Artificial Intelligence Will Check the Beauty Of Your Website

We have trained a neural network to see a website
with the eyes of:

  • webmaster
    who will check if the page code meets the standards important in the modern internet.
  • visitor
    who will instantly form an opinion of the website design and readability of its content.

How it works

A few months ago

We scoured the Internet and selected 12 000 websites — from very outdated websites to landing pages of world-leading IT giants.
Pro designers helped us work out 125 parameters that ordinary visitors use to evaluate the quality of web pages.
A focus group ranked the websites on scale from 1 to 10 so that a machine learning model could have something as a basis.

Now

Our neural network figured out 1000 additional high-level features of a perfect website and the scores of the system have become more precise than the scores of all its teachers taken together.
So we are posting WebScore AI to public access to facilitate its further progress. Right now the system is able to evaluate if a website is attractive in general.

More examples
1
1
2
2
3
3
4
4
5
5
6
6
7
7
8
8
9
9
10
10
110

Show AI the websites that you like or don’t like.
Compare your impression to the score of the artificial intelligence.

What kinds of websites
we can process

  • Landing pages without complex animation

    One-pagers of products, services, events and so on are perfect for the evaluation.

    Learn more

    We use a website snapshot when analysing the design parameters — if a page is heavily animated and a shot the system remembered was not the best one, it may result into a lower score.

  • Business cards and online stores

    Each page of a website should be evaluated separately by sharing direct links with the system.

    Learn more

    Some online stores can get a lower score in case they use too many pop-up windows and have content elements placed too close to each other.

Who needs our service?

  • Website owners

    Here’s a mirror that can help you understand how your site looks compared to other websites. If it doesn’t appear to be the best one — in 2018, uKit AI will update it for free.

  • Beginners in website development

    If you created your website not so long ago, you would definitely want to know what others think of it. Don’t rush out — show your work to AI first and go out there with the result which is truly amazing.

  • Web designers

    Our system imitates an average Internet user: if you create landing pages and websites for the mass audience, you can look at your product with its eyes. Literally.

  • Freelancers

    Convince your clients by saying «Artificial intelligence thinks this design in the best one». Add AI high scores to your portfolio. Use the service if you can no longer look at your website objectively.

How did the project emerge and why?

There’re 1,2 billion websites in the world,

and up to 70% of them might be completely out of date now if their technologies and design weren’t updated regularly.

A problem like this cannot possibly be fixed by humans — it will take too much time and resources. However, that’s something a productive and faster thing like artificial intelligence can handle.

Before
After

This is the reason WebScore AI will soon become a part of a bigger project — a service that can completely update websites with the help of artificial intelligence. We called it uKit AI.

Stay tuned!

Soon uKit AI will be able to come up with new designs for your website and WebScore AI will choose the best of the design options to display.

Update website with uKit AI

Project Team

Roman Steinberg
Data Scientist, Ph.D. in Physical and Mathematical Sciences
Roman developed the concept of the model, in charge of technical management of the project
Yana Valieva
Junior Data Scientist,
Python Developer
Yana is helping with training the system, visualizing and analyzing data
Vladislav Yakovlev
JS-stack Developer
Vladislav develops the assessment tools and tools of collecting website scores
Viktor Kadnikov
Art Director of uKit website builder
Viktor helped with working out website evaluation parameters
Aleksandr Lashchevsky
Project Manager
Aleksandr is in charge of collecting the dataset and assessing it by the focus group for training the system
Aleksandra Lobaskova
uKit team member
Came with «WebScore AI» name and got 60% of team votes among dozen of naming ideas
uKit Group team — 100+ people in 4 cities
The team searches for good and bad websites for the dataset and truly believes in AI technologies
uTeam, Rostov office (Russia)
uTeam, Rostov office (Russia)