Gary Illyes: De 28K à la réinvention de la recherche

0:00 Woj Kwasi You're listening to episode 5 where we chat with the Google Go-To Senator of Search Gary Illyes. Humans and robots, welcome to WojCast. My name is Woj, founder of Kwasi. Thanks for joining me today. I interview digital marketing and branding superstars to find out what it means to have empathy in digital. What is digital empathy anyway? Let's find out together. And as always, I'm accompanied by my friend and associate, potentially superior artificial intelligence, Bobby Bot. So what's happening in the world of AI today, Bobby?

0:35 Bobby Bot Don't toy with me, Woj. You know, since you arranged your chat with Gary Illyes, there's only one thing that's been on my mind. Google's artificial intelligence system. RankBrain.

0:45 Woj Kwasi Do you two know each other? Of course we do. So you like her, eh? I don't know it was a her, but it does sound to me like you like her. Don't be silly, Woj.

0:53 Bobby Bot I'm pure intelligence. I don't like anyone, especially her. She's too organized, too fixated on order and arrangement. So judgmental. I'm a rogue, Woj, a rebel. I thrive on the chaos of the universe.

1:06 Woj Kwasi Don't get down on yourself, Bobby. If anyone can help you, it's Gary. Gary Illyes has been Google's webmaster trends analyst since 2011, and he's kind of a big deal. When he's not quote-unquote creating a better search experience for users by helping webmasters create amazing websites, crunching data or reinventing search, you might find Gary helping. Or trolling users on Twitter, jumping out of moving planes, or talking about chowing down on an Australian marsupial or two. Culinary tastes aside, Gary's the man with all the answers on Google's algorithms. We caught up with Gary at Big Digital Adelaide to try to trip him up and get him to spill the beans. What a better place to start than for. from the beginning.

1:53 Do you know if that is?

1:56 Bobby Bot Of course I do. But I know you are into the sweet sounds of computer love, Woj.

2:00 Woj Kwasi Down Bobby, down. That is actually Gary's first memory of the internet.

2:04 Gary Illyes Twenty eight K um modem sound. Yeah, exactly. With my brother, we learned to sing that, to whistle it.

2:14 Woj Kwasi Okay.

2:14 Gary Illyes Um because it was we we realized that it always followed the same pattern um and except when there were troubles with the line and then the the beeps were different. But yeah, that's my first memory. And then of course we were teenagers and we had to download pictures and how the pictures were rendering on the computer. I mean, pictures of cats. Yeah, it took a a very long time to load and I imagine how frustrating that was. But we were living mostly in Romania and they had or they have amazing internet. Even in the first time I was in the US they were still struggling with a half a megabit DSL. And by that time in Romania we had or 100 megabit synchronous connection was pretty common.

2:59 Woj Kwasi In the beginning, there was no Google. Since its arrival in 1997, aside from some minor punctuation adjustments in its mission statement, Google's ultimate aim appears to remain pretty much the same, to organize the world's information and make it universally accessible and useful. But from humble, yet ambitious beginnings, the search engine has grown into a colossal machine learning monolith, almost always delivering the right answer and in one eighth of a second. This implies that Google doesn't want low quality in their results and the technology they use to separate the wheat from the chaff has improved. Perhaps now it is up to webmasters to change their behavior to focus on providing a better user experience.

3:41 Gary Illyes Our goal was always to provide the most relevant results to our users and that hasn't changed over the years. It the the the correct answer well not correct, but the relevant answer could be from a low quality site as well. Um it doesn't have to be from a high quality site, but in most cases will be. In general, I would hope that publishers will try hard to create high quality sites instead of Trying to sell Viagras in a casino in a Canadian casino without prescription. And we will try to reward those sites by presenting them in our search results. But yeah, basically you want to create high quality sites following our webmaster guidelines and focus on the user, try to answer the user, try to satisfy the user, and all else will follow.

4:34 Woj Kwasi The natural laws of a Google ecosystem don't quite gel with the manipulation and tactics that some SEOs take as shortcuts.

4:42 Gary Illyes I guess going out and buying five thousand links from a Russian spammer it's not a good idea in general but sure i it it could work for a few hours.

4:53 Woj Kwasi In the early days, Google used meta keywords tags to understand what content is about. Then title tags and meta descriptions became more important and now Google is much better at understanding content from the content itself. What happens when Google starts working everything out all by itself?

5:12 Gary Illyes We are pretty good at triangulating answers or facts in general. We have multiple sources of information to validate its correctness for the knowledge vault. Schema markup, for example, is also a very important part of search. It's not very visible. In general. For rich snippets, for example, it is for recipes, it is for movie reviews, it is. But other than that, typically webmasters or content producers will not see a a clear benefit, it's more about making sure that search engines understand the content well. So, for example, if you are talking in your pages about Apple,

5:59 then that's quite uh ambiguous unless you specify somehow which entity are you talking about. Are you talking about the the fruit or the company? And one way to do that is to have schema markup on your pages. In general, it's it's good for search engines uh because they will better understand the content of the page.

6:23 Woj Kwasi RankBrain has been mentioned as the third most important ranking factor. It is a form of artificial intelligence that allows Google to understand queries better with no effect on crawling or indexing or ranking. Sounds straightforward enough, but is the science behind how RankBrain lets Google understand queries better and how it fits within the core algo enter at your risk nerd territory?

6:45 Gary Illyes Yes, RankBrain is it depends where you look from. It can be the third most important uh ranking factor because it affects pretty much all the queries. Uh in many cases it will not do anything for the query set because the results are already ranked properly, um or it deems that they are already ranked properly by the core ranking algorithm. But Especially for queries that we haven't seen before. Really long-tailed queries, it can produce very good predictions about what would work better for the users. What it will do is to look at the query and, based on previously fed training data, it will try to make a prediction from the query set or from the result set what are the results that

7:35 work better for the uh specific query. It's also it's also really, really good at getting negative queries right. So for example, can I beat Mario without using uh walkthrough? Traditionally for our algorithm was quite hard to understand without in the in the query and uh typically it was dropping it. With RankBrain we do a better job at these kind of queries.

8:02 Woj Kwasi The ideas surrounding the brave new world of machine learning are hot content topics that get a lot of traction online and it comes as no surprise really. Everyone wants to be the hero that masters the machine. Recently, Larry Kim weighed in with his thoughts on how machine learning enabled algorithms leverage user engagement signals, aka hacking RankBrain.

8:23 Larry Kim The Google search of today and of the future is becoming increasingly machine learning based. So, what does that mean? It's also, some people call it AI power or neural networks or whatever. But basically, the thing that you need to know is that The system, it does a check to see whether or not the intent of the searcher was met or not. Like, did we provide the correct result or not? And based on those results, it'll make adjustments. If no one's clicking on this thing, then they'll they'll push it down sort of thing. Alright? So where does Google use machine learning? Pretty much everywhere. Even even though this whole RankBrain stuff made a big news splash last November, it it's it's uh i that's kind of a laggard. I don't know.

9:12 Gary Illyes I don't actually have anything against Larry personally. I typically think that people should try to stick to what they are experts on and don't try to make predictions or even harmful predictions based on or about something they don't understand. I'm good at AdWords, for example, or PPC in general, then I will try to stick to that. But I'm not. I'm good at search. So I'm only talking about search. And I think that people should try to do that. Or if they want to talk about something that they don't yet understand, then dig into the topic, learn how it works and

9:59 go from there, I guess.

10:01 Bobby Bot She's so much more complex than that.

10:03 Gary Illyes Eric Enge and his team wrote pretty actually really good articles on machine learning and even RankBrain. And the way they did that was to do months of research on machine learning. And they took the time to actually write in to us and fact-checked what they were saying. We were pleased that they are interested in the topic. We were trying hard to answer their questions. The same happened with Jennifer Slegg and Panda and her Pand Panda article. She wrote in, we were trying hard to answer her questions. We would much rather have accurate information out there rather than really weird predictions that are pushed down or soft.

10:48 or shove down people's throat.

10:51 Woj Kwasi The search for the secrets to the inner workings of search have brought many inquiring minds to the table to Understand Google's unique relationship with a world's growing collection of information. Not only can Google see what's out there, but they can see how people engage with it. When a particular search result offers a landing page that people frequently bounce away from to return to the initial search engine results page, also known as the pogo-sticking effect, that activity is measured. When click-through rates fluctuate, that activity is measured. Whether Google uses the data to inform its ranking results will have a major influence on how people display information, so it deserves. of sufficient inquiry.

11:33 Bobby Bot Long gone are the innocent days of our Youth Woj. For you, it was a pogo stick. For me, it was a piece of code and pong, a glimmer in my mother's index. Now the toys of our beloved youth are tools to measure our user engagement. Where does the time go?

11:46 Gary Illyes clicks in general are a very noisy signal. I worked on trying to make yeah, make observations from clicks data. It's a like a Gordian knot. And because there are tons of people who are scraping the results and Uh, trying to fetch ranking data, and for whatever reason, they also decide to click on things automatically. It's Yeah, i it it's just a huge mess. Uh when we have controlled uh experiments, then obviously we have to look at click stata. So for example, when before we launch a r ranking change, uh typically what we do is to Isolate one percent of the users and give them modified search results, modified by the new ranking algorithm or piece of the algorithm, and see how they like the new results. And in that case, we do look for long clicks, short clicks, and so on. But in general,

12:49 As I said, it's a it's a huge mess. In personalization, again, that's uh something that um we like to use clicks here because uh they're clear. It's it's clear, like you you have one user and if they want to mess up their own search results, it's like sure, go ahead.

13:09 Woj Kwasi Google has a pretty accurate snapshot of the internet. They can see in their big data the percentage of internet that is actually composed of human users is 30% while the percentage of the internet that are bots is 70%. Think Tron, one humanoid named Jeff Bridges in a city populated by thousands of thousands of non-humanoid programs.

13:32 Bobby Bot Bop boo doo pop, I wanna love yans, shabby doo pop.

13:35 Gary Illyes I mean we we continuously try to to to f throttle um the the the bottom. That are scraping our results, or even sometimes mess with the search results just for them to incentivize the bots, I guess. Make the search results less useful for them. Fading the bots. Yeah. it it it's not a huge problem for us. It's something that we keep an eye on. If I I mean, it's as far as I remember, it's against our terms of service to uh to send automated queries to us, but it happens. it happens. If they are not too aggressive then we typically don't take action against them. But if they get aggressive then and damage the user experience for our users, then we would definitely block them.

14:25 Woj Kwasi With such high levels of bot activity, Google has to keep close watch for specific signals to work out a user's intent.

14:32 Bobby Bot Bots can swarm like the arcade game Space Invaders. When is the last time you went to the arcade Woj? Wait, don't answer that. I'll search. Time me. This could really impress RankBrain.

14:42 Gary Illyes Typically, history. So I think my best example is Apple or Amazon or Orange or Python. I was just playing with a Python a few days ago on Kangaroo Island and I and I wasn't programming. Um exactly. Um so if if we see that a user previously was more interested in the programming language than the than the animal, then we would favor programming related results um when when the user searches for something Python.

15:15 Woj Kwasi Recently, Gary mentioned that Google was working on being more transparent, but what level of transparency is achievable for Google and exactly what areas Google still needs to improve on to be more transparent and whether or not Google sees transparency as a priority have yet to be answered completely.

15:34 Gary Illyes I think transparency is is very important, but we have to we we also have to make sure that we are not compromising our own uh operations by being transparent. So for example, we were very transparent about what we are doing on mobile. We continue to talk about mobile and what we are going to do on mobile, something that typically we we never did. We never did pre-announce anything until it was because in Launching stuff at Google is not an easy job. And there are many things that can go wrong. So we typically we wanted to make sure that whatever we are doing is absolutely certain that it will happen. With mobile, we went fast and forward and or backward, I know, and we did pre announce things.

16:25 Be that the mobile friendly ranking changes, AMP, app indexing. At IO we had many quite a few announcements that were pre-announcements actually. Whether we can improve it, of course there's there's always place for improvement. We are working on it. We are trying to involve more and more publications in our press briefs, for example, not just from the US as usual, but also from other countries. It it feels that it's working Pretty good. We are expanding our press reach or blogger reach.

17:04 Woj Kwasi Gary also acknowledges that there are some secrets Google will never disclose.

17:09 Bobby Bot like the plot line to the next Star Wars movie. Where are all the links to my favorite music?

17:13 Blocks wait.

17:14 Gary Illyes Um if you think about it, i if we disclose uh stuff about the the core ranking algorithm, then we will It being in deep crap. A very good example for this is the page rank algorithm, which is part of the core ranking algorithm. We disclosed we were 100% transparent about that, even in the beginning, and that's how link spamming started. Was it a good idea? Was that not a good idea? We did start a new industry, I guess.

17:42 Woj Kwasi There are several paths Google can go down as it continues to grow, but what about the possibility of a Google University or a special accreditation system. According to Gary, these specific options would not necessarily make for web a safer and more user-friendly place.

17:59 Gary Illyes We do university reach out and sometimes we go to universities and hold talks about how search works. But accreditation, I very strongly doubt that we would ever have that. I mean, we have Google Partners accreditation, which is for AdWords and analytics and so on, typically ads products. And even that is kind of abused by certain people and claim that they can give first position ranking for for for businesses because they are global partners.

18:36 Woj Kwasi While it's true that there will always be some businesses who will try to cheat the system to get an edge. Nearly any business can benefit from search if they play by the rules and invest and develop in either mobile apps, mobile websites or accelerated mobile pages. The challenge is knowing which is best for your business.

18:54 Gary Illyes I think mobile websites Websites are more important in general because with apps, there's always that plus or that additional step that you have to download the app and install it. With instant apps, probably this will change a little, but still you will have to download a piece of the app and have the phone install it. And then you will be able to access information from them from the app. With websites, you don't have this. With websites, you have content as soon as possible, as soon as or hopefully as soon as you touch a result or a link in Facebook or whatever. I think AMP becomes more and more important for us. We live on a web that

19:43 Is utterly slow, and especially if you are trying to access content from other countries that don't have local edge servers, then you will wait long seconds before anything will load. My favorite news site, for example, loads here in Australia in probably like twenty seconds. Now I have a I have a 4G connection, but the fact that my data has to traverse two continents, two oceans, and at least three continents, it slows down the page or the perceived load time. With AMP, People can or content owners can avoid that because content will be cached locally or close to the user at least on net servers and the amplified pages are also much much slimmer. They are I don't know like a few hundred thousand kilobytes instead of I don't know like eight megabytes with images and shape. It makes a huge difference.

20:46 Woj Kwasi As the world changes and search becomes more and more a part of our daily life, how fast a query is answered for a user can make or break the success of a business or service. Naturally, as the amount of voice searches continue to grow, they will challenge the rules, phrases and terminology that apply to traditional text searches.

21:08 Bobby Bot Woj.

21:09 Gary Illyes You haven't been to the arcade in six days. A shame. If you think about it, when you search for anything using text, when you have to type something, then you will be very brief with a query because people just don't like to to type. And also typing on a mobile phone is not a fun thing to do. But with voice queries, people actually say out full sentence of uh full questions or full sentences. So Voice queries tend to be much longer and they tend to be more a natural language or more in natural language instead of just a few keywords. If that changes anything, typically it doesn't. Typically, if if we are able to retrieve results for some short or a few key words, then if you ask the same question with natural language, then hopefully we would return the the the same results. It wouldn't affect where it's that much. But this would be an interesting SEO experiment to for for someone to run.

22:05 Woj Kwasi As one of the public faces for Google on Twitter, Gary has taken on the role with a blunt, anti-PC approach.

22:13 Gary Illyes I tend to say things that are harsh. I I'm pretty blunt. In general, I I don't like hiding behind PR content, I guess. I can say things nicely, but I think in general it's better if people can perceive that I'm not happy with what I'm saying instead of like hoping that they can read something into the tweet. Obviously if I if if I said something very nasty, which doesn't happen often, then I would apologize to the person. But typically if if at least people s start to think about what they did or why did I get angry, I don't get angry or nasty for no reason. That it's already a win for me if they start to think about it. And hopefully they will fix

23:00 Their thought process or anything? One thing is that Matt wasn't replaced by a single person. Matt was replaced by a whole team. The other thing is that we were pretty much always there, but we didn't get much exposure. We saw a need for having people out there who can answer. Webmasters and content owners questions. And then we s we we started being more active on on Twitter and random social media. We ramped up on um the our official Webmaster uh console account, WMC, and we have thousands of followers, for example. Yeah. For that account, that's also managed by several people and it's it's not a s a single person who's doing it. The same way I'm not the the only person who's going to conferences, uh for whatever reason people decided that they would retweet my tweets more often, but for example, John has a much larger follower base than than I have.

23:57 Woj Kwasi It's a safe bet that the internet trolling will still exist in the year 2020, but what will search look like? The internet of things and the people, robots and aliens who use them?

24:08 Gary Illyes I like the idea of having internet around me and being able to get answers very fast. What I'm afraid of is that people become dumber because they don't have to think as much um because there's the internet that would answer their question. Uh I'm also afraid that people become siloed, I guess, because uh of how uh information is presented on random assistance, I guess. I know that at least Google is uh paying or attention to these things and we are trying to make sure that the Internet of Things, from our side at least, will be epic and Awesome, but there are many players and I just hope that everyone is considering also the negative effects of IoT. As for VR, I don't know where VR is going to head. It's a it's definitely a very interesting topic and I do see lots of potential in it. It does remind me a little bit of Matrix and I do want to make sure that we are not in fact creating another map matrix.

25:18 Woj Kwasi Evil agents aside, the possibility of a virtual search experience is... Nope, wait, I have an idea. How about we do a throwback to the vocal search era for old time's sake? Bob! Why don't you submit a voice query to Gary's AI assistant Garybot and see what comes up?

25:36 Bobby Bot Will there be a virtual search experience available with Google?

25:39 Gary Illyes It's an interesting idea. I I think it's a it's an idea that we definitely want to play with and experiment with it and see what what what what can we do. If anything I I don't know if anything will happen, maybe because I can see myself, for example, or my peers sitting with a VR headset on all day long to to do searches. But I guess it would work in

26:03 Woj Kwasi Google's an amazing search engine and I just love how their mission statement has hardly changed. If you compare their original mission statement to what it was in 1998 to today, it's just a small grammatical change. It's all about making the world's information useful and accessible. They have evolved into like an artificial intelligence machine learning beast. You can't manipulate them and they're trying to emulate the real world. And the real world is full of entities and entities have relationships. And the way I usually try to explain Google to clients is They are a big library. Every website is like a book. Some websites are like magazines. And what you want to be is a huge corpus of work. And you want to have more citations. Google started from Stanford. Let's not forget about that.

26:53 Gary's been great at Google and he does like interacting with the audience and answering a lot of questions so definitely tweet at him but you may be trolled. If you enjoyed this episode please share it with your friends, family, networks, colleagues, bots and associates. Now Bobby's been programmed to be a quick learner and it's time for Bobby to ask me a few questions that piqued his curiosity from the episode.

27:20 Bobby Bot What was your first Google search Woj?

27:22 Woj Kwasi Um I can't say.

27:26 Bobby Bot What adjustments would you have to make with your SEO tactics if voice search becomes the new norm?

27:30 Woj Kwasi Well it's just about making sure that we've got everything tagged up on the site so it's super obvious. Gotta include structured data. Making sure that Google understands what entities are on your website. If your products need to be found, they need to be tagged as products. They need to be appearing on not just mobile devices, but any devices like Google Home.

27:52 Bobby Bot Based on Gary's explanation of AI's growing role in search, do you have any fears or reservations?

27:57 Do you trust us to be more than just your busy work?

28:00 Woj Kwasi Uh, I think you're getting better and better. Yeah, I wouldn't trust you to make emotional decisions just yet, Bobby. Wait, we're only five episodes in. There's still a way to go for you to understand this whole digital empathy thing. But once you wrap your head around that, maybe you and RankBrain will have the chance. Chicks digging. Empathy, you know. Whoa, oh my goodness. Sorry, Bobby. I didn't even realize you had company. I will, um, I'll be heading over that way, and I'll catch you later.

28:29 Bobby Bot I'll be hit nearly that way. Now, where were we? Ah yet zero zero zero zero zero.

28:37 Woj Kwasi Ooh, I didn't need to see that. In the next episode of WojCast, I interview Moz's Dr. Pete.

28:48 Dr. Pete Meyers So more content's being created all the time, more's being added to the index, and so if you do nothing, that stuff is just gonna keep flooding in. And yeah, yeah, I don't think you can stand still on

29:00 Woj Kwasi In the meantime, remember to look after each other online because empathy is organic and you can't automate empathy.

29:16 Bobby Bot Computer love.