Google I/O 2023 - Above the fold real estate to be 100% AI generated

[bernard]

Yeah people talk about it being a game-changer for SEO, but I don't know how much of the translation industry this thing wiped out. More than 50% for sure. You'll still need someone to read through it & double-check for important documents, but everything else? Man. Fast, accurate & free. That's longer-term though, right now the translators will just be using it to speed up their work.

I'd be worried if I was in the translation industry for sure. ChatGPT 4 is much better at translation than ChatGPT 3, so I can only imagine what 5 will do. My guess is that we're close to voice translation already. There's really not much to improve with 5 in terms of accuracy if you use the method I show above imo. Of course it will always be lacking in personality and focus.

On the other hand, translators can just use it themselves and read through it and correct. Should drive down costs for the individual pr. word, but doesn't have to necessarily do it for the business as a whole.

 

[ianovici]

@bernard the translations are shit.

The worst one is the second. Nobody talks like that and the translation is wrong. In this instance, Google Translate is more accurate.

He just flourished the text to appear more capable. But in essence, is wrong.

Google has some human edits. For phrases, context, it’s doing a better job because of that. It fails if the verbs and pronouns are not in the correct time and form.

I can write a comparison if you’d like but i think it beats the purpose.

 

[bernard]

@bernard the translations are shit.

The worst one is the second. Nobody talks like that and the translation is wrong. In this instance, Google Translate is more accurate.

He just flourished the text to appear more capable. But in essence, is wrong.

Google has some human edits. For phrases, context, it’s doing a better job because of that. It fails if the verbs and pronouns are not in the correct time and form.

I can write a comparison if you’d like but i think it beats the purpose.

Interesting.

Did you check the third translation?

 

[ianovici]

Interesting.

Did you check the third translation?

Yes, I did. The first one, the one from Google Translate is the most accurate considering the context of the text.

The second and third(from chatgpt), are worst. Of these two, the first one is better. The problem with the third translation is that it ignores the message Poplanu was trying to express and adds text to make it seem more "something". I don't know what this "something" is supposed to be.

It is possible that chatgpt is "censoring" itself. Try with a more positive text and without any kind of "aggressivity" in context.

 

[bernard]

Yes, I did. The first one, the one from Google Translate is the most accurate considering the context of the text.

The second and third(from chatgpt), are worst. Of these two, the first one is better. The problem with the third translation is that it ignores the message Poplanu was trying to express and adds text to make it seem more "something". I don't know what this "something" is supposed to be.

It is possible that chatgpt is "censoring" itself. Try with a more positive text and without any kind of "aggressivity" in context.

Well, that's very interesting.

I do think it might have to do with the aggressiveness. Yes, ChatGPT is very self censored.

I will say that for me, using it in more traditional content, it's a no context with Google Translate. While I agree that Google Translate can sometimes be better, it makes some very obvious grammatical mistakes that simple are unacceptable. I've found that ChatGPT do not make these mistakes.

This is a good example that it's not a good idea to use ChatGPT going in "blind". I'm sure you as a spanish speaker could respond to that prompt and make it do it better than Google Translate, where as I would not be able to. You could try and report back?

 

[ianovici]

Well, that's very interesting.

I do think it might have to do with the aggressiveness. Yes, ChatGPT is very self censored.

I will say that for me, using it in more traditional content, it's a no context with Google Translate. While I agree that Google Translate can sometimes be better, it makes some very obvious grammatical mistakes that simple are unacceptable. I've found that ChatGPT do not make these mistakes.

This is a good example that it's not a good idea to use ChatGPT going in "blind". I'm sure you as a spanish speaker could respond to that prompt and make it do it better than Google Translate, where as I would not be able to. You could try and report back?

Will give it a go later.

I use chatgpt to extend or trim content. I edit pages where I lack some context or maybe refresh some parts. I've noticed that even with the same data, the results are not always the same. Also noticed that changing a small and insignificant part, that has no real effect on the result, it may drastically change the output.

Sometimes, it can work really well and I'll be amazed at how well it does a task. But most of the time, it fluffs shit.

Now something on topic with the post. Google will take the content and train its A.I. to the extent that all results will have a response from Bart. All sites will lose revenue, jobs lost, and so on.... But, I think that the EU will pass some laws to protect news sites and content creators.

Google does not seem to realize, without fresh content and on topic, having AI in search is gonna be useless. If I lose money, I'm not gonna put content just to help Google. And I assume that most people will do the same. We transform all sites into paid subscriptions to read, like news sites?

Let's take fishing, is the AI gonna write guides on how to target different species, or troubleshoot gear? Will he write tutorials and reviews of rods/reels and lakes/rivers? I just don't see how you can maintain something like that without fresh content. Lures and baits pop up daily on the market, and the same with gear. Yeah, it's gonna work for some time, but you'll hit a brick wall eventually.

 

[Ryuzaki]

Google does not seem to realize, without fresh content and on topic, having AI in search is gonna be useless.

Another thing that I feel gets overlooked is that as real people publish less or put their content behind paywalls, what these large language models (LLM) will have access to is exponentially increasing amounts of AI content. Of course, they can be selective about what sites to draw content from, but they will inevitably be unsure what sites are mixing in AI content, etc. And ultimately we'll have AI language models being trained on AI language models.

And this is already happening. Google had an internal memo leaked where they were freaking out saying not only Google but Open AI / Microsoft are both getting creamed by open source AI workers, who exist in much larger quantities. And the way the open source guys are getting around having the funding to retrain their next LLM from the ground up like Google and Open AI are doing is to simply train it on the previous LLM. So not only are they deploying faster, they're spending way less money and have access to way more human minds and man-hours. So while the language may not improve, all of the surrounding abilities are going to outpace the corporations' abilities incredibly fast.

Which is great news. We need AI to be decentralized and readily available, not locked behind the vault doors of large corporations. Since AI is going to fundamentally change the game no matter what, we might as well get there faster, cheaper, and open source so the overlords can't use it to choke out content creators or whatever else.

But yeah, my main point is eventually LLM's will be trained on AI content and it's just going to get goofy until someone ponies up the money to start from ground zero with the training and pays to have access to more and more quality content behind subscriber walls.

 

[bernard]

But, I think that the EU will pass some laws to protect news sites and content creators.

This will definitely happen. Unsure of how it will work out, but yes, this will happen in the EU.

 

[secretagentdad]

I mean if you're gonna make ur website by scraping combobulating and processing data from my website I wanna get paid. Ad revenue should be shared with the sources. Ahrefs gang is barking up the right tree.

https://yep.com/about
How do we get anyone, I don't even care who, to live up to half of that.
Or at least start the discussion about the topic of sharing the indexing rent.

Playing pretend that its just a bunch of complainers nooooooticing things seems pretty anti productive.
I get it you can compete moar harder and alpha dog webmasters just grind more and win the new way or what ever. Lets hypothetically discuss how to go in a direction that allows more than a ever consolidating minority of participants to harvest some value.

 

[tziahs]

This will definitely happen. Unsure of how it will work out, but yes, this will happen in the EU.

Italy already banned ChatGPT mainly because it gathers sensitive personal information about its users. Another major concern was that it simply makes up data. Also, it causes major problems in the education sector.
The EU has strict GDPR rules and it's expected to regulate ChatGPT in the next few months.

 

[Ahmed71]

Actually, SEOs saw it coming when chatgpt was first launched. And to the naysayers, every AI announcement looks like scarier than the last one.

Personally, i run a content, an SEO and a lead-gen/email marketing team. And i believe there will be more demand for contents curated by proven niche experts.

So i'm pivoting my content business from in-house generic writers to niche expert, part-time writers. That said, it'll be possible to deliver more E.E.A.T.'s to clients this way.

 

[Politico]

Actually, SEOs saw it coming when chatgpt was first launched. And to the naysayers, every AI announcement looks like scarier than the last one.

Personally, i run a content, an SEO and a lead-gen/email marketing team. And i believe there will be more demand for contents curated by proven niche experts.

So i'm pivoting my content business from in-house generic writers to niche expert, part-time writers. That said, it'll be possible to deliver more E.E.A.T.'s to clients this way.

I think this is the way.

Content by experts in the field and on the ground of whatever industry you're in will always be valued over generic (and sometimes censored/incorrect) content. The reason Google was so great and grew into a verb, was it was able to (at one point in history) find and serve up the best information created and cited by trusted sources. It didn't just make up an amalgamated answer based on it's index of information sources like ChatGPT and other LLMs are doing.

To Ryu's point, if AI starts getting trained in a recursive loop on already AI generated content, it's going to become absolute garbage, and fast.

You can't trust GPT for shit (yet). A lot has to be double checked by experts or people with knowledge anyways, as we're seeing in this thread.

AI is a game changer in lots of ways, but I think we're scared of the wrong things right now. - I don't think its main threat to content sites or SEO is its ability to generate content. That's bottom of the barrel type shit.

 

[DevMystic]

Another thing that I feel gets overlooked is that as real people publish less or put their content behind paywalls, what these large language models (LLM) will have access to is exponentially increasing amounts of AI content. Of course, they can be selective about what sites to draw content from, but they will inevitably be unsure what sites are mixing in AI content, etc. And ultimately we'll have AI language models being trained on AI language models.

I wonder if people will start blocking Google from indexing their sites so that it won't be able to use their content.

However, that can only happen if a competitor (e.g. Bing) can provide the same amount of traffic to the publishers. Which might be a long time from today.

 

[secretagentdad]

I wonder if people will start blocking Google from indexing their sites so that it won't be able to use their content.

However, that can only happen if a competitor (e.g. Bing) can provide the same amount of traffic to the publishers. Which might be a long time from today.

The big hold up to this is they just push your brand tail to your competition and no matter how hard you try you can't dehostage your userbase.

 

[OnlineDopamine]

I think this is the way.

Content by experts in the field and on the ground of whatever industry you're in will always be valued over generic (and sometimes censored/incorrect) content. The reason Google was so great and grew into a verb, was it was able to (at one point in history) find and serve up the best information created and cited by trusted sources. It didn't just make up an amalgamated answer based on it's index of information sources like ChatGPT and other LLMs are doing.

To Ryu's point, if AI starts getting trained in a recursive loop on already AI generated content, it's going to become absolute garbage, and fast.

You can't trust GPT for shit (yet). A lot has to be double checked by experts or people with knowledge anyways, as we're seeing in this thread.

AI is a game changer in lots of ways, but I think we're scared of the wrong things right now. - I don't think its main threat to content sites or SEO is its ability to generate content. That's bottom of the barrel type shit.

GPT-generated content was never the concern; Google (and other search engines) answering the complete query is.

A food blogger friend of mine tested Bard quite frequently. Literally just took all of his recipes and answered it without attribution.

Pichai said during the I/O event that they’re going to run ads within those AI snippets.

If that’s the case, then it’s truly over for the display-ad business model (+ cookie removal in 2024).

And I have a hard time believing that the likes of Mediavine or AdThrive will file class-action lawsuits on behalf of their publishers given their deep ties to Google’s AdExchange.

Time will tell but right now I’m extremely pessimistic..

 

[bernard]

Paying for data use is long overdue.

GPT-generated content was never the concern; Google (and other search engines) answering the complete query is.

A food blogger friend of mine tested Bard quite frequently. Literally just took all of his recipes and answered it without attribution.

Pichai said during the I/O event that they’re going to run ads within those AI snippets.

If that’s the case, then it’s truly over for the display-ad business model (+ cookie removal in 2024).

And I have a hard time believing that the likes of Mediavine or AdThrive will file class-action lawsuits on behalf of their publishers given their deep ties to Google’s AdExchange.

Time will tell but right now I’m extremely pessimistic..

You can't just steal people's content like that, not even in the US.

New laws will have to happen.

 

[zelch]

I'd say (and I didn't watch the presentation) that it'll wholly depend on if that big ass AI box immediately loads and covers the entire above-the-fold or if it's a "click to engage" kind of thing that then pushes everything down. If I had to guess it'll depend on the type of query, and in some cases will start expanded like a featured snippet.

I've been looking at my biggest converting phrases and trying to see what bard says for them. I just type my biggest converting/money phrases into bard and see what it says. (I mean, dont ask bard a question, just type the phrase as if you were searching google "dog snuggie" "solar panels dallas" etc) If bard is able to predict what you're looking for and give a good answer, and if it seems that users wont need to scroll farther, I'm assuming google will track that and keep showing it. If the answer is crap and people scroll past it to the serps, I'm assuming google will track that also and hopefully show the AI box less.

 

[tyealia]

GPT-generated content was never the concern; Google (and other search engines) answering the complete query is.

A food blogger friend of mine tested Bard quite frequently. Literally just took all of his recipes and answered it without attribution.

Pichai said during the I/O event that they’re going to run ads within those AI snippets.

If that’s the case, then it’s truly over for the display-ad business model (+ cookie removal in 2024).

And I have a hard time believing that the likes of Mediavine or AdThrive will file class-action lawsuits on behalf of their publishers given their deep ties to Google’s AdExchange.

Time will tell but right now I’m extremely pessimistic..

Anything that can be answered in a paragraph is pretty much going to be -5-30% traffic. If its a quick recipe you don't need images for, I'm not in the niche but would assume that would drop. Personally I would still like to see some images of someone actually doing it and I would click a link.
To me for info content this essentially looks like featured snippet 2.0, where whatever featured snippets had cut off for zero click searches is going to get further cut down. They showed they will have 3 helpful content links at the right of the answer still so its not complete armageddon at all. I think a bigger issue is all the one click ai generated spam sites brute forcing the index, I think that's a much bigger threat than maji.

 

[Politico]

GPT-generated content was never the concern; Google (and other search engines) answering the complete query is.

A food blogger friend of mine tested Bard quite frequently. Literally just took all of his recipes and answered it without attribution.

Pichai said during the I/O event that they’re going to run ads within those AI snippets.

If that’s the case, then it’s truly over for the display-ad business model (+ cookie removal in 2024).

And I have a hard time believing that the likes of Mediavine or AdThrive will file class-action lawsuits on behalf of their publishers given their deep ties to Google’s AdExchange.

Time will tell but right now I’m extremely pessimistic..

I get what you're trying to say, but Google (or whoever) answering the complete query using AI is quite literally GPT (or whatever) generated content. And I still think for anything but relatively simple/straight-forward queries, there isn't enough expertise/nuance there.

If you're looking for a TL/DR, then sure it will definitely satisfy the query most of time. But if I really want to learn something deeply, AI is only a starting point. Will that change? And eventually take away the need for more context/expertise/nuance - yea, most likely. But I don't think we're as close to that as everyone thinks.

 

[CCarter]

But I don't think we're as close to that as everyone thinks.

The behavior of people within the SEO community would suggests otherwise. They are changing their patterns very quickly.

I'm not going to lie, and I know I shouldn't say this lightly, but we are at an existential crisis point. Alot of SEOs are giving up and moving on. People have been losing their jobs, link sellers are quitting because people are switching to brute forcing with content and A.I.

This isn't being reported, but I estimate the SEO community has shrunk by 60%. But... perhaps I'm wrong.

 

[Politico]

people are switching to brute forcing with content and A.I.

Actually just watched that video you shared here and it's got some wheels turning in my head. With good prompt engineering, the content (and SEO) game is going to get REALLY interesting.

 

[googlealchemist]

And this is already happening. Google had an internal memo leaked where they were freaking out saying not only Google but Open AI / Microsoft are both getting creamed by open source AI workers, who exist in much larger quantities. And the way the open source guys are getting around having the funding to retrain their next LLM from the ground up like Google and Open AI are doing is to simply train it on the previous LLM. So not only are they deploying faster, they're spending way less money and have access to way more human minds and man-hours. So while the language may not improve, all of the surrounding abilities are going to outpace the corporations' abilities incredibly fast.

Just gonna ask the dumb question here...what prevents Google and Open Ai from

Paying for data use is long overdue.

You can't just steal people's content like that, not even in the US.

New laws will have to happen.

I could see them just 'rewriting' it enough to bypass laws like that

Just spitballing here...this is going to ramp up what we've already been seeing with growing above the fold ppc ad space, then the lsa ads, before even getting to the 'normal' local maps (which has its own ads section) and accompanying organic section.

I'm obviously mainly thinking of local brick and mortar service type stuff here. But it applies one way or another to anything else.

To me, this just lights a bigger fire under my ass to get paid ad campaigns properly understood and scaled out.

Alot of it will depend on how they are going to create the ai stuff above the fold with the embedded/paid links. Will it mirror how it serves up the existing local maps/organic listings or just be a faintly veiled monster of nothing more than who is paying the most for ads...

My brain is in a bit of a frazzled loop thinking about it actually.

Anyways, the past few days I've had this mantra in my head for some reason that keeps saying "Make hay while the sun shines" so I'm hitting it harder than ever focused on lead gen for local service type stuff.

And getting back into a project or two where the searcher is definitely not going to be satisfied with any sort of abbreviated, or even long form, ai generated shit...and still want/need to buy my digital/physical product.

 

[OnlineDopamine]

Just gonna ask the dumb question here...what prevents Google and Open Ai from



I could see them just 'rewriting' it enough to bypass laws like that

Just spitballing here...this is going to ramp up what we've already been seeing with growing above the fold ppc ad space, then the lsa ads, before even getting to the 'normal' local maps (which has its own ads section) and accompanying organic section.

I'm obviously mainly thinking of local brick and mortar service type stuff here. But it applies one way or another to anything else.

To me, this just lights a bigger fire under my ass to get paid ad campaigns properly understood and scaled out.

Alot of it will depend on how they are going to create the ai stuff above the fold with the embedded/paid links. Will it mirror how it serves up the existing local maps/organic listings or just be a faintly veiled monster of nothing more than who is paying the most for ads...

My brain is in a bit of a frazzled loop thinking about it actually.

Anyways, the past few days I've had this mantra in my head for some reason that keeps saying "Make hay while the sun shines" so I'm hitting it harder than ever focused on lead gen for local service type stuff.

And getting back into a project or two where the searcher is definitely not going to be satisfied with any sort of abbreviated, or even long form, ai generated shit...and still want/need to buy my digital/physical product.

It'll certainly be query-dependent. However, I could see smaller blogs that focus on display ads + PAA type queries (e.g., "can dogs eat chocolate") be wiped out. attribution will, if at all, only be done for the top 3-5 sites since those blogs don't bring anything new to the table and often just copy stuff.

I'm going to wait until the end of this year and continue loading up on cash. if it's as material as I fear, then sell existing sites and pivot into travel (using real photos, videos, and personal tone), which AI will never replace.

 

[zelch]

Not sure if everybody has seen this but some 12 or so hours ago, Google announced that in the near future, all above-the-fold results will be generated by AI.

I didn't see the announcement, did google say that ALL above the fold results will be A.I. ? or Just some? Because I've been playing with their preview version and some have AI and some do not

 

[DevMystic]

I didn't see the announcement, did google say that ALL above the fold results will be A.I. ? or Just some? Because I've been playing with their preview version and some have AI and some do not

Right, so when I made this post it wasn't very clear but it's good to know from beta testers like you the frequency!