AMA with Google Search SMX 2015: Danny Sullivan and Gary Illyes
Hello all - I have the pleasure of being at SMX Advanced 2015 currently.
Several thousand words later, I have successfully live-scribed the AMA that Danny Sullivan and Gary Illyes of Google presented for us. Sort of. Forgive me for missing the occasional banter between points, but the meat of the message is all there below.
Last year we covered Matt Cutts' Keynote with Danny Sullivan. This year we're back at it again.
Matt McGee of Search Engine Land also live blogged the event, if you'd like to see their coverage.
The conversation was fast-paced, humorous, and incredibly insightful. I did my best to capture the entire conversation as it happened. To make this conversation easier to digest, I've broken out each question topic immediately before it's discussed.
I'd also like to note that Danny and Gary joked back and forth quite a bit. I've tried to note that throughout the conversation.
Hope you enjoy, and if you have any questions feel free to leave them below or hit me up on Twitter.
Full Transcript of Gary Illyes AMA at SMX 2015
Note: the event opened with a RIP Matt Cutts in memorium. Danny mentions that there will not be a new Matt Cutts.
Question 1: The New Matt Cutts
Sullivan: So to start off, what’s it like being the new Matt Cutts?
Illyes: As you said a few moments ago--probably you’re getting older or something--there is no new Matt Cutts. There is a team taking on his role rather than a single PR person. There is a person leading the web spam team and a team leading PR. 12 years of experience “leading teams”, brilliant engineer, doesn’t want to go out and talk to people. Which we totally understand. He’s an amazing guy and I’m sure he’ll do amazing stuff with his team.
Question 2: Panda and Penguin - Update and Refresh Status
Sullivan: Since you're new to the SMX stage I thought we’d start with something easy. What’s the deal with Panda and Penguin? First we are told they regularly refreshed, then they need manual updated, etc.
Illyes: Our definition of “easy” is different. Panda – there will be a Panda update every 3-4 weeks. There is a data refresh coming in a few weeks. I don’t want to exactly say when, because I would probably set an expectation for the team they can’t meet and that would just look bad. There is a data refresh in the works.
Penguin. Penguin, that’s a hard one. We are working on making it update continuously and to not have data updates, but that’s months away. That’s huge work. We have to change many things, many many huge working things. Many months away.
Sullivan: So Panda, seems still like, are we aiming towards a month, 2 monthish, 3 monthish thing?
Illyes: Our goal is to refresh it more often. If you think about it, if we don’t update it regularly then it hurts our search quality. But, uh, yeah. It’s not that easy to refresh Panda too often. We have to get our data in order to do a data refresh and sometimes that data is either scarce, or some noise was introduced, or something went wrong with the data and we have to wait another month or two. Painstaking process. Our goal is to update more often because it hurts our search quality.
Sullivan: So the goal is to update both more often, but is Panda more often than Penguin?
Illyes: Penguin, the goal, is to have real time. But we’re not there.
Sullivan: So you’re saying that Panda will always be monthly and Penguin, the goal is to be real time, but it isn’t now [real time] and it isn’t even close.
Illyes: Yes. Panda will always need to collect data and refresh monthlyish.
Question 3: Can Google Employees See Which Algorithms Affect a Site?
Sullivan: Can John Mueller tell if a site is algorithm or manual [impacted]?
Illyes: Yes – we can see any penalty for any site.
Sullivan: Can we see it?
Question 4: Google Reporting Algorithmic Filters, More Transparency
Sullivan: Is there any hope of Google finally telling us--since you have this ability--which major algorithmic penalties or actions have been taken against us? If it happens during a 3 month period and a bunch of things update, Panda, Penguin etc. and then it’s announced, how can we know what to fix?
Illyes: No, I don’t see that happening any time soon. Don’t get your hopes up. What we are working on though, is trying to be more transparent about what we are launching. A while ago we had, on Inside Search, a blog post coming up with launches we had in the past month, or something like that. We were trying to dump there pretty much anything, and in my experience that caused more harm than good.
Sullivan: Oh I remember, that was great, you’d list like 40 things and hidden around #10 and #11 in the list were the important things you’d hope everyone would miss.
Illyes: Those people were taking apart what we said, and trying to outsmart the algorithms. What we’re trying to do is something similar to the mobile-friendly update. Or mobilegeddon as you call it.
Sullivan: Technically, we called it the mobile-friendly update until it launched, and day of launch was mobilegeddon. The next movie will be San Andreas – the mobile-friendly update II. I would have called it Platypus.
I bet next time you name an update you won’t leave it to us. We’ll name it “kitten”.
Illyes: For Pigeon, it was like, if you don’t name it something, we’ll name it “Titanic”.
(Ongoing banter about naming mobilegeddon, SEO community stoking fear, etc.)
Sullivan: Wait a second, what question was I asking you? Did you misdirect me?
Illyes: That’s what I was trying to do.
Sullivan: Ah, in the interest of transparency.
Illyes: Haha. But seriously, we are trying to do more transparency. Stuff like the blog post that announced the mobilegeddon. I mean mobile-friendly update. We are planning to do something with hacked sites.
Sullivan: So you have a “hack site” update coming?
Illyes: No. We are trying to be more transparent with notifying webmasters when we identify their site has been hacked. It’s easy for webmasters to tell if their site is hacked so there is no reason we shouldn’t be transparent about that. Likewise, there are other good “candidates” for more transparency.
Sullivan: That’s nice, because we used to have this kind of weather report type of thing, we promise to not put “-geddon” on the end of everything. So, last month we had the quality update. We said, did something happen last night? Google said no. We’re like you’re sure? Because lots of people think they’re getting hit by Panda. “Oh…. We thought you meant like something bad happening, we just adjusted how we identify site quality etc…”
Illyes: That in my opinion was a bad PR mistake. PR team assumes Barry is asking about Panda, Penguin, etc. when he asks if there has been an update. So they said no.. But there had actually been a quality update.
Quality updates we don’t typically confirm, because there’s not much we can say about it. It would be leaking stuff that could hurt us, either short term or long term. One thing I want to change there as well is to actually confirm “quality updates”. So for example if Barry asks if there is an update, we’d say, "well there was no webspam update but maybe we could say there is a core ranking update, quality update, etc." But we can’t say what it is targeting. I wrote it up to our search quality leads, and they weren’t too excited about sharing that information.
Question 5: Google's Ranking Signals and Click Through Rate
Sullivan: Okay… So can you detail what ranking signals Google uses?
Illyes: Yes. This morning I woke up and decided, “I want to lose my job.”
Sullivan: Well… you have come to the right place. Let’s take you out with a bang!
Okay. How about click through rate? We know you measure what clicks are going on. Is that part of the algorithm?
Illyes: So… We use… clicks… in a few different ways. The main things we use clicks on are evaluation and experimentation. These are the two main things. There are many many people who are trying to induce noise in clicks. One would be Rand Fishkin. Using those clicks directly in ranking, would be pretty...
Sullivan: is Rand just clicking on stuff to mess things up?
Illyes: I think what he’s doing is hiring people to click and stuff, experiment etc. Using clicks directly in ranking would not make too much sense with that noise.
Sullivan: But do you use it AT ALL?
Illyes: Okay, yes. In certain cases. Okay, let me give you an example. In certain cases it makes sense to use clicks directly. For instance if you have personalized results, and you search for apple, the first time you searched for apple we would most likely serve you a disambiguation box. Do you mean the company, or the fruit? If you had clicked on Apple the company in the past, we know you are most likely interested in Apple the company. The second time you click on Apple the company, we become more convinced that’s what you’re looking for.
If you’re a programmer, after a few searches, your searches will be dominated by the programming language results.
Sullivan: So you’re using it for personalization?
Illyes: Yes exactly, the thing [click through rates] is about personalization, if you want to mess up your own search results by randomly clicking on stuff, go ahead.
Sullivan: If I have to search my own content, see an article of mine, I’ll do it in incognito mode so I don’t give it bounces.
Illyes: Are you serious? It doesn’t matter.
Sullivan: I don’t care. I’ll still do it. I’ll cross my fingers. Does it matter?
Question 6: Google's Mobile-friendly Update and Mobilegeddon
Sullivan: Okay, so how is Mobilegeddon going?
Illyes: No idea. The mobile-friendly update was a complete success though.
Sullivan: Some people felt it didn’t happen, some felt it was dramatic to them..?
Illyes: There are two things there. Some people didn’t understand what the impact meant. I said “its impact is bigger than Panda/Penguin combined.” Because it was.
Sullivan: Yeah. Hence the geddon.
Illyes: Okay.. uh.. calm down. You know what? Aside to the audience: Last SMX he was obsessed with buttons. He wanted buttons in webmaster tools, in search results, what else?
Sullivan: Let me put it to you differently.
Illyes: Wait wait wait. The number of affected URLs was much much greater than Panda and Penguin combined. The number of search queries [affected] was much bigger as well. I think we should have said it, not just leave it to people to interpret what we said.
Sullivan: The potential impact was bigger than those things that would have hit, but afterwards people saw it as relatively minor, even though it affected more pages?
Illyes: The other thing was that sites actually did switch right before April 21st [the release of the update] to mobile-friendly, and that was a great thing. We expected ~50% impact to search results, that decreased 17% by April 21st after many people switched to mobile friendly setups prior to April 21st.
Sullivan: It feels like we’re getting to this place with Google where, no app, no need to apply. A good example is I’m searching on my phone and with app indexing it will suggest do you want to install this app, I know you don’t have it, do you want it? In contrast, when we search for a website, we don’t get a little email sign up box that appears next to the site, would you like to sign up? Or a Facebook share button in SERP that suggests you share the URL, etc.
Illyes: Sometimes you have to make an app, but it isn’t for everyone. If you have an information site, maybe it isn’t for everyone. If you have a car dealership, of course you should have an app so people can search your stock. But back to your point, we do have different universal results. For example we have universal results for images or news, apps will be kind of like that. We want to present users another way to consume content and if they like it, they can install it from the search results. Taking out one step from the user journey is actually a really good thing for the user.
Question 7: Direct Answers and Feature Snippets
Sullivan: What is the deal with direct answers? They have grown over time, last time I looked they estimated it would appear a single percentage of the time, now it’s like over 30%. Will it eat all the other results and is SEO DEAD?
Illyes: How can you phrase [say] stuff like that?!
Sullivan: You know, I say it, and you say “yes”, and we’re done! Paid search, and there we are!
Illyes: I definitely don’t think SEO is dead, I think it is always there and good for the users in general except for in the few cases where people try to game stuff with stupid crap.
Sullivan: What about smart crap?
Illyes: I don’t think there are smart craps? But it sounds cool so we might introduce it.
Yeah, we call them actually feature snippets. No geddon, it’s boring.
Sullivan: Oh I like it! I always like having an official definition so I can go back to update all my website to match it.
When a lot of people first saw this stuff happening, they thought “wow, that’s a lot of nice free content for Google, eh?” Some people now though, are saying we like the traffic this snippet gives us. What kind of traffic can people get from these?
Illyes: I don’t have the numbers, but there is a guy looking at me right now (Eric Enge), he likes them. He did a presentation on feature snippets at SMX West (and this morning, sorry).
SMXGEDDON. SMX. It sounds cool. He [Eric Enge] did a great presentation on feature snippets and pretty much nailed it. It can be extremely helpful for the site as well to have feature snippets because, one, it can send traffic to your site for example when we don’t give a complete answer and people have to click through for it. You get a sense as a user that your answer will be answered on that page. Another thing is with branding. Branding, is pretty hard to measure and that’s one problem we have. Our ultimate goal is to actually have data about featured snippets as well in Google Webmaster Tools.
Sullivan: Oh yeah, we could have a button that says “Google Featured Snippet Site”. It could be a button. We’ll make our own.
Do we get a benefit by the way, if we are picked as a feature snippet? Is that a sign that Google likes us, that more authority is passed to our site more than others?
Illyes: No. It means your site provided an answer to a user query that was good.
Sullivan: …. You could have just lied. Made us so happy. By the way, is there an opt out if people don’t want to be in a feature snippet?
Illyes: No, these are snippets we show in the search results, so no.
Sullivan: Oh! So Moz did something great. They made a page, they made the answer you didn’t have, then they took the page down, but you still remembered the answer.
Illyes: If the page is still indexed, then yes we will show it still.
Sullivan: But when you get these answers to these things do they go off into the knowledge graph etc?
Sullivan: By the way, if you search “king of America” we are a featured snippet. It’s pretty awesome. It’s Obama. Check it out! Don’t break it. It’s pretty cool. Don’t worry about that.
Illyes: Don’t worry, if you were to Bing that, it’d just be like… I dunno!
Question 9: Search Console, Previously Google Webmaster Tools
Sullivan: So, will we get anything [new] in the search console, which is Google Webmaster Tool's old name, which is webmaster central’s old name?
Illyes: It’s been webmaster tools for 10 years.
Sullivan: In search console, you said we might get something that shows us feature snippet stuff?
Illyes: I really hope we can get some data there. We are looking into how we can provide that data. The thing is it’s just not that easy to interpret the data and we’d have to give it in a way that is not confusing. And is easy to interpret. It’s like, brand recognition, how do you track that?
Sullivan: We can come up with something.
Illyes: No, it’s better that you don’t.
Question 10: Google's Ability to Recognize Entities
Sullivan: Let me ask you about entities.
When Google knows about a thing, a business, a brand. I’m wondering how strong a signal this is versus links? What I mean is, in the past, Google wanted to know about the content and also how it was being linked to. Now we know with the Pigeon update, [and] Google knows a lot about businesses, even if they aren’t being linked to. John Mueller was saying that possibly you guys are looking at brand entities for web page ranking so that you would understand, say, something about SEL because someone said something about SEL in the NYTimes and NYTimes didn’t link to SEL, because they think links are against the law… Does Google know [about this connection]?
Illyes: Okay, so most of that comes from structure data and wiki data. If there is no structure data, it would be a pretty hard thing to extract entities. In general if you have structure data then we can add structure data to our database or something like that, but if there is no structure data it’s pretty difficult. However, we’re looking, [for example] like stuff that looks like a link but isn’t a link. If we see searchengineland.com but it isn’t a link, we might treat it like a nofollow [link]. We might use it for discovery, and say, we should visit that site because it is a URL from an authoritative site but not a link, but we wouldn’t pass PageRank.
Sullivan: So does that mean you’ll use reputation signals like tweeting, social following, followers, etc.?
Illyes: So I think right now we are not doing anything like that. We are in early stages of Twitter indexation. In the past it was really hard when we integrated social engine content in our real time content search results. I don’t think we want to go down that path again. Also, in general, we want signals that are stable and are not that noisy. As that’s able, if someone closes that firehose, then our signals would just disappear.
Sullivan: Well you’ve always got G+ to fall back on. Signal is there, it’s stable. Facebook likes, they’re important because they help you rank.
Sullivan: Just a test. They don’t. They don’t.
Illyes: Okay. So webmasters, server managers, they could robot out, [disallow via robots.txt] the Facebook "like" code and we’d be without that signal. We need signals that are stable for long term, otherwise our ranking would be messed up pretty bad.
Question 11: Google's Ability to Crawl Ajax
Sullivan: You said you’re going to depreciate crawling ajax soon. What’s happening? What are people going to do instead?
Illyes: I said we are deprecating ajax crawl. Since we are rendering the content anyways, the ajax crawling scheme says you should render the content on your site, we just want to save you from doing that. We would be able to render it on our end without the webmaster doing it, but … (lost this part in my notes. It seemed like this was a very technical and complex issue for Gary to explain and he had to keep going back on his statements as he made them. Complex issue to explain in a quick overview). We want to make sure we are understanding all the crazy ajax stuff, then we’ll release a blog post for people to learn from once we understand enough of it. (I took this to mean that soon, Google will release a clear and transparent blog post that will help webmasters better understand the nuances of this soon).
Question 12: Global TLDs
Sullivan: As more global TLDs get introduced, how will Google take this into account? And are you going to do anything special with your own things, like searchengineland.google will we get a ranking boost?
Illyes: No, but if you put meta value name PR = 10, your PageRank will just shoot right up. (Editor note: He's joking people, joking.)
Nothing changes with new TLDs. They are treated just like .com. No special benefit.
Question 13: URL Structure in Search Rankings
Sullivan: How important is URL structure in today’s ranking? Do you care if something is buried in a website? Do we need descriptive URLS etc?
Illyes: The length of URL does not matter. Don’t pass the HTTP limit. The structure kind of does matter, but for discovery not ranking. For example if something is far away from the root of the domain it kind of signals it’s less important to us as a content owner than things that are close to the root of the domain. Of course, please fill them with important keywords (sarcasm). For example, if you link important content from your home page, that also signals you think it is important. That is for discovery, it will be prioritized during ranking.
Sullivan: So if we have 1000 pages we should have 1000 links on our home page of our domain.
Illyes: Yes (sarcasm).
Question 14: Google's Crawl Factors
Sullivan: Okay. So once I had 8k out of 10k [pages] crawled, the next day it’s only 750. What’s it mean?
Illyes: So, we know about over 30 thousand trillion URLs. I know the expectation is to index them all but we don’t have enough storage to index the content from them all. Sometimes we have to get rid of our stuff from our index that we perceive people won’t ever search for.
Sullivan: My international content is .com and .com.cn with over 100 pages each in their native languages. Google Search Console shows 15k per day for .com, but .cn is getting 150k per day. Why so different?
Illyes: Number of the crawled URLs, or number of URLs we can crawl per day comes from like probably 5 factors. Most important is PageRank from other external URLs. It is an indicator of importance so we crawl more because people link to it. Another is actual host load, if we perceive we can crawl as much as we want to from a server, but we’re pushing the server to it’s limit and slowing it down, we’ll back off from the server. Could be the .cn server can handle more traffic than the .com. It’s probably just we can’t crawl as much as we want to.
Sullivan: Does Google use user behavior stuff when determining crawl?
Illyes: Same thing as what I said about rankings, it is very noisy. I’m not sure if anyone at Google can make sense of that data.
Sullivan: What about Google Analytics?
Illyes: No. I can do this – I can say we don’t use Google Analytics.
Question 15: Panda and Duplicate Content
Sullivan: Does the Panda duplicate content penalty only stem from text on the page, or video?
Illyes: There is no duplicate content penalty, only Panda.
Sullivan: Okay, what about any penalty? Do you care about video content embedded on a page?
Illyes: In general, no. Not for web search. But you can’t ignore video search. Video search will try to index the video and serve it up to the user if the query is video seeking. We might surface video universal in web search if we think the video is a good match. If you can produce videos, it’s important to link it properly, structure it properly, I.E. VIDEO XML sitemaps. Because you can get quite nice traffic from video universal results.
Question 17: Duplicate Content
Sullivan: Let me go back to duplicate content because there are like a bazillion questions, probably all from the same person. You don’t have a duplicate content penalty, but sometimes you see things that are similar and split the difference, only rank one because you don’t want dupe results in the SERP.
Illyes: Okay, when you have 3 URLs with the same content, you only want to serve one in the search results.
Now, in other issues you have problems where two different domains are similar but no canonical tag.
I think you don’t have to do that, we can figure it out.
Sullivan: So people shouldn’t worry about it?
Illyes: I think there are better things to worry about. In general we can figure out which URL should be served in the search results. If you are not happy with our decision then you can use rel=canonical to influence that decision.
Question 18: Slow Loading Pages Flagged in Search
Sullivan: Any plans for flags like slow or not mobile-friendly?
Illyes: Not mobile-friendly, we tried, people didn’t like it. Experiments showed that people don’t like it. We also ran user studies and we were testing with a bunch of different icons and text, if you put a big ol’ red phone there, people will think you click it and you’ll call the company. Not what we wanted. A few countries we’re running an experiment with transcoding, in those countries I think we do label some sites with a “slow” tag… Not sure what the status of the experiment is.
Question 19: SSL as a Ranking Factor
Sullivan: Do you think SSL will have a larger impact in ranking?
Illyes: I do hope, yes. For me, secure search is super important and I do want to improve that algorithm. I have a bunch of ideas, but really I should just spend more time actually working so I can do actual work. Yes, I do have plans to improve that and make it stronger.
Sullivan: A lot of sites will have overlays come up, is that a problem ranking?
Illyes: Right now, no.
Illyes: No. Just go. [next question, this was during the "lightning round" Q&A].
Question 20: Author Rank
Sullivan: Author rank, using that for normal search results?
Illyes: I don’t know. I don’t want to say something stupid or not true. But, when we deprecated authorship images, we didn’t deprecate the whole other authorship concept. I would imagine that data would be very valuable for certain other things. I can say with some certainty that yes, we are using it. (Author's note: This is kind of a big deal folks. We all know that authorship images are gone. We have only conjectured that the data is still being used in other ways).
Question 21: How Does Google Determine Quality Content?
Sullivan: How do know “what quality content is?”
Illyes: If you go to SEL for example, there are columns. You go to SEMrush, there is good content there. Something that is valuable. If it is valuable for the user, if it can satisfy an information need, then users love that content, that is quality content. If you have quality content, users will link to it. On Facebook people will like, G+ will plus +1, Twitter will share, etc. I’m not sure why we are actually discussing this question.
Sullivan: Because everyone wants to know what “quality” is in Google.
Illyes: It bothers me because it is so obvious. Something that satisfies user’s information need, that is quality. If people like, share, link your content – that is quality content.
Sullivan: If you are trying to find a proxy for quality, if a lot of people are sharing your content (and you didn’t buy those shares), then that’s a good indicator.
Question 22: Emojis in Search
Sullivan: Ok, quick question. It’s all down to you. When do the emojis die in search results?
Illyes: Hopefully quite soon.
Sullivan: Search for Panda update, we got a little panda in ours!
Illyes: Generally thinking about it, it looks like a Christmas tree. I asked the Titles Team (Author note: it was cool to hear Gary talk about the different "teams" that exist at Google and their various responsibilities), and they were like, “uh… ummm….. eh….”
Sullivan: Ok well I want more emojis and animated gifs in my search results as well.
Illyes: We have that! In knowledge panels!