChatGPT. I doubt these technologically conservative editors are using ChatGPT to write official reports. But they usually use a very formalistic template-based style. Often it's even just copy & paste of paragraphs from different sources. So yes, news are already very AI-like today, even without AI involved. The human usually only brings in their skill in evaluating the content and composing them with matching parts. > Instead, they'd need to understand the topic at hand to judge whether a page is useful in answering the user's question. And at that point, they have an answering machine and no longer need to show search results. Yes, but people have not always questions to answer. And it's also a kind of trust-issue whether I want to see the original source, or some more context to the answer, or whether I blindly trust the answer from Google. I mean they already today have common Q&As in their search results, and you can see also where they are coming from. But often they don't quite satisfy my demand. Sometimes they are to short, sometimes they are plain out wrong. Sometimes they are just not what I search. Traditional search still has its place even with AI, until we reach the age of completely trustable strong AIs.

luckylion 4y ago
They are doing more than links, but I'd compare it to a race where you do 500 miles by car and the last 5 miles are on foot. Only when the cars are very similar in speed does the "on foot" result matter. It's why everyone in SEO is buying themselves into whitelabels on large, well-linked sites. You're right that AI-generated wouldn't matter if the AI-text was of high quality. So far, Google says that AI-generated is against the guidelines (read: you can/will be penalized if you do, like you might be for cloaking). Maybe they'll change their stance if AI provides high quality texts. But if it can, will anyone still need Google, or will they go to the AI directly? And again, to judge the quality, Google would probably need to be able to generate the result themselves, in which case they'd have the trustable AI.

pards 4y ago

> Relying on PageRank / links Search engine companies learnt long ago that parsing all that content was time-consuming and costly, so they simply use the 'meta' tags instead.[0] [0]: https://calpaterson.com/metadata.html https://calpaterson.com/metadata.html