Test of AI-logo & CI designing tools
Can the AI-based online logo-design tools replace brand designers? You will find out in my comprehensive test.
You might have noticed that this AI Marketing Hacker blog doesn’t have a logo. The reason is that a blog about artificial intelligence in marketing, of course, must have a logo generated by artificial intelligence. And that’s more work than I expected. To be more precise, there is a default logo of my blog, created by its coder Michal Halaj; I think he made it quickly with the help of artificial intelligence. I embarked on a redesign and, how else, with the help of AI tools.
The biggest aces in the field of visuals are of course Midjourney and DALL-E, followed by Stable Diffusion, and then the rest of the herd. These tools, however, do not have niche positioning. There are a number of those that do, promising to deliver a logotype and complete CI. I decided to look at those that are specifically created for the creation of a logo, CI or even whole brand.
Firstly: what makes a CI?
To begin with, here’s a summary of what comprises Corporate Identity (CI):
- Name and claim (tagline) embodied in typography
- (Optionally) Symbol
- Color scheme
Unfortunately, many companies stop here. When I work on brands, I also strive to include:
- Shapes and forms (form language)
- “Brand assets” (e.g., Ronald McDonald is a brand asset of McDonald’s)
Complete branding also includes:
- Vision/Mission (since working with DesignStudio, I’ve been using a blend of these two)
- Tone of voice
The brief for all tools
The assignment was the same for all tools. Design a logo for “AI marketing consultancy and blog”, include the word Hacker with a big H, and use the following key words for the creative concept:
- + a claim “Cutting through the hype“.
Don’t be scared of the longer text describing the first service. We’ll go briefer and shorter as we go on…
Brandmark.io, with its brand promise, states “Create a unique, professional logo for your business” with the subtitle “Kickstart your brand with business card designs, social media graphics, app icons, letterheads and more”. The service claims to be the only one offering the option of customizing a design for free. This simply isn’t true. Even if it was meant to be “the only one among AI-infused services”, it’s clearly a lie. That’s a really bad card to play for a service that’s supposed to deal with branding, isn’t it… Let’s move on to the test.
You can see the outputs in 3 steps. In the first one, you input the name and slogan, in the second, the key words. In this step, the service suggests that if you don’t know what to input, it has inspiration for you: “keyword ideas” like “robot” “rocket” “planet” “flower” “organic”. This step, in my opinion, is a far bigger problem than it initially seems. If someone is serious about branding, they simply have to ask at the beginning. It’s necessary to ascertain from the client whether they will be expensive, fast, or organic. Sure, a small ice cream shop or corner store won’t do positioning according to Ansoff, Porter, Kotler, and Simon Sinek’s “start with why”, but even graphic designers on Fiverr ask much more than Brandmark.io does. I understand that a startup’s consideration was “in 3 steps, the results should be visible and then we should immediately aim for conversion into a lead”, but this is simply wrong. Regardless, I input several key words and added an axe as a symbol (semantic link to the name Hacker):
I choose several results that I think can be described quite succinctly. The tool generates logos based on the input, primarily on typography, or primarily on the symbol. With short names and claims, after tens of minutes of refining, it copes as well or as poorly as small suppliers on services like Fiverr or Upwork. The outputs are identical based on one (type of) input. By that I mean, until you change the input, which can’t be dramatically changed because it’s always about the name + claim and a set of key words and then clicking in the menu, you technically get the same outputs (e.g., diagonal text + symbol above it). These can be judged purely by taste. Some we can rate as terrible:
Some quite fine:
The service then swiftly develops the logo into a Corporate Identity (CI). This part works excellently for the service: just like with ‘human suppliers’, the logo appears entirely different in the application, and much better than in isolation. The logo above, after tens of minutes of playing with it, applied:
Eventually, it comes down to the worst part: payment. Let’s go through the payment in detail as it’s common for all the following services:
- One output, that is, a CI with source files, will cost you 65 USD
- Forget the lowest, 25 USD version; it includes only a PNG logo and is a career suicide from a working perspective – if you used it for your company, you’d cause more problems and prospective financial losses than you would gain.
- The ‘Enterprise’ version for 175 USD offers, I couldn’t believe my eyes, beyond the medium package up to 10 designs from… a team of human designers.
In conclusion: Brandmark.io is definitely not an AI service. It’s a ‘combinator’ of pre-prepared elements. It works like LEGO, and it can be used to create a basic CI ranging from very poor to ‘standardly below average’ in the ‘for a handful of dollars’ category.
Looka claims to ‘Design your own beautiful brand’ and at first glance, it’s indistinguishable from its competitor, Brandmark.io.
The creation process again begins with entering a name, in the second step Looka asks slightly more about the context than its competitors, in the following steps, it tries to learn more about your taste in style and color palette. Standard stuff.
The next step, selecting a logo symbol, looked very promising. You can either choose from keywords or select a symbol directly. I chose five symbols and was excited to see the promised AI creatively combine the styles of the symbols and what they represent into an abstract symbol. And then came… a big disappointment. The result was, just like with the previous tool, text in various fonts combined with the chosen symbol and color scheme. Everything can then be manually altered. No surprise there.
The output is then a Brand Kit, which includes hundreds of applications from stationary to merchandise to large objects, such as a booth. This is entirely correct and again, the mockups look much better than the isolated logo. The output is comparable, albeit volumetrically much larger, than Brandmark.io. The pricing of the service is also identical, a nonsensical single PNG with a logo for 20 USD, a one-off CI package for 65 USD, and to that a more intelligent subscription pricing for a hundred annually, including a brand package with flexible update possibilities.
Looka offers, if you put in the effort, decent, yet again “LEGO style” outputs. It doesn’t do anything completely wrong, let’s put it this way. And offers smart pricing.
But the question again is: where is the AI?
The SEO domain LogoAI hides another service that claims its outputs are used by 250,000 (!) companies worldwide. Personally, I hope that’s not true, because the result really isn’t good. In 3 steps, identical to those of the previously mentioned services, you end up with a group of admittedly bold, but in my opinion, disharmonious and overall poor logos. There are a number of examples on the homepage, which also are not good. Overall, so far the worst tool, which I would recommend for branding only to my enemy. The pricing is again similar, around 20, 60, and 100 USD.
The logo applied to mockups again looks better than the logo itself. Obviously, if there really is something more than just a promise behind the AI tools, it can be stated that AI is an assistant in drafting, not in designing the basis.
Logomaster promises that you can create a logo (type) in 5 minutes. In my opinion, that’s quite an unfortunate claim: all these tools actually design a logo in 5 seconds, some in 3 clicks and 1 typing of the name, but at least a decent logo that meets even very benevolent ideas takes about half an hour. So, 5 minutes is both too slow and too fast. Logomaster guides the user through the creation process in a slightly different way, starting according to the type of branding (personal, business, non-profit organization, etc.). However, it is then completely standard, we again see the familiar icons (yes, really the same ones as with Looka) and the result of working with the tool are practically identical outputs as with most other tools.
The same old story. 3 steps, a maximum of 3 symbols, no frills apart from the stereotypical color scale descriptions (red is love, passion, energy…), and we’re at the logos.
Logo.com was the first to offer me a bolder graphic approach. Text imitating various vectors (arc, etc.). A “swipe of color” as a background. The icons are again standard, but in the end, using the mix & match system, I was able to prepare some fairly wild combinations. This tool can offer a person who “can’t draw” the feeling that they are expressing themselves creatively in a given context.
And the price? The basic package is for 0. SVG, PDF, PNG, and especially EPS for 0. So, Logo.com really allows you to own a free logo package that you can work with. Yes, it’s not much. Yes, it’s not original. Yes, it’s not going to be a trademark. But for a small business or the first few months of a startup, why not…
Just a question comes to my mind… where is the AI?
By the way, Logo.com also offers a company name generator. It doesn’t even promise AI in this case.
Designs.ai promises a whole suite (the similarity with Adobe Suite is certainly not accidental) of assistants. They are all supposed to be AI-based and offer graphic outputs, both static images and video, audio, and text. The selection of tools is so diverse that it makes up a robust homepage. Some of them present themselves as helpers that are supposed to help you go really deep, e.g. Color Matcher, but in fact, they are just named steps in creating graphics, for which other services have no extra names. The “Font Pairer” is just step no. 3, 4, or 5 in another tool when you are choosing a font for a logo.
Designs.ai has the same brand promise, identical procedure, i.e. in several steps from entering the name of the business, through the selection of sample logos and color scale, user receives the first set of designs. Not in 3 clicks, there are more inputs, but with quick decision making within a minute. The result is really bland, boring, and uncreative output. For virtually the same price, I would definitely go for Logo.com or other tools.
This service offers assistance in setting up a company (LLC) right in the first tab. An interesting proposition. Tailorbrands is also a slightly more complex service than standard “logomakers”. The output is available only after (free) registration. The variability of the outputs is much greater than the standard. Here you really choose between different styles, not just variations on the same style. What surprises me, even Tailorbrands does not create logo symbols by some combination, overlap, but simply slaps a standard, unchanged icon or shape to the typography.
You choose an octagon or an axe symbol, and the exact octagon or axe symbol appears in the logo. Yet AI should excel in an infinite number of small combinations, and thus changes and adjustments.
I’ll save you another same description of the same creation process and just state: even a fifteen-year-old freelancer from Bangladesh takes care of overlapping fonts over each other. Logopony doesn’t. Artificial intelligence?!
On Logomakerr.io, in 3 steps and you have a really ugly creative logo, or a bland, averagely ugly logo. You have the option to edit within 3 days of purchase, a “designer fix” is available for $40. Prices? You’re guessing right, and that’s down to the dollar exactly like most of the competition.
Summary – and the winner is:
As this is already a very long article, as briefly as possible in bullet points:
- All AI tools are an alternative to non-European freelance graphic designers who sell their services on job marketplaces like Fiverr. And they are slightly more expensive than them.
- I would choose at least those tools that allow you to navigate to a reasonable result through editing. The more the tool convinces you that it will solve everything on its own, the worse the result will be.
- “AI” in tools is a marketing label, not a reality. If all these tools are swearing by AI, then Gmail has every right to claim “AI-driven email solution” and Office Word “AI-powered text editor”. Combining fonts with default icons from a photo bank is about as much (artificial) intelligence as I am the Chinese god of war.
- I expected all the tools to deliver a bland combination of fonts, colors, and icons that will be averagely bad and averagely bad in the context of the invested 60 USD. However, some outputs of the tested tools are really bad: typographically bad, with errors like overlapping texts and graphic elements so that they blend into each other, etc. Some also navigate inexperienced users into the tragic trap of “I’ll save and just download the logo in PNG”, which lays the groundwork for big problems.
- Even online tools have confirmed that people evaluate outputs much better in context, i.e. logos always look better on mockups.
- The winner of the test is Logo.com followed by Looka.