Matt Cutts At Pubcon 2013: Moonshots, Machine Learning & The Future Of Google Search
This morning, the head of Google’s webspam team Matt Cutts gave a keynote speech at Pubcon in Las Vegas. The keynote comes on the heels of a scathing day — 1 keynote from Jason Calacanis, who said that Google rules everything, that they were essentially evil. On Twitter yesterday, Matt asked if Jason wanted the polite response, or a thorough one. All of us here in attendance are hoping for “thorough.”
Matt starts with the state of the index talking about where will Google go in future.
He’s proud that Google as doubled down on ‘moonshot’ changes, specifically:
- Knowledge Graph
Google has been trying to understand entities — not just the searches. So essentially they are trying to learn about “things not strings.”
- Voice Search
- Conversational Search
- Google Now
Matt is proud that today, sometimes you don’t even have to search to find information you need.
- Deep Learning
Google is looking more into the relationships between words. Google will be able to read at a higher level and interpret the relationships between words. Works well with voice search when a user asks Google, “Who is the Prime Minister of Turkey?” then searches again for “How old is he?” and Google can reply with the previous context.
Core Quality Changes
This change targets better natural language translation. Search is more than just matching words — instead it’s looking at specific words that are more meaningful for intelligent scoring. For instance, a voice search for “what is the capital of Texas, my Dear” the “my Dear” isn’t that important — Hummingbird will be able to detect this. While Hummingbird affected 90% of queries, it was a very subtle change that most users didn’t recognize but will help users get more pertinent results.
- Panda Softening
This is something that Google has looked into to help bring some sites and content back.
- Detecting/Boosting Authorities
Not done by hand, but applies by topic areas. Webmasters can keep deepening their content on a topic to further their authoritativeness on a specific content area.
- Smartphone Ranking
Doesn’t have flash? won’t display a site that has flash to you then.
- Penguin 2.0 & 2.1
Penguin 2.0 was released – not that intensive. Black hats said wasn’t big, so Google then released turned it up in 2.1. More changes will be continually coming, so buckle up.
- Spammy Query Algorithms
Items like porn and payday loans will be targeted for better results. Right now the SERPS aren’t great, but they will be working on it.
- Advertorials/Native Advertising
Google has cracked down on publishers selling ads that blended in as editorial with dofollow links.. You shouldn’t be paying for links that pass pagerank.
- Spam Networks
They’ve got a pretty good list, just working their way down them. Matt joked that he should talk a poll to determine who to axe next.
Google has done a great job of increasing the communication with webmasters, especially:
- New videos for malware/hacking
- Concrete examples in guidelines
- >100 speaking events, Hangouts on Air, webmaster office hours
- How search works website
Future of Search
- Machine Learning
Google’s goal is to provide the world information. The word “search engine” isn’t anywhere in their mission statement. They want to be able to give answers to specific queries.
Mobile is coming faster than anyone expected. 40% of YouTube videos are now served to mobile devices. If you haven’t thought about mobile, it’s time to start thinking about it.
Matt starts with “Facebook did a great job of social and knowing who people are.” Then talks about the fact that signal is not just likes/+1s/Tweets but in the long terms; social signals are a sign of authority. You are someone worth listening to — search engines will think you are worth listening to as well.
Next 6 months – it’s going to look like we aren’t working on much. Now working on next generation of hacking. Queries like “buy viagra” still looks bad because people are breaking the laws.
- Hot Topics
Items like child porn, international issues and really nasty queries are being addressed.
- No Toolbar PageRank scheduled for rest of year
The pipeline for updating PageRank broke this year and PageRank stopped updating. Google realized that it wasn’t that bad and stopped updating as people seem to pay too much attention to the metric. It’s something they will reassess at a later time.
Get ready, you need a mobile plan.
- Request Autocomplete
New item in Chrome that allows users to auto-fill forms. Saves users time by using the standard to pull in all information and increase chance of conversions.
- Ad-heavy pages above the fold
Some tweaks are coming to “turn up” this algorithm. Users shouldn’t see a barrage of ads above the fold when they visit a site.
- Tightening Authorship
Matt mentions that a tightening of Authorship may provide better results. Google is looking for a 15% reduction to ensure that the quality of the authorship is still high and relevant.
- Rich Snippets
The ability to have and use rich snippets may be taken away for low quality sites in the coming months
Now to the Q and A section:
Matt talks about +1’s specifically and that they are a short term signal, but very bullish on long term signal of authorship. Next Matt talks about Negative SEO. Worked on Negative SEO for years. With Penguin, it not only removes sites, but can actually have a negative effect on the site. Disavow tool announced last year, use as a last resort. Use Webmaster Tools, find links and disavow at link or domain level. Webmaster tools is now giving better backlinks, not just A-Z, so use Webmaster Tools to help identify, can see 100,000 links.
In response to Jason Calacanis’ claims from yesterday, Matt polls the crowd on whether or not to go into the matter. Crowd wants to hear the response. Matt talks about the initial version of Panda and whether or not they should have rolled out slowly. Matt says that this wouldn’t have been good and cites multiple articles showing the degrading quality of the search results. Google needed Panda. A Googler made a personal blocklist to block specific sites and nearly 200k users installed — people did not want these content farms.
In response to Jason’s claims that Google wasn’t a good partner, Matt talked about the fact that no companies have partnerships with Google. There are YouTube partnerships, not Google search partnerships. In aggregate, Mahalo simply wasn’t a quality site and they came to an impasse at a personal meeting. This wasn’t even a webspam issue, it was a quality issue and nobody received special treatment.
With the Mahalo issue behind, Matt talks about press releases. “If you are paying for PageRank, you probably aren’t doing something right.” Google has identified “a lot” of the top Press Release sites and ignores the links but doesn’t penalize those who are using them.
On infinite scrolling issues, Matt recommends using some type of paginated versions as a safety guard to index all content. On the growing size of the Google bar, Matt mentions that they are aware of the size and pixels being taken up by Google.
That’s a wrap folks.