How Artificial Intelligence Changes SEO: Asking Experts

I think Artificial Intelligence is the core of both Text and Voice Search. Search Engines use sophisticated AI to predict which results will satisfy any given search.

There's a growing amount of SEO tools on the market that use AI to conduct content and keyword research and track social media mentions, among other things. These can help you identify opportunities for content and keep track of how they're ranking, two key components of SEO.

When AI gets significantly smarter, we can use it to automate parts of marketing that are either tedious, or take a long time. For example, Grammarly is an AI-powered tool that can spot grammatical errors in a minute. Something that'd take you an hour of focused text analysis to do.
Another example is content writing. Writing first drafts is often the hardest part of article creation.
So if we could input data and links to a program, and then it spits the draft we can then make shine, that'd be a huge time saver.

Google is continuously working to improve the user experience and to provide the user with the best possible answer in the shortest possible time. The rich snippets have been with us for a longer time now. If you optimize your content for rich snippets, you may increase the visibility of your search results, which in turn can significantly increase the CTR.
Another interesting product is Google's "Discover". The product is developing into a good traffic supplier. AI optimizes the users news feeds on their smartphones by showing relevant news.

Let's face it: most websites fail as they lack, e.g., content strategies. So we shouldn't focus too much on trending topics such as AI or machine learning. There are definitely use cases for AI, but first things first: make up your mind on what you want to rank for. Check if your website is good enough and identify areas for improvement.
When you got the basics right, then you can have a look how and if topics like AI can add the remaining percentages to your website.

Understanding how Google might potentially use AI as part of it's algorithms, and then trying to work with the theory rather than game is, for me, a crucial part to modern day SEO.
The underlying result of trying to optimizise for AI is that, as webmasters, we're trying to increase the number of topics and phrases covered within our content, and create more logical site structures that flow between said topics. Whether or not this is done under the guise of AI optimization (which John Mueller has been on record saying it can't directly be optimized for) or through a wild spate of "LSI keyword optimization" (again, not a thing), the end result will only benefit users.
The difference between those rewarded in organic search vs. those who don't, comes down to those working with the theory vs. those trying to game it.

Ask 5 people to define Artificial Intelligence and you'll get 6 answers. The challenge, when talking about AI, is coming up with a definition that everyone can get behind. This makes it incredibly difficult to start making predictions and nearly impossible to start talking about business critical actions based on AI. So, I'll do some disambiguation first. When it comes to search technology and SEO, I see Machine Learning (ML) as a learning system where human curation, review and refinement as critical to its learning abilities. Artificial Intelligence (AI) as a learning system that has the ability to remove most or all human review from the learning process and still get the results right most of the time. So, how can AI be used in SEO? I think on the search engine side, Google is using a highly advanced version of Machine Learning that continually removes parts of the human curation aspect. Therefore, right now, AI by search engines are primarily used for intent detection, disambiguation and the collection and use of non-textual signals for personalization and prediction. I have yet to see any SEO professional or software use AI for our primary jobs of improving tech, architecture, links, content or rankings. However, I am seeing some really interesting uses of ML to automate tasks like identifying content gaps to make content strategy more accurate. I've also seen some ML being used to assist content creators curate information across topics to make the content they product to be more authoritative.

Artificial Intelligence (though I prefer the term "Machine Learning" when referring to practices in the SEO field, Al being so vast in scope!) can help the SEO practitioner in many ways - such as:

• Cut down and cluster an avalanche of operational signals to look for anomalies;
  
• Forecast time series more accurately;
  
• Detect user intent and find subsequent opportunities via clustering;
  
• Auto-generate content at scale;
  
• Auto-label your keywords, URLs etc.

I could go on, yet the thing to bear in mind is that ML can automate and speed-up many of your daily processes. Tasks that would usually take hours on spreadsheets may only take minutes with a well thought-out script!

Yes, I use a tool that analyzes the content of the top 10 ranking websites for any given query and based on their content, I know which words or sentences are being used that may help to improve my content's semantics.

We certainly try to take advantage of AI advancements in Google around natural language and voice searches. We've recently begun a process of optimizing our content for voice search given the recent development in Google algorithms and the growing share of such searches. We also employ several tools that utilize AI to help us with SEO research, monitoring and evaluation.

We work on some tag and NLP stuff and I use Google itself a lot, which should use AI too :)

Use cases vary from client to client, but in general we use AI, e.g., to cluster search queries into groups or to find patterns in web analytics data. So nothing super spectacular here, as we need to get the basics right in the first step and then constantly improve the website.

We use a low-form of AI to perform mass categorization across some publishing websites; by using training sets to educate it that certain phrases such as "Manchester United" or "Troy Deeney" are suitable for a sports classification, so this can be run en masse.
We also use some grammar and content tools within the wider content optimization process, based on the theory that if the tool is working with data sets to make suggestions to improve content quality, it's good to have as a second phase to our own internal processes.
This still requires a form of human checking however, as AI isn't yet a substitute for logic.

No. At Red Hat, for our site-search, we have several ML projects running to look at intent detection and personalization for better search results.

While dabbling with Javascript and R. I code mainly in Python. It is intuitive, versatile, and is a great pick for Machine Learning, as ubiquitous ML frameworks such as Sci-kit Learn, Tensorflow or PyTorch are all written in Python!
Currently, I use ML for regression problems, forecasting and text classification. As a server log aficionado, I'm also experimenting with ML for anomaly detection.
It's only recently that I've integrated Machine learning in my daily workflow. Needless to say I'm only scratching the surface of it - there is so much to learn and for me that is super exciting! :)

Yext is a very great tool that uses AI to monitor reviews, optimise for voice search, and many more useful solutions.

Kafkai, Ai Writer, Articoolo (AI content creation tools); MarketMuse (AI powered content optimization); CanIRank (tool that uses AI to determine whether you can rank for any given query).

Wordlift, Market Muse, HubSpot.

There are a number of tools out there that work on combining performance data, competitor data, and then making optimization suggestions based on the small dataset within.

For SEO's who want to learn how search engines rank information, using the LTR (learn to rank) ML plugin for a search platform like SOLR is a fantastic way to start learning how to incorporate ML into your search, which will give actionable information to point your SEO efforts into good directions.

There are a plethora of SEO tools leveraging ML out there, from Ryte's great Anomaly Detector and GetRedirects, to Quill (which cleverly use NLP models to deliver personalised stories). Also check out WordLift from my friend Andrea for some crazy Semantic wizardry!

And, while they're not SEO tools per-se, I'm fascinated by "Automated" ML solutions (Google AutoML, H20.ai, Uber's Ludwig, Spacy.IO etc.). It's great to see how these tools are democratising ML in the organisation, allowing anyone to build quality custom models with limited coding expertise. Sometimes, removing the hassle of tasks you'd otherwise need to perform manually (hear hear hyperparameter tuning!) can be a godsend!

Last but not least, I've recently fallen in love with Streamlit. I think it is a game changer for web applications, even outside of ML!

BERT helped search engines to better understand search queries either with a text or voice search. This makes search results to be more accurate, especially in long-tail searches.

According to Google, this algorithm will affect up to 10% of results by using AI to better understand language in context. Common examples are how prepositions are used in a sentence, and how their meaning changes in different situations. This is something that speakers of a language easily understand, but has been difficult for previous algorithms to grasp. All of this means that searches using more natural, full-sentence type language will lead to more accurate results. This is particularly important for voice searches since they often use natural language.

It boosted Google's understanding of relevancy. Google wants to rank the most relevant result, and not the one that has the most links. Google uses links as the necessary evil to gauge the "popularity" of any content asset, but BERT enables them to understand content at a deeper level so they don't need link data to make the call on who should rank and where.
Bottom line, BERT made it easier for weaker and smaller sites to beat the giants in their verticals. If their content is better and more relevant they have a good chance of ranking highly.

According to Google estimate, BERT algorithm affected roughly 10% of all U.S search queries. Informational and long-tail queries BERT algorithm impacted organic search results in rewarding pages with content that better match user intent as Google is getting better to understand natural language with NLP.

BERT Algorithm update has been one of the biggest improvements in how search engines can recognize entities and answer questions in the recent history. With BERT came also a new shift to open-sourced technology allowing everyone to use this algorithm.
BERT has helped Google better understand the nuances and the context of conversational queries. Allowing the search engine to serve more relevant content, recognize entities, and make SERPs more interactive.

Based on our analysis, the BERT algorithm update gave the Google algorithm more of a focus on content quality with a change of user intent, and subsequently a move away from the importance of links (although we have seen a move back towards this in the core algorithm updates that followed the BERT update).

BERT had, and still has, a positive effect on the SERPS as it tries to rank content based on a searcher's intent to provide the most relevant content to them first. Although Google mentioned that the update would only affect 10% of results, BERT is a bi-directional, contextual model of NLP meaning that results should reflect better the complete answer that a searcher needs from their search query, compared to the search results in the past.

I don't think there is a formula or a checklist to optimise a site considering AI. I think the best advice would be to write for users, search engines will understand content and queries like a normal person would, so a well-written text that focuses on addressing the user's needs is what we all should do and should have been doing for a long time.

The best advice continues to be to focus on providing quality content that's aligned to your audience's needs. To understand those needs, I would advise using an AI tool to conduct research. In that way you can learn about what people are interested in, what questions they have and what type of content they're looking for. Once you have this it's a matter of producing high quality content that is focused first and foremost on helping people. Google is getting better every day at rewarding just such content.

Since we don't know the exact use of A.I in the algorithm (and even if we did know it right now, it wouldn't really matter because the program's always learning and evolving); I think our best bet is to use the AI software we have available to streamline our processes and make life easier for us.
So use AI for what it's worth but never forget that it'll probably never be able to beat human login, intuition, and common sense. So going forward, it's content; and it's links. And it's also the smart use of AI where possible, to get an early advantage over those that don't.

Google is constantly evolving, but the goal for the search engine is and remains the same: it wants to provide the best possible user experience in the shortest possible time.
We always write our content with this goal in mind. We optimize our texts largely to answer questions such as "When will shoe XY be released" or "Where can I buy shoe XY". We always try to answer these questions in our content.
Trying to answer users' questions in our content has always been our goal in SEO and that seems what Google rewards in the long run.

Focus on users and their needs! What do they expect to find? What's the best approach to present it? We aren't optimizing for search engines and their algorithm – they aren't using your website and they won't convert. So simply focus on your target group, their needs and constantly improve your website.

Work with the theory, and don't try to game it. AI isn't TF*IDF, LSI, Keyword co-currence... AI is aimed at better understanding and matching content documents to user searches, so if you want to really optimize for AI, you need to be writing highly expert, authoritative, and trust-worthy (EAT) content that covers a users initial base query, subsequent queries, and potential stacked queries.

Take your time to understand ML & AI, but remember it's Google's job to understand you, your job is to build great websites with authoritative content that seres your customers. If and when you do use ML, use it to understand and automate tasks to shorten the time between discovery and execution.

While over the past half-decade, improvements in computing performance hugely benefited the AI field (Google achieving quantum supremacy is now real!), the basic rules to perform well in organic Search haven't changed:

• Build strong technical foundations. Design lighting-fast websites. Stick to the crawlability rules to ensure content is discovered often and well.

• Oh, and did I mention to write for users? ;)