How To Create Content That Is AI-Friendly
How do I create AI friendly content?
1. Use Driver analysis
2. Apply Text mining
3. Integrate Time Series Prediction
Artificial intelligence (AI) is widely applied to create personalized, unique experiences for consumers in a wide variety of industries. A recent study by Business Insider predicts up to 85% of consumer interactions will be managed without human intervention by 2020.
What are the implications of this on content? To answer this question, we need to answer what AI is first because there is a lot of confusion about the term. It’s predominantly associated with robots and movies. To help clarify: our world currently runs on Artificial Narrow Intelligence or ANI, machine intelligence that at least equals, but usually surpasses that of human beings. Examples include Google’s self-driving car, our phones, and even email spam filters.
When we talk to Siri or navigate using our map app, we are using ANI. Our spam filters start with lots of data on how to determine what spam is. They learn and customize their intelligence to users as they become familiar with our preferences.
Companies like Google use artificial intelligence to deliver better search results to their users. In 2000, Google cofounder Sergey Brin told Online Magazine that artificial intelligence would transform the company. He said the ideal search engine was smart and would have to understand a query, and that was clearly AI. However, concrete steps were not taken immediately. The development is relatively recent, having started in 2011.
2011: AI “Rises” out of Google’s Labs
Google CEO Sundar Pichai fondly recalls the day this happened. An engineer named Jeff Dean, who helped build Google’s search engine, wanted Pichai to look at a project he’d started working on in 2011. Dean and colleagues had built a huge network of interconnected computers, connected in ways modelled on the human brain. They had engineered 1,000 computers with 16,000 processors, which were capable of making 1 billion connections combined. Although still was far from the human brain’s capacity, which exceeds 100 trillion, it was an unprecedented achievement for a computer system.
2014: Google Acquires DeepMind and Launches DistBelief
In 2014, Google bought AI startup DeepMind for £400 million. This company specializes in machine learning and algorithms for positive impact. That same year, Dean harnessed the power of his massive computer network into a new framework to train machines in mass-scale thinking. This project was called DistBelief. Many teams, including YouTube, Android, and Maps, started using its findings to make their products smarter.
Unfortunately, DistBelief was not designed to adapt to technological changes such as the emergence of speech as a complex data set and the rise of GPUs.
2015: Google Introduces TensorFlow
The successor of DistBelief, TensorFlow, did away with these limitations. It was opened to developers outside of Google in 2015. Twitter uses it today to rank tweets, build bots to monitor conversations, and encourage people to spend more time in the medium.
2019: Google Holds I/O Developer Conference to Introduce AI
On May 7, 2019 in Mountain View, Google presented new services and features introducing AI into people’s day to day lives. One example is “knowledge” panels, which news searches will soon generate. For example, the search engine will index podcasts by crawling content and making it available in search results. The user can save the podcast that comes up on the results page for later or play it. They will also be able to click on 3D icons to access 3D models appearing in search results.
Artificial intelligence has progressively learned the characteristics of what makes published articles valuable or not. It automatically classifies these web pages and determines their rankings with high precision.
Creating AI Friendly Content: Profitable Action
Driver analysis is a very effective tool when you’re overwhelmed with data, but you’re not sure which is important for your desired outcome. In this case, machine learning software is immensely useful. You enter the data and it highlights the significant elements. An analysis will typically end with information about the most crucial mathematical relationship between a set of variables and a desired objective.
Driver analysis in content marketing shows which factors generate the highest number of leads. Implementation of driver analysis on Spin Sucks, a marketing blog, showed email was the strongest driver and organic search was the third-strongest. This led the blog’s team to turn more attention to email marketing as the most powerful source of traffic.
A ranked list of drivers can help you set priorities and plan resources and budgets more effectively regardless of your objective, be it leads, page views, revenue, or social shares.
Driver analysis uses natural language processing to narrow down, provide context, and ultimately improve online shoppers’ search results. Another example of a company applying it is the tech startup Clarifai, who initially focused on the visual elements of search. The company is allowing developers to create more intelligent apps that perceive the world like human beings do. Video recognition and advanced image empower businesses to create customer-centric content.
Automatic Content Generation
AI has made it possible to generate content automatically for simple stories such as sports news and stock updates. There is a high probability you’ve read content written by an algorithm without even realizing it. Companies like Fox, Yahoo, and AP have been using AI to generate content automatically for quite some time.
Text mining is an approach used to reveal keywords, topics, and hidden issues. This AI application ingests content to categorize, rank, and make sense of it. It utilizes what is known as vectorization, which converts words into numbers. This form of learning looks at the mathematical relationships between numbers and establishes if words are similar or not.
Developers have been able to “reverse engineer” Google to illustrate key terms and topics. The Google algorithm is itself a deep-learning system utilizing a great deal of AI. Its search algorithm is extremely complex and chances are no one really knows how it works. Google itself has a very small degree of interpretability of this algorithm.
An example of applying text mining to reverse engineer the algorithm is a top-down approach to analyze search terms. Let’s say you want to analyze the term “artificial intelligence.” Text mining will show the words, which the top 15 or so result pages have in common. The list of words will indicate what categories or concepts to cover when creating content around the reverse-engineered keyword. You’ll get better organic search results when you have this list of common words than you would if you simply tried to write a good article about artificial intelligence.
AI-Friendly Content to Generate Leads
AI has helped some e-commerce businesses solve lead generation and other challenges. An example is software company Mintigo, which provides AI solutions for sales, marketing, and CRM systems. This company’s products have helped Getty Images capture data showing which websites display images from their competitors. Getty generated significant new leads as a result.
Gain Hidden Insight
Darden Restaurants carried out a full management reshuffle in 2014, in the wake of which certain changes were implemented across the board. One of these involved enforcing the company’s breadstick policy, which was to serve one breadstick per guest. The staff now had to count the number of breadsticks served based on the number of guests.
Text mining was used on more than 2,000 public reviews written by Darden staff. Results indicated breadsticks were an issue for employees. A manual analysis of the reviews would have been misleading, as most of the poor reviews involved typical employee complaints like long hours and low pay. These were actually a consequence of the Sisyphean efforts to enforce the restaurants’ breadstick policy.
Text mining’s applications go above and beyond commerce and retail. It can be used to gain hidden insight into well or poorly performing blog posts, customer success phone call transcripts, and more.
Analyzing Competitor Brand Search with Time Series Prediction
This is an approach combining AI, math, and statistics that enables content marketers to predict what their competitors will search for. It then becomes easier to staff, plan, set budgets, and set up an editorial calendar. An exercise of time series predicting was done on search data on hotels in Cleveland. Branded searches over the period of one year were analyzed, where searchers named specific hotels such as the Cleveland Marriott, Hilton Cleveland, Hyatt, and Holiday Inn Cleveland. The results accurately forecasted when search volumes would increase and decrease for each hotel.
One specific finding was that the Cleveland Marriott would have more search interests at the end of September than its competitors. Management could then launch campaigns to increase their market share at that exact time. The advantages of predictive forecasting aren’t only for hotel brands. Any brand can conduct a search when its performance is less than stellar and use the data to bid on competing brand names with a promotional offer or relevant content asset.
What is more, you can predict more than search volume. It is possible to predict lead generation based on software for marketing automation and revenue from ERP or CRM. You can predict any regular data over time.
Putting Theory into Practice
Having a hard time wrapping your head around this stuff? It’s understandable. However, you do have options, and you don’t need to be a mathematician or programmer. Below are content creators, curators, and marketers’ options when it comes to approaching AI.
If your company doesn’t have data science and AI experts in-house, you’ll find this option to work quite well for you. Simply outsource to agencies or freelancers with the necessary AI experience and know-how. Consultants and agencies can help you understand and apply the methodologies. If you have a frequent or ongoing need for this service, these professionals can help you develop software to use on a case by case basis.
Use your Data Scientists
Do you have a staff data scientist in your organization? You might if it’s medium-size or large. Staff with data science knowledge and skills can be of great assistance because they know how to use the relevant technology correctly. In terms of time-series predicting or text mining, an in-house data scientist will have a good grasp of your goals and objectives. They will be able to develop the right models and implement the codes as needed.
Finally, a small number of content curators and marketers will find DIY a fitting approach. This is especially true for those with a genuine interest in machine learning and data science. These people are interested in going deep with statistics, math, and probability. They also know or at least are ready and willing to learn how to write code.
If you choose this option, you might benefit from Google’s free online Machine Learning Course. It features 25 lessons, 40+ exercises, and lectures from Google researchers, all against the backdrop of real-world case studies.
The Python and R programming languages form the basis of many AI libraries and tools, so if you’re interested in coding, those are worth taking a shot. Be warned you need quite some time on your hands. More specifically, it takes 9 months to learn the languages on average and another 9 to understand the data science.
Content marketers who do not wish to write code may avail of the drag and drop, intuitive user interface made available by IBM Watson Studio. It is possible, but not mandatory to write code on this platform.
Experts draw an analogy between AI learning and buying a computer. To buy a computer, you don’t have to be a developer, but you do need to be able to compare their components (ex. processors) and determine which one is better. Likewise, if a vendor tells you that they integrate the latest AI techniques in their predictive analytics product, you have to know what questions to ask in order to see if that’s really the case.
AI Friendly Content in Retail
It will always pay off to learn about the role AI plays in content marketing if you want to create AI-friendly content. It will become more and more important in the future. In retail, AI is already playing a crucial role.
An example of a company applying AI to improve their understanding of their customers’ needs is The North Face. They use IBM Watson Studio’s AI solution to help online buyers find the ideal jacket. They ask customers questions via AI voice input, such as “when will you be wearing the jacket” and “how often will you be using it.” Then, the software scans hundreds of jackets to find the perfect match according to its own research and customer input. Pertinent factors like local weather conditions are taken into account.
Amazon has begun using AI to improve operations involving speech recognition, answering questions, natural language understanding, and dialogue systems in general. Amazon converts customer speech to text accurately and effectively using deep learning, a category of machine learning algorithms. In their efforts to begin answering questions automatically, the retail giant is applying AI by leveraging customer reviews, product descriptions, and other content within website pages.
AI Friendly Content – Writing Tips
Our final section deals with writing tips to ensure maximum organic reach. These include aligning your headline with the text of your post or article, using floating share buttons, making your content easy to scan, and using high-quality visuals. Read on for the details.
Keep your Promises
If your headline is about the worst city in the US to live in, your article should actually be about that and explain why it is the worst city. According to a LinkedIn study, 23% of readers are likely to stop engaging with a brand if a headline misleads them.
We Don’t Read Online, We Scan
A study by NNGroup found that when a reader looks at a web page, they follow an ‘F-shaped’ pattern. This means they will only read about a fifth of your content, no matter how good it is. Your content should be easy to scan and concise to make sure people get the important facts. This means short sentences and paragraphs and bullet point lists.
Want more shares on social media? Enter “floating share buttons.” Fortune Lords report they increase social traffic by 27%. Effective tactics to this end include short sentences (up to 12 words long) and rhetorical questions. These questions make it more likely for the reader to perceive the content as a conversation. You could also break an argument up using a colon into smaller points to sustain readers’ engagement.
On a final note, you need to not only tell, but also show readers what they need to know. Your content needs charts, images, graphics, or other supporting visuals to be more AI-friendly. High-ranking content (page 1 of Google) contains at least one visual according to this study by Backlinko.
AI is the future – and the present, for the most part – and it’s becoming more important to marketable content than ever. We hope you can use these technologies to your advantage in 2020 and wish you another year of even better content!