Atomic Smash homepage splash

Our wish list for machine learning in websites for 2018

Words byPiers TincknellJanuary 5, 2018

What do the websites of the future look like? Will there even be websites in the future? Will websites write themselves? What will people be viewing websites on? Will Google take over the web entirely with semantic markup? With a new year and new goals for Atomic Smash we have been imagining what websites could be like in the not too distant future.

Technology is advancing at a rapid rate, even though Moore’s Law isn’t technically correct at the moment we are still speedily advancing with processor speeds and computing power. There has been a lot of talk in the last few years about AI or Machine Learning being used in everyday life. We are now at that stage where a lot of the major tech companies have been using machine learning for a few years now.


Netflix uses machine learning to help predict what you would like to watch next.

Black Mirror - Credit Netflix

"We take all of these tags and the user behaviour data and then we use very sophisticated machine learning algorithms that figure out what’s most important. How much should it matter if a consumer watched something yesterday? Should that count twice as much or ten times as much compared to what they watched a whole year ago? How about a month ago? How about if they watched ten minutes of content and abandoned it or they binged through it in two nights? How do we weight all that? That’s where machine learning comes in. What those three things create for us is ‘taste communities’ around the world. It’s about people who watch the same kind of things that you watch." Todd Yellin Netflix’s vice president of product innovation.


Facebook uses machine learning to order and prioritise people’s timelines and serve up suggested adverts.

When you log in to Facebook, we use the power of machine learning to provide you with unique, personalized experiences. Machine learning models are part of ranking and personalizing News Feed stories, filtering out offensive content, highlighting trending topics, ranking search results, and much more.

FBLearner Backbone Diagram


No one is really sure how it is allowed but Instagram uses machine learning to listen to your conversations and then serve adverts based on what you have been talking about.

Facebook, who own Instagram have said they are not doing it on Facebook, but they have not denied doing it on Instagram.

The list goes on, eBay uses it for fraud detection, Amazon uses it for suggested / related products.

It is only a matter of time before machine learning starts to make its way into every other industry.

So we have started thinking about our predictions of how we could imagine machine learning benefitting our clients in the future.


This has been a buzz word online for a few years now, a lot of marketing companies have been talking about personalisation. There has been a huge increase in personalisation being used in email marketing with great results.

Hubspot have put together a list of personalisation experiences used in email that go beyond “Hello [name]” (Even if they do have themselves in the list)

How can personalisation be used online?

There are lots of ways to introduce personalisation into websites that have transactional purchases; if you bought this you might like this. It gets more tricky when your website might not sell anything. If your website is a content delivery platform (like many of our clients) then you can use personalisation to suggest content and reading to your users – If you enjoyed reading this then you might like this other article. When it gets very interesting is if you are able to keep a note of what they have read to suggest the next article rather than manually always choosing the related content. (This is essentially what Netflix is doing with your viewing history)

Dynamic article length based on user history

Other intelligent ways you could look to implement personalisation is for example by the length of articles. If your website detects that a user is not reading articles and are just scrolling through the text why not serve up shorter articles automatically? This could be done by calculating the average read time for a page and determining what is served up based on this. Also you could personalise a website based on the number of times someone has visited you, if they are a repeat visitor they are more likely to respond favourably to an email pop up rather than if they have only visited once.

Other ideas which are more Minority Report

Other ideas for personalisation could be if you get someone to sign up to your website and grab their Twitter handle, if their profile is public you could crawl all their previous Twitter history and serve up suggested content based on their history. You could also grab their photographs and drop them into placeholders across the website. This feels very minority report but would definitely be possible.

Why not take it one step further to take someone’s name and scan their entire online profile, searching through forums, Google Reviews, photographs and build a level of personalisation based on that. In the future machine learning will be intelligent enough to attribute content and online activities to a single person. I am sure the NSA or MI5 have digital profiles of everyone already.

Websites that write their own content?

Every content marketer’s dream or worst nightmare! No more writing long blog posts on topics that you have only just learnt. We believe that in the future it will be possible for websites to generate their own content in an informed manner. Machine learning could do its own keyword research to determine what is a hot topic, it could identify which of those keywords have the least competition and look to write a piece of content based on that topic.

People are already writing software to pen novels so it is only a matter of time before they start writing web copy.

Imagine a truly self generated website

Once the website has written its own copy it can then self edit once it knows how the content is being engaged with and can optimise itself! Is that CTA button not being clicked on? Then move it to the top of the article and see if that works, if that doesn’t work try putting the CTA in the text instead. The potential number of layouts for a web page is a lot! Machine learning could help websites to optimise themselves by changing and testing automatically.

Dynamic navigation

Much like the ability to add their own content websites of the future could also change their navigation dynamically based on social trends or real time events happening in the world. This dynamic navigation could also work to help remove pages / sections that are not getting any engagement by tracking the number of clicks and re-prioritising accordingly. The intelligence would be able to determine if certain links are getting more clicks due to placement or popularity.

Custom alerts!

With the rise of chatbots imagine if you had your own little bot helping you to keep on top of your website management? There are already services offering some of these actions but they are all standalone, the ideal would be to have this level of notification built directly into your website or into a bot.

You haven’t posted a log for a while, a bot reminds you and suggests some content to post.

You have received a new enquiry through your website; through machine learning your bot alerts you to a new business enquiry and prompts you to phone them straight away. If the bot determines that the enquiry is not critical it just sends it through as a normal notification. This could be a great way to respond rapidly to people who are having significant issues with your service or those new business enquiries that need a speedy response.

Other examples

  • Your website has dropped out of the first page of Google, your bot tells you why and what to do to help get back on that first page.
  • You haven’t even logged into your website for a while, the bot can remind you and prompt you to login to check out what has been going on.
  • Some of your content is going viral, your bot can alert you of this and suggest actions accordingly.
  • Some of your content has been stolen and is being passed off by someone else as their own!
  • One of your regular users has not logged in for a while and your bot suggests for you to get in touch with them directly.
  • A new competitor has popped up and their website is outperforming yours, a bot could alert you to this.

In summary

The future of websites is very exciting; with all the technology currently available there will be a lot of opportunities on the horizon for technology to assist us when it comes to creating online platforms. When we think of good website design we think beyond fonts and colours and into technical design of websites.

In 2017 we did quite a few sites that had technical integrations with other systems and we look forward to pushing these technical advancements further in 2018.

Got any ideas you would like to add?

Head over to Twitter and drop us a message!

Tweet Us
Profile picture ofPiers Tincknell

Piers Tincknell

Piers is a co-founder of Atomic Smash and heads up the user experience design and project management in the studio. He is also the best contact for any new business enquiries.

Go back to top

Keep up to date with Atomic news