Eric Enge – Search Engine Land News On Search Engines, Search Engine Optimization (SEO) & Search Engine Marketing (SEM) Fri, 27 Aug 2021 22:20:11 +0000 en-US hourly 1 Ask the expert: Demystifying AI and Machine Learning in search Thu, 26 Aug 2021 16:02:46 +0000 Artificial intelligence encompasses multiple concepts, deep learning is a subset of machine learning, and natural language processing uses a wide range of AI algorithms to better understand language.

The post Ask the expert: Demystifying AI and Machine Learning in search appeared first on Search Engine Land.

The world of AI and Machine Learning has many layers and can be quite complex to learn. Many terms are out there and unless you have a basic understanding of the landscape it can be quite confusing. In this article, expert Eric Enge will introduce the basic concepts and try to demystify it all for you. This is also the first of a four-part article series to cover many of the more interesting aspects of the AI landscape.

The other three articles in this series will be:

  • Introduction to Natural Language Processing
  • GPT-3: What It Is and How to Leverage It
  • Current Google AI Algorithms: Rankbrain, BERT, MUM, and SMITH

Basic background on AI

There are so many different terms that it can be hard to sort out what they all mean. So let’s start with some definitions:

  • Artificial Intelligence – This refers to intelligence possessed/demonstrated by machines, as opposed to natural intelligence, which is what we see in humans and other animals.
  • Artificial General Intelligence (AGI) – This is a level of intelligence where machines are able to address any task that a human can. It does not exist yet, but many are striving to create it.
  • Machine Learning – This is a subset of AI that uses data and iterative testing to learn how to perform specific tasks.
  • Deep Learning – This is a subset of machine learning that leverages highly complex neural networks to solve more complex machine learning problems.
  • Natural Language Processing (NLP) – This is the field of AI-focused specifically on processing and understanding language.
  • Neural Networks – This is one of the more popular types of machine learning algorithms which attempts to model the way that neurons interact in the brain.

These are all closely related and it’s helpful to see how they all fit together:

In summary, Artificial intelligence encompasses all of these concepts, deep learning is a subset of machine learning, and natural language processing uses a wide range of AI algorithms to better understand language.

Sample illustration of how a neural network works

There are many different types of machine learning algorithms. The most well-known of these are neural network algorithms and to provide you with a little context that’s what I’ll cover next.

Consider the problem of determining the salary for an employee. For example, what do we pay someone with 10 years of experience? To answer that question we can collect some data on what others are being paid and their years of experience, and that might look like this:

With data like this we can easily calculate what this particular employee should get paid by creating a line graph:

For this particular person, it suggests a salary of a little over $90,000 per year. However, we can all quickly recognize that this is not really a sufficient view as we also need to consider the nature of the job and the performance level of the employee. Introducing those two variables will lead us to a data chart more like this one:

It’s a much tougher problem to solve but one that machine learning can do relatively easily. Yet, we’re not really done with adding complexity to the factors that impact salaries, as where you are located also has a large impact.  For example, San Francisco Bay Area jobs in technology pay significantly more than the same jobs in many other parts of the country, in large part due to the large differences in the cost of living.

Vector isolated illustration of simplified administrative map of USA (United States of America). Borders and names of the states (regions). Grey silhouettes. White outline.

The basic approach that neural networks would use is to guess at the correct equation using the variables (job, years experience, performance level) and calculating the potential salary using that equation and seeing how well it matches our real-world data. This process is how neural networks are tuned and it is referred to as “gradient descent”. The simple English way to explain it would be to call it “successive approximation.”

The original salary data is what a neural network would use as “training data” so that it can know when it has built an algorithm that matches up with real-world experience. Let’s walk through a simple example starting with our original data set with just the years of experience and the salary data.

To keep our example simpler, let’s assume that the neural network that we’ll use for this understands that 0 years of experience equates to $45,000 in salary and that the basic form of the equation should be: Salary = Years of Service * X + $45,000.  We need to work out the value of X to come up with the right equation to use.  As a first step, the neural network might guess that the value of X is $1,500. In practice, these algorithms make these initial guesses randomly, but this will do for now. Here is what we get when we try a value of $1500:

As we can see from the resulting data, the calculated values are too low. Neural networks are designed to compare the calculated values with the real values and provide that as feedback which can then be used to try a second guess at what the correct answer is.  For our illustration, let’s have $3,000 be our next guess as the correct value for X. Here is what we get this time:

As we can see our results have improved, which is good! However, we still need to guess again because we’re not close enough to the right values. So, let’s try a guess of $6000 this time:

Interestingly, we now see that our margin of error has increased slightly, but we’re now too high! Perhaps we need to adjust our equations back down a bit. Let’s try $4500:

Now we see we’re quite close! We can keep trying additional values to see how much more we can improve the results. This brings into play another key value in machine learning which is how precise we want our algorithm to be and when do we stop iterating. But for purposes of our example here we’re close enough and hopefully you have an idea of how all this works.

Our example machine learning exercise had an extremely simple algorithm to build as we only needed to derive an equation in this form: Salary = Years of Service * X + $45,000 (aka y = mx + b). However, if we were trying to calculate a true salary algorithm that takes into all the factors that impact user salaries we would need:

  • a much larger data set to use as our training data
  • to build a much more complex algorithm

You can see how machine learning models can rapidly become highly complex. Imagine the complexities when we’re dealing with something on the scale of natural language processing!

Other types of basic machine learning algorithms

The machine learning example shared above is an example of what we call “supervised machine learning.” We call it supervised because we provided a training data set that contained target output values and the algorithm was able to use that to produce an equation that would generate the same (or close to the same) output results. There is also a class of machine learning algorithms that perform “unsupervised machine learning.”

With this class of algorithms, we still provide an input data set but don’t provide examples of the output data. The machine learning algorithms need to review the data and find meaning within the data on their own. This may sound scarily like human intelligence, but no, we’re not quite there yet. Let’s illustrate with two examples of this type of machine learning in the world.

One example of unsupervised machine learning is Google News. Google has the systems to discover articles getting the most traffic from hot new search queries that appear to be driven by new events. But how does it know that all the articles are on the same topic? While it can do traditional relevance matching the way they do in regular search in Google News this is done by algorithms that help them determine similarity between pieces of content.

As shown in the example image above, Google has successfully grouped numerous articles on the passage of the infrastructure bill on August 10th, 2021. As you might expect, each article that is focused on describing the event and the bill itself likely have substantial similarities in content. Recognizing these similarities and identifying articles is also an example of unsupervised machine learning in action.

Another interesting class of machine learning is what we call “recommender systems.”  We see this in the real world on e-commerce sites like Amazon, or on movie sites like Netflix. On Amazon, we may see “Frequently Bought Together” underneath a listing on a product page.  On other sites, this might be labeled something like “People who bought this also bought this.”

Movie sites like Netflix use similar systems to make movie recommendations to you. These might be based on specified preferences, movies you’ve rated, or your movie selection history. One popular approach to this is to compare the movies you’ve watched and rated highly with movies that have been watched and rated similarly by other users.

For example, if you’ve rated 4 action movies quite highly, and a different user (who we’ll call John) also rates action movies highly, the system might recommend to you other movies that John has watched but that you haven’t. This general approach is what is called “collaborative filtering” and is one of several approaches to building a recommender system.

Note: Thanks to Chris Penn for reviewing this article and providing guidance.

The post Ask the expert: Demystifying AI and Machine Learning in search appeared first on Search Engine Land.

SMX Overtime: Opportunities and challenges for conversational voice search Wed, 18 Dec 2019 20:02:24 +0000 Expert search marketer and SMX speaker, Eric Enge, explains why brands need to better understand voice apps and interactions to meet their user needs.

The post SMX Overtime: Opportunities and challenges for conversational voice search appeared first on Search Engine Land.


At the SMX East conference I spoke on the topic of voice, and in particular, I discussed how to go about building a personal assistant app. During the Q&A for that session there were a few questions that went unanswered, and in today’s post I will address five of them. In the first of those questions I will go into great depth into my thoughts on the total amount of overall voice search usage today. Check it out because you might see the data that you expect!

1. To compete with others on Google search, do we need to build an app for voice search?

While I’m a fan of building personal assistant apps, my experience suggests that the usage level for most such apps remains relatively small, with some notable exceptions. I do think that it makes sense for a large number of organizations to have a voice app program, but for most of these, it’s about gaining experience in voice while it’s still relatively early days in this new arena.

I believe you should be experimenting with voice today. Not because voice is huge right now, but because it will be and it’s coming fast. In other words, having a personal assistant app is not a requirement to compete in Google search, but in a few years, it likely will be.

So why start now? Simple: voice as a source of input will change user expectations of how the dialogue between humans and machines should work, and learning how to deal with speech recognition and natural language processing is hard. You will need to experiment with this and learn how it works and it will take time to build that expertise.

Many of you are probably thinking to yourselves that you’ve seen many statistics that suggest that voice is already huge, and you’re wondering if I’m simply out of touch with what’s happening. However, the reality is that those statistics are deceptive (and in some cases completely misquoted). There are three basic reasons why this is true:

  1. Frequently cited metrics about voice search are inaccurate.
  2. The nature of voice queries is quite different than the make up of typed search queries.
  3. Personal assistant usage is only a portion of total voice search

To address the issue of voice search volume, little real data is published by anyone on this topic. We have the oft-quoted prediction about voice search making up 50% of all search by 2020 that is often attributed to Comscore. First of all, the prediction was made by Andrew Ng, who at the time was Chief Scientist at Baidu.

Second of all, his prediction was for the combination of voice and visual search. And thirdly, the available data that we have suggests that his prediction was greatly exaggerated.

We also have Sundar Pichai’s keynote at Google I/O where he noted:

“in the U.S. on our mobile app and Android, one in five queries, 20% of our queries are voice queries and their share is growing.”

Twenty percent of queries already sounds like a great number, but let’s break that down a bit more:

  1. This did not include desktop queries, where the voice query share is probably close to zero
  2. This did not include queries via browser on Android devices, and it maybe that significantly more queries happen in regular browsers than the Google Assistant
  3. This did not include Google queries that take place in iOS devices, where installations of the Googl Assistant are probably quite low (so nearly all Google queries happen via browser)

Even so, it still seems like it might be a significant number, right? But now let’s dive into the makeup of those queries. We’ll start by looking at Bryson Meunier’s analysis of 3,000 voice queries and Greg Sterling’s report on consumer survey performed by NPR and Edison Research. First, let’s have a look at data from the usage of a Google Home by Bryson’s family.

Click to enlarge

Between playing music, telling Google Home to stop, setting timers, home control and adjusting the volume, we have 72% of all the queries. There is not much trace of anything actionable for the great majority of businesses with web sites (unless you run a music service).

Bryson breaks this down further by analyzing the overall query intents. He uses the intents as they are defined in the Google Search Quality Raters Guidelines (the full definitions start at page 71). A brief outline of their definitions is as follows:

  1. Know – the user is seeking information on a topic
  2. Know Simple – a special case of a Know query where all the user wants is a simple fact
  3. Do – queries that indicate that the user wants to do something
  4. Do Device Action – a special case of Do queries, where the device the user is using is able to complete the action on their behalf (such as Play Music queries)
  5. Website – locate a specific website or web page
  6. Visit in Person – when the user wants to go somewhere specifically
Click to enlarge

Looking at this one family’s data set, we don’t see a ton of opportunity for most organizations to gather in lots of search traffic. The basic reason is that a lot of the usage is for new types of use cases that Google can satisfy without needing third party assistance (i.e. your web site).

One of the interesting data points was the dayparting graphic created by NPR that shows how usage of smart speakers varies throughout the day.

Looking at these queries we see that these mostly do not look like traditional search queries these are as well. Further, this is what Greg Sterling had to say about this data:

It’s a broad and diversified mix, though actions/skills discovery remains a problem on both Google Home and Alexa devices. People are not entirely sure about all the things smart speakers can do, and there’s no great discovery mechanism right now.

For one last source, let’s consider data from a voice usage report from PWC.

Cumulatively, what these three different sources confirm for us is that the voice revolution hasn’t fully taken over yet. We’re seeing a significant number of things being done by voice, but a lot of them are NEW or fall into highly repetitive actions, such as paying music, getting directions, setting timers and the like. So they add a new dimension to total search volume, but not much opportunity for most businesses.

The key obstacles to a broader level of adoption, including shifting more traditional search queries to voice are:

  1. While the reported accuracy of speech recognition is the same as it is for humans (around 95%) the types of errors that happen with voice queries are quite different and often very frustrating.
  2. Speech recognition is not the only issue and in fact the more important one is natural language processing. The major personal assistants still have a LOT of work to do here.
  3. Users still do not fully understand the wide range of capabilities that personal assistants offer, and as Greg Sterling noted, discovering new capabilities is not easily done.

All three of these areas need to improve for voice to realize its full potential. Those improvements are coming, and we will get there, but it will take some time.

But brands need to work on developing their understanding of voice apps and interactions, and how to build apps to meet the related user needs. This is something that I’d start working on today. Not to compete on Google today, but to be ready to compete on Google tomorrow.

2. How do you think about using speakable schema for the voice contents and its possible results?

I’m glad that this question was asked because there has been a big change recently. It used to be that speakable markup applied only to news sites. In fact, if you go to the page on on speakable markup, you will still see the following:

To be eligible to appear in news results, your site must be a valid news site. Make sure you submit your news site to Google either through the Publisher Center or setting up a valid edition in Google News Publisher Center.

So as of Dec. 12, Google has let us know that it will look for speakable markup on sites, so go ahead and implement it! Of course, this does not guarantee that you will be used in voice search results, but it likely increases the chances that you will.

3. What is the strategy for voice for verticals like the weather, where the answer gives no credit to the source or drives brand awareness?

That’s a tough one. If your goal is to earn a featured snippet (and therefore presence in voice results) for a query like “Falmouth Weather,” it’s going to be very tough for you unless you are the publisher of, or Also, Google does actually provide attribution when a screen is available:

However, as the question suggests, Google does not provide attibution when delivering the results through voice. I discussed this with Barry Schwartz, and we both believe that Google has a deal with for the weather results, so the way this behaves is likely the way it was specified in that deal.

So to be clear, the verticals that I know of where there is no attribution provided, I believe that those are the direct result of a negotation between the source of the content and Google. These scenarios should not impact your overall voice strategy.

Unless you are in a position to negotiate a specific deal with Google, you should consider how you structure the deal to get the best result for yourself. However, if you’re not in a position to negotiate such a deal there is probably not much you can do to rank for voice search on these results because someone else (like will.

4. How hard was the implementation of Google Assistant/Alexa tracking in Google Analytics?

As I noted in the presentation I did on voice at SMX East on building voice apps, it’s possible to track the usage of your personal assistant apps in Google Analytics or Adobe Analytics.

It does require a certain amount of programming skill to setup involving integration via API calls. A reasonably skilled programmer can probably work out how to do this with a few weeks of programming effort (including testing and debugging).

5. When should you use your own voice app, and when should you use existing tools (like Google My Business) to get your information to users?

I’d phrase the question a bit differently. The real question is when you should do both. If have physical locations for your business and you want to bring foot traffic, you should be in Google My Business. It’s that simple. Further, if there are opportunities on those same queries to be featured in voice related results, which you can now enhance your chances of doing by earning featured snippets or using speakable markup, you should do that too.

The post SMX Overtime: Opportunities and challenges for conversational voice search appeared first on Search Engine Land.

Bing’s holistic view of search on display at SMX West Wed, 20 Feb 2019 18:08:21 +0000 Bing's keynote at the San Jose event took a closer look at search integration for internal documents and how intelligent image search will expand.

The post Bing’s holistic view of search on display at SMX West appeared first on Search Engine Land.

The second day of SMX West began with a keynote by the Bing team from Microsoft. Speakers Christi Olson and Junaid Ahmed focused on outlining Bing’s view of how search should work, their general approach to it and features they have in their search engine that Google doesn’t have. The Bing team discussed some of this previously (read my recap from SMX East recap here) but there were several new projects discussed.

Bing’s holistic approach to search overall strives to improve their organic results, and for the ads, their goal is to be as helpful as possible. To do that, Bing made 75 significant changes in 2018.

Within Bing Ads, two big areas for investment were the introduction of Bing Ads Scripts and improvements to their Google Import capabilities.

Bing Ads is also moving away from keywords and towards intelligent audience marketing with its Microsoft Audience Ads program. This includes features such as in-market audiences, LinkedIn profile targeting and the ability to address multiple language targets within one ad group.

Bing ads chart

Also, the Microsoft Audience Network is benefitting from their recent deal with Verizon to access all their properties, including native advertising across the Yahoo network. This also gives Bing Ads access to, the Huffington Post and provides them with a strong mobile market share. Microsoft is planning to complete its integration with the Verizon Media network by March 31.

AI and the scale of the Bing Knowledge Graph, which is about 5 billion entities, has the goal to build a better semantic understanding of each phrase. Trying to discern what a user wants based solely on a two or three-word phrase is a task is one of the most challenging aspects of search.

Christi highlighted the Microsoft Search Graph Integration for Businesses that allows Bing to do more than search the web. It can also search internal business networks with information in Microsoft office docs, spreadsheets, Powerpoint files, and Sharepoint sites, all in one seamless experience with web search.

Junaid, who was new to presenting at SMX as he spends more time at machine learning conferences, explained how Bing’s goals align with SEO goals.

Bing’s QnA capability handles search refinements to deliver multi-perspective answers. They have partnered with the Trust Project to increase news transparency. This includes a new type of schema that many are not using yet, but they’re promoting it. Bing can provide multi-perspective responses and not only does this with regular search results, but it is extended to news results as well.

Real-time indexing is a big push at Bing, too. When a presentation followed by a blog post should mean that by the time someone wants to do a search, it should be there in the results. For example, when the wrong Oscar winner gets announced you’re going want to get that info NOW.

Bing is also expanding what they’re doing in both image and visual search. One example is the use of a camera as a direct input into the Bing mobile app. Imagine that you are in Paris and walked to dinner, had a nice meal and now you’re walking back. You should be able to take a picture of the bridge you’re approaching and use visual search to identify the bridge as well as other nearby landmarks within the image. Another image search feature tells you how many pages on the web include a particular image.

Bing is addressing over-crawling issues and outlined how the volumes of their crawls declined over time.

With the ability to submit up to 10,000 URLs per day to Bing (read more about this announcement here),the need to crawl is eliminated so the webmaster of a site can let Bing know when something has changed.

To facilitate this, Bing will interact directly with larger companies (e.g., Amazon). They see this as a major shift in how search engines work. To that end, the current limit of 10,000 is just a number, and this could get increased over time.

The post Bing’s holistic view of search on display at SMX West appeared first on Search Engine Land.

Solving complex SEO problems require a new ‘discovery’ approach Fri, 09 Nov 2018 15:24:32 +0000 The SMX presentation with Hannah Thorpe and Arsen Rabinovitch reviewed Google's latest updates along with diagnostics and tools to get your site back on track.

The post Solving complex SEO problems require a new ‘discovery’ approach appeared first on Search Engine Land.

Image of the statue of liberty and other NYC landmarks at the SMX East show in New York City

At SMX East I attended the “Solving Complex SEO Problems When Standard Fixes Don’t Apply” session with presenters Hannah Thorpe, Head of SEO Strategy at Found, and Arsen Rabinovitch, Founder and CEO at Here are key learnings from each presentation.

Hannah Thorpe highlights Google changes and the user discovery experience

Hannah starts by sharing a slide with an SEO checklist used by Rand Fishkin in a past episode of Whiteboard Friday:

Hand-written checklist for SEO

The problem is that a lot has changed since Rand published that Whiteboard Friday. One aspect of what has changed is the advent of neural matching. For example, according to Danny Sullivan, 30% of all queries use neural matching, as shown in this example query.

Screenshot of Danny Sullivan tweet on neural matching

In this example, we see a user asking why their TV looks strange, matched up with the common label for that (the “soap opera effect”). This is an example of neural matching in action.

Another major change is that Google is becoming a discovery engine, where the focus is more on discovery than search. This is quite literally true in the case of the Google feed. Another significant change is the addition of the topic layer.

mobile screenshot showing search results for Yorkshire terrier

In this result, above the knowledge graph, you can see the most relevant sub-topics to the query result.

Another significant shift is driven by Google’s increased understanding of entities. This enables Google to do a far better job of finding the content that is most relevant to the user’s query, and in fact, their actual needs (Note: my view is that this is not a deprecation of links, but an improved methodology for determining relevance).

Hannah then provides us with a table comparing the differing natures of search and discovery.

Chart with words distinguishing between search and discovery

As SEOs, we tend to optimize for search, not discovery, where the entity holds a lot of the value. With this in mind, we need to start thinking about other ways to approach optimization, as suggested by Rand Fishkin.

screen shot for tweet from Rand Fishkin about SEO goals

Hannah then recommends that we attack this new SEO landscape with a three-pronged approach:

  1. Technical Excellence
  2. Well-Structured Content
  3. Being the Best

Use your technical expertise to reevaluate your focus

The need for this is a given, but there are four causes of failure for audits:

  1. Sporadic implementation
  2. A changing ecosystem
  3. No change control process
  4. Too many people involved

How do you get better audits? It’s about your focus. Think about these things:

  1. Being actionable
  2. Prioritize time vs. impact
  3. Understand the individuals involved
  4. Quantify everything

Be smart about your structured content

This is about making it easier for users to find what they want. Here’s a smart way to think about formatting your content:

Image illustrating how to structure content for SEO

Basically, you want to make the main point (answer the question) in the first four lines of the content. Expand upon that with examples, and then comment on that.

The specifics of how you format the content matters too. Some basic ideas on how to do this are:

  1. Avoid large blobs of text – keep paragraphs short
  2. Keep tables clean and avoid large chunks of text in them
  3. Avoid excessive text in bulleted lists, and don’t double space them either
  4. Add structured markup to your content

Be the best in your own space

This is critical to success in today’s world. It doesn’t mean that you need to be the best at everything, but carve out your space and then own it.

This starts by understanding where the user is on their journey when they reach your site. What are their needs at the moment that they arrive on your web page? Understand all the possible circumstances for visitors to your page, and then devise a strategy to address the best mix of those situations possible.

Remember, Google is focusing on users, not websites. You need to focus on your target users too.


Arsen Rabinovitch outlines how to evaluate SEO problems and what tools can get you back on track

This was a big year for Google updates and that prompted changes in traffic, some significantly, for a lot of sites.

Here’s a short summary of the updates that occurred.

timeline of Google's algorithm updates related to quality

For some sites, the impact of this was quite large. You can see an example in this SEM Rush chart of one site that suffered two large drops during the year.

Once you have experienced significant traffic loss, it’s time for SEO triage. The first step is to look at the symptoms.

  1. We lost positions
  2. We lost organic traffic
  3. We lost conversions

It’s also helpful to create a timeline, as shown here.

Once this initial research is complete, it’s time to do some diagnostics. It’s helpful to start by ruling out as many potential causes as possible. Some of the things to check for are.

  1. Check for manual actions
  2. Look to see if the symptoms are environmental
  3. Check if the symptoms just started presenting themselves
  4. Determine if the symptoms are chronic

It’s also useful to see what’s being affected. Is it the entire site or just some pages? You should also check to see if competitors are affected as well.

Further digging can involve checking which queries lost clicks (or rankings). You can also use third-party tools such as SEMRush to confirm your findings. Make sure to also check the URL inspection reporting in the new GSC to see if Google is reporting any problems there. Check to see if the SERPs themselves changed. You can also check the index coverage report, and look for URLs that are indexed but not in sitemap.

Checking coverage reporting is also a good idea. Some things to look for there include:

  1. Look for large chunks of pages moving in or out
  2. Blocked by robots.txt
  3. Excluded by NoIndex tag
  4. Crawled – currently not indexed
  5. Alternate page with proper canonical tag

Search Console is truly a goldmine of diagnostics, and here are some more things you can look into:

  1. Check crawl stats
  2. Check blocked resources report
  3. Check html improvements report

Another rich area of information is your web server logs. These are files kept by your web hosting company that contain a ton of information in search and every visitor to your web site. Things to look for here include:

  1. Weird status codes
  2. Spider traps
  3. Performance issues
  4. Intermittent alternating status codes

Don’t forget to look at your backlinks, as you may spot problems there that impact the entire site. Use a crawler tool such as Screaming Frog or DeepCrawl to crawl. Identify and fix all the issues you find there, and look for pages that are not getting crawled by your crawler and find out why.

The post Solving complex SEO problems require a new ‘discovery’ approach appeared first on Search Engine Land.

The age of voice assistance and a new input paradigm Fri, 02 Nov 2018 15:38:04 +0000 SMX East's opening keynote from Google's Naomi Makofsky outlined how marketers can find opportunities for the explosion of conversationally powered experiences with customers.

The post The age of voice assistance and a new input paradigm appeared first on Search Engine Land.


NEW YORK – On the second day of SMX East, the opening keynote from Google’s Naomi Makofsky entitled, “The Age of Assistance” reminded us that people are connected to more devices than ever before, and that more and more of these interactions are conversationally powered.

We’ve gotten here as a result of a long evolution. In the 1970s, we had the revolution of personal computers, where input was driven by typing. By the 1980s, more and more people had home computers, and we had the mouse, so a new input format was clicking.

Then came the next major shift with the advent of smartphones and another new input paradigm came forth—tapping.

Now we’re moving into a new environment where voice becomes a major input paradigm. With that comes the concept of proactive assistance provided by the machine, and the idea of helping people get things done. This involves anticipating the consumer need or intent, to get assistance to the right person at the right time.

Looking at the evolution of voice search itself, we can see that voice search from Google launched in 2009:

The Knowledge Graph debuted around 2012, and today has over one billion entities. Google Now, the first assistant, followed later that year, and it tried to anticipate user needs.

Today, personal assistants need to work across a widening landscape of different devices, and help users with a wide variety of needs, from helping them get from point A to point B to booking a restaurant.

This involves more than running on and seamlessly transitioning between mobile devices—users may be connecting via their refrigerators, vacuums or thermostats.

So how fast will all this happen? According to eMarketer, 87 percent of B2C marketers believe that personal assistants will play a big role before 2021. Can it really move that fast?

Consider the rate at which the smartphone became a core part of our world. In 2008, no one had a smartphone. Naomi takes a quick poll of the audience, and demonstrates that by the time of her keynote (at 9 a.m.), nearly everyone in the audience had already either used an app ride service such as Uber of Lyft, checked their social media, or sent one or more texts.

The bottom line is that this can happen fast, so get ready for it!

Today, Google Assistant is installed on more than 500 million devices. Its focus is to help you get things done. For business owners, our opportunity is to create “Actions,” which is how we can offer our own customers opportunities for conversational experiences with our brand.

One of the driving factors for this is that the accuracy of voice technologies has improved dramatically, as shown in this chart:

In fact, computerized voice recognition software has recently begun to beat humans in voice recognition tests. Speech generation is also getting much better. These are key drivers for voice adoption.

Next, Naomi shares five key insights with us:

1. Voice Is About Action: Interactions in voice are 40 times more likely to be about actions than traditional search queries:

list of queries

2. People Expect Conversations: Voice interactions between humans and devices are much more conversational:

As a result, we’re evolving from the keyword-based query to something more dynamic. For example, there are over 5,000 ways to ask to set an alarm.

3. Screens Will Change Everything: Nearly half of all voice sessions use a combination of voice and touch input:

The world of voice will be multi-modal, and will involve a mix of tapping, typing, and talking.

4. Daily Routines Matter: People will use their personal assistants to support their daily routines:

They will naturally take action based on the context they’re in, not the device they’re on.

5. Voice Is Universal: No manual is needed. We all (well, nearly all of us) learn how to speak, and it’s intrinsic to human interaction. We learn a common language of communication, and our personal assistants will be tasked with understanding us as we are.

Naomi’s final words of advice are for us to “show up,” “speed up” and “wise up.” We show up by creating great content, using schema markup for our content, and by posting videos on YouTube. We speed up by creating voice experiences that make working with our voice apps faster than the alternatives available to users.

Finally, we wise up by starting to get our hands dirty now. Start learning to create conversational interfaces now. It will take work and experience to get good (or great) at it. We’re still in the early days, but this is likely going to come upon us quite fast.

Last, but not least, think about what your brand is going to sound like in the world of voice. You need to learn to project your brand persona through this environment, as voice is inherently social. How your brand sounds online will become a core marketing area of focus. Make sure you sound good!

The post The age of voice assistance and a new input paradigm appeared first on Search Engine Land.

Bing’s ‘Quest for Intelligent Search’ at SMX East Fri, 26 Oct 2018 22:29:05 +0000 Search has come a long way and, in a keynote talk Wednesday, Christi Olson and Frederic Dubut painted a picture of what the future might hold.

The post Bing’s ‘Quest for Intelligent Search’ at SMX East appeared first on Search Engine Land.

The first SEO-related keynote at SMX East this week came from the Bing team and was entitled, “The Quest for Intelligent Search.” Featuring Christi Olson and Frederic Dubut, the presentation contained many insights into the core areas that search engines are investing in to make search better.

Christi and Frederic discussed the topics interactively, and the initial discussion focused on the shifts and changes that Bing has made in many areas of search.

When considering search, you might think that the goals for a search engine are:

  1. Discover and index everything
  2. Create (expand) a knowledge graph
  3. Super fast access to information

In fact, the way that Bing thinks about this is that the goal is to:

Provide searchers with timely, relevant results that they can trust to provide insights about their search queries.

To back this concept up, Christi showed old search-result screenshots to show how search has evolved. Here is one from 2006 or earlier, when search results still consisted of just 10 blue links:

Here is another example of that, but this one includes some advertising:

Of course, search has evolved. Here is an example that includes search results other than just those blue links (what we call “blended search results”):

Finally, fast forward to today, and your results might look like this:

In this latest result, you now see two featured snippet-type responses. What’s intriguing about this one is that it provides multiple perspectives. Here is another example of a multi-perspective result offered by Bing:

Finally, here is a third example of a multi-perspective result, notable because it also provides a timeline of events:

This is an area where Bing has pushed hard in their technology. You might even find, for example, that your article that ranks for a query like “is coffee good for me” also ranks for the query “is coffee bad for me.” Often when users search for something, they really do want both perspectives.

Another area of innovation for Bing is in how they think about the nature of the information that they index. The knowledge graph is no longer enough — people want the graph of themselves too. Bing is integrated into Office 365, and it will index all of the information within your Office 365 environment, and make that much easier to find.

To test this, use your search bar to search your OneDrive and your hard drive at the same time. You can think of this as your personalized knowledge graph.

Bing is also pushing to make it easier to search through images. For example, two years ago, Bing released functionality that allows you to shop within image search:

This capability is driven by data feeds, so implementing schema is important in order to appear here. This functionality also includes showing related images in a people-who-like-this-also-like-this-type format.

Another area of innovation is the concept of fact-checking, as shown by this screenshot:

This functionality is dependent on Schema ClaimReview markup to provide context to show if it has been fact-checked. In the example screenshot, the claims have been fact-checked by Snopes (and revealed to be false).

Bing also provides full support for AMP. Many people think of this as a Google-owned project, when it is, in fact, fully open source. Bing has its own viewer and its own AMP cache located at

The only thing you need to do to enable support is to allow Bingbot to fetch AMP content from your site, and to allow cross-origin-resource-sharing (CORS) to permit AMP content hosted on to render resources hosted outside of Note that Bing’s support for AMP is currently limited to the U.S. only.

Bing also recently had to deal with an issue in their Webmaster Tools, because the URL submission form was broken. They removed the anonymous URL submission functionality because spammers loved it. As a first step, the Bing team switched to authenticated accounts, but spammers found ways around that too.

This triggered a throttling of how fast they could process URLs, and resulted in a slow down in the processing of legitimate submissions. That is fully resolved now. Nonetheless, they are continuing to work hard to fight spammers and slow them down going forward. This will likely result in tweaking the number of URLs that can be submitted through the URL submission form sometime soon.

Before SMX Advanced in June, the Bing team had received feedback from some webmasters that their sites had been getting hit too hard by the Bing crawler. This might include spending too much time crawling pages that were largely static or that were changing in minimal ways.

Bing worked on improving processes that can reduce crawling levels by as much as 40%. This is discussed in this post on the Bing blog that cites a Cornell University case study.

Many people ask about Bing’s support for JavaScript. This is something that Bing does support, in a similar way to Google. On a related note, Google recently deprecated support for the escaped fragment protocol, and,while Bing has not done so yet, Frederic tells us, “don’t use it.”

Ultimately, the search experience that Bing wants to provide us is:

  1. Intelligent
  2. Personalized
  3. Pervasive
  4. Predictive
  5. Conversational

The goal is to have the power of the internet everywhere we are. In time, the primary interface will be our voice, and we will be speaking with a digital assistant, though you may also be working with a companion screen, or even other formats.

Christi then shows us a video that captures this concept. In this video, Cortana (their personal assistant software) is engaging with a woman throughout her entire day, taking care of all of her needs, trying to stay one step ahead of her all along the way. Some of the key things it does include:

  1. Nails the music she wants every time.
  2. Identifies that there is a park near her next appointment, and gets her to bring her sneakers so she can run to help her meet her training goal.
  3. In her car, it offers her dynamic re-routing around heavy traffic to help her get to her destination and tells her how long any delays might be.
  4. The car also makes its own repair appointment and arranges to drive itself to that appointment.
  5. When she learns that she has just gotten a new emergency assignment, she lets the car drive itself so she can work on it during the drive home.

The post Bing’s ‘Quest for Intelligent Search’ at SMX East appeared first on Search Engine Land.

How Google’s March, April and August updates might fit together Wed, 26 Sep 2018 19:39:00 +0000 In Google's mind, does brand authority trump breadth and depth of content? Here's a new perspective connecting the dots between three significant algorithm updates.

The post How Google’s March, April and August updates might fit together appeared first on Search Engine Land.


It’s been an interesting year in Google-land, with a number of large updates, including those in March, April and August.

Many have written at length and shared related data about all of these updates. In today’s post, I’ll cover in depth one aspect of the August update that hasn’t received much attention: brand authority.

I believe that one significant component of what Google did with the August update — increasing rankings for sites with general brand authority — was an adjustment to the impact of the March and April updates.

To be clear, the August update clearly had many different components that are also of critical importance. I’m calling out brand authority because it was not explicitly identified by other articles on the update.

I’ll go into this more deeply below, but I’ll begin with a brief recap of what others have said about the March, April and August updates. Feel free to skip ahead.

A brief recap of the March and April updates

I’m not going to repeat what’s already been said in the many excellent articles about these updates (though I link to some of them below). I will summarize some of the basics, then add some additional thoughts about what happened.

Here are some of the key quotes from Google on the update:

Per Google’s John Mueller in the Webmaster Central office-hours hangout on April 23:

The updates that we made are more around relevance, where we’re trying to figure out which sites [are] relevant for certain queries and not so much a matter of quality overall. That might be something where we just don’t think your site is exactly relevant for those specific queries. It doesn’t mean that it’s a bad site, it’s just not relevant for those specific queries … That’s something that I think happens to a lot of sites over time in that they might be really high-quality content, but over time [they’re] just not seen as relevant in the overall picture of the web.

Overall, there is a general consensus that the main components of this update were about how your content’s relevance is measured and Google’s adjustments around its understanding of user intent. This is consistent with statements from Google, as well as the data shared and analyzed by a variety of people in the industry.

Here are some of their analyses:

  1. Recap by Barry Schwartz on March 12, 2018
  2. Recap by Marie Haynes, last updated April 23, 2018
  3. Recap by Glenn Gabe on May 16, 2018 (Part One)
  4. Recap by Glenn Gabe on June 5, 2018 (Part Two)

My supplemental comments on the March and April updates

One aspect of the March and April updates that didn’t get much attention is the idea that the breadth and depth of a site’s content were considered as a ranking signal. Sites with large volumes of content that thoroughly and completely address a topic of interest did extremely well in these updates. Here is how I would characterize breadth and depth of content:

  1. Content created by true subject matter experts.
  2. Content created in high volume (from tens to hundreds of pieces of content per month).
  3. Content that addresses key topic areas both deeply and broadly. In other words, they address many subtopics that are important to users, not just the surface level of the topic area. This depth and breadth may be (and probably should be) accomplished across many articles.
  4. And, of course, content that isn’t sliced thinly for the sake of volume. Each of the different articles has a real reason to exist.

I saw many sites with these four characteristics experience a major uplift with the March and April updates. Here is an example of the Searchmetrics data through April for one of those sites:

As you can see, SEO visibility nearly doubled during the course of these updates. That’s a pretty serious set of gains! This is a phenomenon seen with many sites that follow this pattern of publishing quality content in volume. But, as noted, I do believe that a big key to this is the perceived depth and breadth of coverage of a topic.

To preserve anonymity, let me share what I mean with a fictitious example. Let’s say you want to be seen as an authority on replacing a kitchen sink. You might create a comprehensive article on the topic and include a companion video. That would be a great start. But perhaps some portion of your audience might be interested in one or more of these related topics:

  1. Disposing of the old sink.
  2. Installing a kitchen sink countertop.
  3. How to install kitchen sink plumbing.
  4. What type of caulk to use.
  5. How much it costs to replace a kitchen sink.
  6. What tools are needed for the job?
  7. Installing a garbage disposal with the sink.
  8. What would a plumber charge to install one?
  9. Changing a sink faucet.
  10. Special considerations for brass sinks.
  11. Special considerations for copper sinks.

I could keep going, but you get the idea.

A brief recap of the August update

People have called out many different aspects of this update. Some of the principal ones have been:

  1. Health-related sites being impacted more heavily, hence the “Medic” name Barry Schwartz gave to the update. However, it’s become clear that many different types of sites were impacted, not just health sites.
  2. An increased focus on expertise, authority and trust (E-A-T). In this context, authority tends to mean using subject matter expert writers, citing other authoritative research (including outbound links to same), managing your reputation online and so on.
  3. More speculation on aligning your content with user intent.
  4. Basic SEO factors like crawlability, avoiding thin content, mobile readiness and more.

There is not quite the same level of consensus that there was with the March and April updates, probably partly because Google made fewer statements specifically about it. In addition, I think it’s quite likely that between April and August, Google collected a lot of data on the March and April changes and decided to make a series of adjustments as a result. More on that in a minute.

Here are some of the recaps written about the August update:

  1. Recap by Barry Schwartz on August 8, 2018
  2. Recap by Barry Schwartz on August 9, 2018
  3. Recap by Ignite Visibility on August 14, 2018
  4. Recap by Marie Haynes, last updated on August 8, 2018 (Part Two)

Digging deeper into the August update

I already noted that when Google does any large-scale update, they continue to collect data on how the SERPs are performing, and they’re able to compare that with what they were seeing prior to a given update. Based on this, they can make adjustments to build upon the success of the earlier update and correct its weaknesses. It’s a process of continuous improvement.

For example, here is the Searchmetrics data for one Fortune 100 retailer, showing a large drop to their traffic in April:

This site is for a very well-known brand, but it has fairly thin content on the e-commerce pages. The products are there, but there’s not much description or detail about them. And the site took a hit. However, they appear to have seen some level of recovery in the August update.

Here is a look at another site from a large, well-known brand through the same updates:

This site had the same problems with a lack of content on major e-commerce pages, and it took a substantial hit in the March and April time frame. However, it also recovered in the August update. So, I went looking for more of these. Here is an example from the travel industry:

Yes, another major brand with some content problems that sees a recovery in the August update. Here is yet another example of a prominent e-commerce site taking a hit in March and April but recovering in August:

To try and figure out what was going on, I did an analysis of each of these sites (as well as several others). In each of the above cases, and in several others I looked at, it seemed like the March/April evaluation of the site’s relevance was hurt by a lack of good, in-depth content on their e-commerce pages.

Why did all these sites recover during the August update? Based on the data I’ve seen, my speculation is that the weight of brand authority signals was one of the things that was adjusted in the August update. When I talk about brand authority, I don’t mean authority in the E-A-T sense, but in the pure and simple strength and power of a brand. How does Google measure that? There are probably many components to it, including factors like links, mentions and overall user engagement with a brand.

Why should brand authority matter so much? Think of it from a user perspective for a moment. Users develop a strong affinity for brands. They learn to trust them, and they give them the benefit of the doubt. As related to this series of updates, it means they may prefer sites from prominent brands they trust, even though the content of those sites is materially weaker.

In addition, for curiosity’s sake, I also looked back at my example site that I shared earlier, the one that did really well with the March and April updates. How did it fare?

It kept on soaring upward! For that first example site, the depth and breadth of their content has kept them going strong.


There were undoubtedly many components to each of this year’s major updates. I’m by no means saying that brand authority was the single focus, or even necessarily the primary focus of the August update. But it does appear to me that it was one of those factors.

What does that mean for us as publishers of websites? What you see in the wild about E-A-T does matter a great deal. You need to keep investing in content and user experience. That makes you a better site for Google to rank higher in their results.

But brand authority still matters, too. That means all the things that cause people to have a great experience on the site, links to your site and mentions of your site or brand remain a key part of the overall mix.

The post How Google’s March, April and August updates might fit together appeared first on Search Engine Land.

A closer look at Chrome’s User Experience Report Fri, 07 Sep 2018 18:00:00 +0000 Google's Ilya Grigorik offered this breakdown at the SMX Advanced event in June. Here's a refresher.

The post A closer look at Chrome’s User Experience Report appeared first on Search Engine Land.

The second session of SMX Advanced 2018 was the SEO Keynote: A Conversation with Google’s Ilya Grigorik.

Ilya is a Web Performance Engineer and also the co-chair of W3C Web Performance Working Group and author of the book “High Performance Browser Networking” (O’Reilly). He jokingly refers to himself as an internet plumber.

Here is the session overview:

Join Google Web Performance engineer Ilya Grigorik for this fireside chat where he’ll discuss web performance, site speed, user experience, security and mobile-friendliness, top-of-mind issues for advanced SEOs confronting next month’s Speed Update, and already dealing with the mobile-first index. We’ll also take a look at how the evolution of the web and Google search have impacted both user experience and user expectations, and discuss where developments for each are headed. Bring your questions and curiosity to this interactive chat!

The following are are my notes and insights from his interesting talk.

The keynote

The discussion centered on the Chrome User Experience report (CrUX report), and how we can use it to better understand how users experience our sites and our competitors’ sites.

The CrUX report is a good source of real-world data on user experience on a given page. The data is assembled from actual user sessions across the web, based on:

  • Users who have opted in to sync their browsing history.
  • Users who have not set up a Sync passphrase.
  • Users who have usage statistic reporting enabled.

The data can then be accessed via Page Speed Insights. Here’s an example of what the report looks like:

Ilya explained that the FCP metric stands for “First Contentful Paint.” This denotes when a user first sees a visual response. This metric is important because the first visual response provides users with an indication of action, and it helps keep them engaged.

DCL stands for DOMContent Loaded. It measures how long it takes for the document to be fully loaded and parsed. Stylesheets, images and subframes are the exceptions here. They may not show as complete in terms of loading.

Looking at our sample CrUX report above, notice how users are bucketed into three categories: fast (green), average (yellow) and slow (red). What Grigorik said next was an important insight: not all users get the same experience with your site; the percentages vary by category.

In the above diagram, 57 percent had a fast FCP, 29 percent average and 13 percent slow. For the DCL, we see 36 percent had fast, 41 percent average and 23 percent slow results. You also see that the median FCP was 1.4 seconds, and the median DCL was 2.6 seconds; this places it in the middle third of all pages.

Just to give you an idea of how this works for users, consider the following chart from this User-Centric Performance Metrics post by Google:

The sequence of site loading stages helps us understand what it is you need to work on optimizing. Note the additional metric of Time to Interactive (TTI). This is the point when users can begin to actively interact with the content. This is not something the CrUX report currently gives you, but it’s also something that you need to be concerned with.

Since the CrUX data is accessed via Page Speed Insights, you can pull this data for your competitors as well. Note: If you are pulling data on an individual page in Page Speed Insights, the CrUX report may not be available due to a lack of sufficient data. The message you’ll get back looks like this:

Make sure to enter in the full URL for best results. For example: “”, and not “”, as the latter form will assume the HTTP protocol.

If your page traffic is too low to get the CrUX report, you can also enter a “site:” command into Page Speed Insights in this format: “site:” to get sitewide data.

If you enter “site:” only without the URL, you will only get the CrUX data from Page Speed Insights.

Grigorik emphasized the importance of optimizing for real-world experience and getting direct measurements of your own site. You can get sites where the real-world experience is good and the scores are low, and vice versa. Finding other tools that help you with that is a good idea.

Google Analytics

Google Analytics (GA) has some limited capability in this area. The Site Speed report in GA looks like this:

The metrics tracked by Google Analytics include:

  • Average page load time.
  • Average redirection time.
  • Average domain lookup time.
  • Average server connection time.
  • Average server response time.
  • Average page download time.

It’s interesting to drill down on some of the additional dimensions as well. For example, for this site, if we look at the browser breakout, here is what we see:

Did you notice the long load time for the Edge Browser and Samsung Internet? There may be some browser-specific problems with these two sites. Fixing them could help a lot of users (and increase overall scores, of course). These types of browser-specific problems are not something that the CrUX report would reveal.

What I’ve seen over the past several years is a ton of data showing how improving user experiences and page performance can lead to large increases in conversion. For example, looking at the above browser data, it’s highly unlikely anyone using an Edge or Samsung Internet browser is going to convert on that site.

Overall page and site performance are frontiers that have become increasingly important to address. Configuring web servers, content management systems and e-commerce platforms to make this happen is hard work, but the payoff is significant!

Are you planning to attend SMX East in October? Google’s Naomi Makofsky will be our SEO Keynote on day two. Naomi, who works on Global Product Partnerships for the Google Assistant initiative, will take you on a roller coaster ride from where we are today to what the future holds, including how some future tech is already here and having a significant impact on marketing campaigns. Hope to see you there!

The post A closer look at Chrome’s User Experience Report appeared first on Search Engine Land.

SMX Advanced Recap: Bing’s Fabrice Canel keynote Wed, 27 Jun 2018 14:14:00 +0000 Contributor and SMXpert Eric Enge recaps the opening keynote and big announcement by Bing's Fabrice Canel at SMX Advanced 2018.

The post SMX Advanced Recap: Bing’s Fabrice Canel keynote appeared first on Search Engine Land.

For the opening session of SMX Advanced, we were treated to a keynote by Bing’s Fabrice Canel. Here are my notes from that session.

Fabrice started by pointing out that search is becoming more intelligent. For example, search engines are getting better at sentiment analysis. If you search on whether video games are good for you, you’ll get one set of results, and if you search on whether they are bad for you, you’ll get a different set of results.

You can also get multiperspective answers offering different opinions. Bing does this in featured snippets, as shown here by this sample result:

As Bing prepares for continued growth in voice interactions, it is becoming more conversational in its approach to search.

For example, if a user asks a very broad query which could have several intents, Bing will ask a clarifying question, as you’ll see here:

And where it’s appropriate, search is becoming more visual.

More and more technology is being applied to help search become more human as well, allowing interactions with people to become more natural:

With all of this, search engine optimization (SEO) is evolving, too. The basics still apply. Be sure to optimize your web pages, and pay attention to:

  • Using title tags.
  • Meta descriptions.
  • Heading tags.
  • Writing great, descriptive content.
  • Ensuring your site is crawlable.

Basically, don’t forget your SEO 101!

So, what is new?  There are many new destinations for content (desktop, mobile, voice) and many new ways to distribute and duplicate that content. There are new standards such as accelerated mobile pages (AMP) and much more.

Search engine optimization also has to deal with content being cached on content delivery networks (CDNs) and content being hosted in the cloud. This leads to new challenges with setting up integrated analytics and managing content across multiple platforms to help keep it all in sync.

Artificial intelligence

Next, Fabrice told us that “artificial intelligence (AI) is tech that can perceive, learn, reason and assist in decision-making and act to help solve problems.” He said, “One area that Bing is using AI heavily is to invest in making a smarter crawler.”

This is important because the web presents many complex problems, such as:

  • Determining what content has changed and not changed.
  • What content has been removed or disappeared.
  • What content is new.
  • Detecting duplicate content.
  • Handling mobile and desktop.
  • Handling JavaScript.
  • Handling cascading style sheets (CSS).

Dealing with these items on even a single site is complex, but when you bring them to web scale, they become incredibly complex. AI techniques can make these easier to deal with.

In March of 2018, Bing announced support for schema implemented in JSON-LD. At this year’s SMX Advanced, they announced  Bing has extended that support to allow for debugging of that JSON-LD in Bing Webmaster Tools.

Fabrice did clarify that Bing will not prefer JSON-LD over other forms of markup, as the web is an open environment.

Bing also announced extended support for AMP.

This specifically includes support for viewing AMP pages from the Bing CDN when users click on an AMP page in the Bing results. This support is similar to that offered by Google. Below is a screen shot Fabrice showed:

One very popular SEO question came up, which is whether or not Bing intends to show voice queries separately in their Webmaster Tools. Fabrice indicated that they don’t do that, and it’s not on their roadmap to implement.

Overall, this was a great session and an excellent opportunity to get more insight into what’s going on at Bing. The industry tends to focus a lot of attention on Google, but Bing offers a very robust search experience, and understanding what they’re thinking and doing is invaluable.

The post SMX Advanced Recap: Bing’s Fabrice Canel keynote appeared first on Search Engine Land.

Voice search & virtual assistants — SMX Advanced Session Recap Fri, 22 Jun 2018 15:00:00 +0000 With over 420 million voice assistants sold, it's clear that SEOs need to optimize their content for voice search. Contributor Eric Enge recaps the session and covers the opportunities and challenges voice search presents.

The post Voice search & virtual assistants — SMX Advanced Session Recap appeared first on Search Engine Land.

Voice search changes everything

Voice search is all the buzz out there, with some saying that by 2020, 50 percent of all queries could be by either image or voice. Personally, I think that estimate is aggressive, but the fact is that it’s a rapidly growing segment.

It’s also important to be aware that “voice search” may not be the correct label. I say this because many of the “queries” are really commands, like “call mom.” That is not something you would ever have entered into a Google search box.

Nonetheless, there are big opportunities for those who engage in voice early. Learning about conversational interfaces is hard work, and it will take practice and experience.

For that reason, I looked forward to seeing the Optimizing For Voice Search & Virtual Assistants panel at SMX Advanced, and today I’ll provide a recap of what the three speakers shared.

Upasana Gautam, Ziff Davis

First up was Upasana Gautam (aka Pas). Her focus was on how Google measures the quality of speech recognition as documented in this Google white paper.

Pas went over the five major metrics of quality discussed in the paper:

  1. Word error rate (WER).
  2. Semantic Quality (Webscore).
  3. Perplexity (PPL).
  4. Out-of-Vocabulary Rate (OOV).
  5. Latency.

Next, she went into the quality metrics in detail.

Word Error Rate (WER)

This metric, which measures misrecognition at the word level, is calculated as follows:

WER is not necessarily a great measure of what will impact the final search results, so while this is measured, it’s secondary to some of the other metrics below.


Webscore tracks the semantic quality of the recognizer. Higher levels of recognition result in higher Webscores. Temporal relationships and semantic relationships are what WebScore is all about, and Google focuses significant energy on optimizing this metric. It is calculated as follows:


This is a measure of the size of the set of words which can be recognized given the previously recognized words in the query. It serves as a rough measure of the quality of a language model. The lower the Perplexity score, the better. It is calculated as follows:

Out of Vocabulary Rate

This metric tracks the number of words not in the language model, and it’s important to keep this number as low as possible, as it will ultimately result in recognition errors. Errors of this type may also cause errors in surrounding words due to subsequent bad predictions of the language model and acoustic misalignments.


This is the total time it takes to complete a search by voice. The contributing factors are:

  1. The time it takes to detect the end of a speech.
  2. Time to recognize the spoken query.
  3. Time to perform the web query.
  4. Time to return back to the client.
  5. Time to render the result in the browser.

If you have an interest in developing voice assistant solutions, understanding this model is useful because it allows us to better tune our own language model in our conversational interface. One of the things that I’ve learned in the voice assistants we’ve developed is that picking simpler activation phrases can improve overall results for our actions or skills.

Pas SMXInsights

Katie Pennell, Nina Hale

Katie shared some data from a Backlinko study of 10,000 Google Home search results and a BrightLocal study of voice search for local results:

  1. 70 percent of Google Home results cited a website source (Backlinko).
  2. 41 percent of voice search results came from featured snippets (Backlinko).
  3. 76 percent of smart speaker users perform local searches at least weekly (BrightLocal).

The theme of her presentation, reinforced throughout her talk, was that not all topics work great for voice search.

For example, with entity searches the data will get pulled from the knowledge graph, and your brand won’t get visibility for knowledge graph sourced results.

Web pages hosting your-money-or-your-life (YMYL) type content stand less a chance of being offered up as a voice result. As you work on deciding what content to target for voice search, you should carefully consider the overall customer journey:

You can get useful ideas and trends from many different sources:

  1. People Also Ask result shown by Google.
  2. Keywords Everywhere Chrome plug-in.
  3. Answer the Public.

From there you can figure out which topics will be best for voice optimization. You can also consider developing your own Actions on Google app or Alexa Skill. Be careful, though, as there are lots of people already doing these things.

You can see from Katie’s chart that the market is crowded. Make sure you develop something that is useful enough for people to want to engage with it.

In addition, make sure the Google Actions and Alexa Skills fits in with your business value. To do that you can leverage:

  1. Keyword research.
  2. Social listening.
  3. Internal site search.
  4. User research.
  5. Customer service team.

Katie’s SMXInsights:

Benu Aggarwal, Milestone

Benu started her presentation by discussing a seven-step process for voice search:

She also shared interesting data on how people use voice search:

One of the big factors in voice search is that when you get a spoken reply from a device, it typically comes as a solitary answer. If you’re not the source of that answer, then you’ve been left out of the opportunity for that exposure.

Benu also discussed the importance of offering conversational content across each stage of the customer journey and shared an example in the travel industry.

She and her team looked at the rankings for a client on queries like “hotel near me” and saw the client ranking at position 1.5 and at position 6.7 on desktop. This serves as a reminder that you need to check your rankings on a mobile device when selecting candidates for which you might be able to get a featured snippet. You stand a better chance to be the response to spoken voice queries if you do.

Featured snippets are generally seen for informational queries and are more often shown for long-tail queries. The questions they answer frequently start with why, what, how, when or who.

You should also seek to create content for every stage of the journey, but 80 percent of that content should be informational in nature, as these are what feed most featured snippets.

Here are some thoughts on different content types to consider:

  1. Informational intent (guides, how-tos, etc.).
  2. Navigational intent (store locations, services, press releases, customer service info).
  3. Transactional intent (videos, product information, comparisons, product stories).

Additionally, here are some content tips:

  1. Satisfaction. Does the content meet user needs?
  2. Length. Relevant fragment from a long answer.
  3. Formulation. Grammatical correctness.
  4. Elocution. Proper pronunciation.

Benu also proposed a site architecture chart for a voice-optimized site as follows:

Benu’s concept is to integrate FAQs across every major page of your site. In addition, you need to condition your organization to buy into this approach. That includes developers, designers, content creators and more.

She also spoke about the process for creating Actions and Skills and included a flow chart for the process:

Benu is also an advocate for using multiple ways to connect and chat with customers, as captured in this chart:


Lots of great perspectives and info were shared. As you begin your own journey into understanding voice, be prepared to experiment, and be prepared to make some mistakes.

It’s all OK. The opportunity to get a leg up in this emerging new space is real, and if you approach it with an open mind and a willingness to experiment, you can learn a lot and begin building mindshare with your target audience before your competition does.

The post Voice search & virtual assistants — SMX Advanced Session Recap appeared first on Search Engine Land.