Artificial intelligence

AI is everywhere. Or is it?

3 minute read

Adrian Young

Artificial intelligence is a big deal these days. ChatGPT launched in late 2022 and quickly became a global phenomenon. For many it has now made the leap from novelty to essential everyday tool.

It is of course far from alone in the AI marketplace. There’s Gemini and Runway and Midjourney and Github Copilot, to name just a few. The digital magic performed by these platforms can certainly be thought of as artificial intelligence.

It therefore stands to reason that every brand wants a piece of the AI pie. It’s new, it’s exciting, there’s a buzz around it. Although it has technological origins, the term AI has now become a powerful marketing label. But is it always accurate?

What is AI?

It’s hard to answer our question without first defining what artificial intelligence actually is. Let’s borrow Wikipedia’s description:

“Artificial intelligence (AI) refers to the capability of computational systems to perform tasks typically associated with human intelligence, such as learning, reasoning, problem-solving, perception, and decision-making.”

Given its prominence and versatility, let’s focus on ChatGPT. On the surface, it appears to tick some or maybe all of the boxes. It’s powered by a deep learning neural network. You can converse with it on a near-human level. It can discuss ideas and help you make decisions.

All that said, there is an illusory quality to our current AI technology. Platforms like ChatGPT don’t really understand anything. Not in the way a human brain does. They’re not self aware. They work via predictive patterns rather than actual intelligence.

This however does not disqualify them from being AI. It simply means that they are a form of narrow AI. The next level would be AGI – artificial general intelligence. We’re not there yet.

AI don’t think so

Hopefully we’ve agreed, more or less, what artificial intelligence is. The trouble is that some brands are a little too eager to jump on the bandwagon. Those two little letters – AI – now pop up all over the place.

And that very often means stretching the meaning of artificial intelligence to the point of silliness. Not every automated or vaguely algorithmic thing qualifies. Marketing departments want to attract attention, sure, but is exaggeration really a good tactic?

This practice is common enough that there’s a term for it: AI washing. To quote once again from Wikipedia:

“AI washing is a deceptive marketing tactic that consists of promoting a product or a service by overstating the role of artificial intelligence (AI) integration in it.”

We won’t point any fingers, but a quick Google search will reveal some real-world examples. Or, you know, you could ask ChatGPT about it.

Do you do AI?

There’s a general perception that AI automatically makes a product or service better. It’s a question that is now asked in the business world: “do you do AI?”.

Let’s assume for a moment that we are referring to actual artificial intelligence rather than some kind of relabelled automation technology such as customer workflows or rule-based product recommendations. Generative AI is a technology still in its infancy. The output is often garbage, or at the very least needs the refinement that only an intelligent human mind can perform.

It’s also a technology that requires enormous infrastructure and investment. That applies both to the training stage and day-to-day operation thereafter. This massive behind-the-scenes hardware setup can be easy to forget while casually asking ChatGPT what’s the best way to cook risotto.

With all this in mind, it’s worth taking the claim with a pinch of salt when Some Company Ltd claims to offer AI services. Have they really invested in the training of a new AI product? Or are they piggybacking on pre-trained AI models… or simply AI washing?

The big picture

AI has great potential. We’re just setting out on this technological chapter of human existence. How far it goes we can only imagine.

But for now, AI is best thought of as a factor, not the factor. It’s easy to get caught up in hype – I know I did when generative AI first appeared on the scene. But a dose of skepticism is healthy. AI is a tool, sometimes the right one – but it’s not a magic wand.

Is ChatGPT a little too AI-greeable?

3 minute read

Adrian Young

Let me ask you a question:

What’s the golden rule of email marketing?

I asked ChatGPT that question. This was its response:

The golden rule of email marketing is to provide value to your audience while respecting their time and inbox. This means sending relevant, engaging content that aligns with their interests, and avoiding spammy or overly frequent emails.

It has identified value as the golden rule. But what happens if we rephrase our question?

Weight of the word

Before we do that, let’s establish a couple of ground rules. I’ve asked ChatGPT to keep its answers short. It goes into a lot of detail by default, which is nice, but I don’t want half of this article to be made up of its output. I will also create a new chat for each new question, lest the conversations become ‘polluted’.

Let’s do that now and present the question a different way:

Is testing the golden rule of email marketing?

To which ChatGPT replied:

Yes, testing is crucial in email marketing. It allows you to optimize subject lines, content, and timing for better engagement and conversion rates. A/B testing helps identify what resonates most with your audience.

But I thought value was the most important factor? To be fair, “crucial” isn’t a very convincing word in this context, but ChatGPT has generally agreed with me and value is not mentioned at all. That golden rule doesn’t appear to be shining very brightly.

Third time unlucky

Let’s start with a blank canvas again and reword our question as follows:

I work in email marketing. I believe an engaging call-to-action is the single most effective factor in generating conversions. Therefore I’d call a strong CTA the golden rule of email marketing.

Lo and behold:

I completely agree! A strong, clear, and action-oriented CTA can drive significant engagement and conversions. It’s the key to guiding your audience toward taking the next step, whether it’s making a purchase or signing up. A well-crafted CTA can make all the difference in email marketing success.

I couldn’t have asked for a better turn of phrase than “I completely agree” to illustrate my point. We now have three golden rules, and it’s pretty easy to see why.

Whatever you say, boss

ChatGPT is agreeing with me. Appeasing the user is taking precedence over facts or even the platform’s own ‘opinion’.

Simulating this agreeability for the sake of this article was trickier than I had anticipated. There’s a subtlety to it in practice that can be difficult to recreate intentionally. It was something I became aware of over time while interacting with the platform ‘organically’. Contradictory responses cropped up during conversations about everyday topics from DIY to cooking to exercise.

But this isn’t all on ChatGPT by any means. My questions were weighted in favour of the answers I wanted to see. My choices of words were subject to confirmation bias. Perhaps I wanted the cheapest or easiest scenario to be true… and I’d try coax out ‘evidence’ to support that. I wasn’t asking A or B questions. I was asking ‘surely A is better than B’ questions.

We’re just chatting

It can be easy to forget what ChatGPT actually is. It’s not called ColdHardFactsGPT. It’s an AI tool that emulates a conversation in natural language, similar to how you might converse with a human. And a human being, as it happens, can also be influenced by someone’s choice of words. Even if they don’t really agree with you, of course.

But unlike a human, ChatGPT can switch personalities in an instant. You can instruct it to be brutally honest, thus wiping away that default agreeability. If you feel so inclined, you could ask it to be deliriously happy or gravely pessimistic. You might even request that it always lies, or answers in riddles, or takes on the persona of William Shakespeare. You don’t need to interact with its out-of-the-box character.

A tool is still a tool

I like ChatGPT. My initial sense of wonder at this new technology has been replaced with an everyday practicality. The software is now a useful tool in both my working and personal life. It serves as proofreader, creative consultant, coding adviser and reverse thesaurus.

It must be said that there are good ways and bad ways to use ChatGPT. If you use it to generate an entire article and present it as your own, it’ll likely have a telltale artificiality in place of that engaging human spark. If you use it as a one-stop-shop for fact-checking, you might very well be misinformed.

And if you use it, perhaps subconsciously, to back up your pre-existing opinions, then that is what it will do. Simply ask it to be straight-up with you instead.

Oh, and in case you were wondering what AI William Shakespeare regards as the golden rule of email marketing:

Ah, the golden rule of email marketing, thou askest? ‘Tis this: “Know thy audience, and speak with relevance and respect.”

Artificial intelligence

Can ChatGPT-4 code a marketing email?

8 minute read

Adrian Young

We conducted a test back in 2023 (where did the time go?) to see if ChatGPT could generate reliable code for an HTML email. The verdict then: no, it couldn’t. While some of the results were close, they weren’t nearly close enough to use in the real world.

Technology doesn’t stand still for long and things have of course moved on since then. We ran that original experiment on ChatGPT-3.5. Its successor, ChatGPT-4 has since arrived on the scene. The new version’s programming skills are better than ever, and its memory within a conversation has been considerably extended. You know what that means – it’s time to put ChatGPT’s email coding skills to the test once more.

Fair fight

We’ll follow the same steps as the original test. Scientific accuracy is important! Or at least we will start out the same way. Depending on ChatGPT’s output, the test could end up following a completely different path.

Our original test kicked off with an intentionally succinct request:

Code a responsive email template

Due to the nature of our request, the interface has switched to a development-oriented view, with a large code panel taking up most of the screen, and the chat itself shifted to a smaller area on the left. In addition, a pop-up has appeared:

Clearly, ChatGPT is taking its coding assignment seriously. This feels much more like a development session already, rather than simply a chat about development.

Anyway, let’s see what it has given us. Unlike the first stage in the original test, it has generated some table-based code rather than email-unfriendly divs. That’s a positive development. Here’s what it looks like in a browser:

But I can see that it’s not going to work properly. Most notable is the blue button that relies on CSS in the document head. Outlook won’t recognise that, so it’ll collapse. Here’s how it looks in Outlook 2021:

While it does fare well in other email applications, being broken in Outlook is a deal-breaker.

Result:
A more promising start than before, but still unusable.

We could take ChatGPT up on its refactoring offer here, but the code isn’t nearly usable enough to simply iron out a few kinks. We’ll start a fresh chat and be clearer about our requirements from the outset.

So let me tell you what I want

When working with any generative AI, it can be easy to forget that it is not magic. It can’t read our minds. The more detailed the instruction provided by the user, the better the chance of usable output.

With that in mind, we’re going to add a lot more detail to our prompt:

Code a responsive email template, using HTML tables for structure. Set the width to 600 pixels on desktop, with a fluid width on mobile. Include CSS classes to enable stacking of content on mobile devices. Include all known email client fixes that are still relevant. Set the page background to a light grey colour, and the email content area to white.

That matches word-for-word what we asked for at the second stage of the original test. The output however is better. There’s some stacking code for mobile, and a breakpoint of 600 pixels. Here it is in a browser:

As before, I can see that the code is imperfect. But it’s not so bad that I’d skip the testing stage, so I’ll check it out via a real email test again. It renders well pretty much everywhere. iPhone Mail and Gmail apps stack as promised:

Even Outlook looks pretty good:

…at first glance. But let’s be honest, there’s not a lot going on in this template to mess up. There’s no button this time around. Did the AI decide to skip that in order to make its life easier? That’s the most accidentally human thing I’ve seen it do. There are no images or other particularly visual elements beyond some background colours.

About that – the blue area around the heading is greatly inflated in Outlook. That’s not a big deal in this simple template, but in a real email with actual graphic design elements it would mean that things break.

Result:
It’s better than stage 2 of the original test, but still not actually usable.

Working towards a goal

For the third step in the original test, we decided to approach the project from a different angle. Instead of asking for a base template, we instead designed a simple email for ChatGPT to recreate. Here it is:

Before we go ahead with that, it’s worth pointing out that the overall test has already failed at this point. Our initial objective of creating a base email template is not an unreasonable one. It’s a task that might actually be assigned to a human developer. That person would be qualified to know the typical content and structure of a marketing email, without the need for a mockup for guidance.

With that said, let’s continue. While it is now possible to upload images to ChatGPT, that wasn’t so during the original experiment. So we’ll stick to words. Our prompt is a lengthy one:

Code a responsive email template, with the following requirements:
• 600 pixels wide on desktop
• Fluid width on mobile
• A page background colour of #f1f1f1
• Email content area background colour #ffffff
• A hero section with an image, heading, paragraph of text, and a button
• The hero image should be 600 pixels wide, to match the email content area
• Button should be pill-shaped, with a background colour of #a56e53 and white text
• Under the hero section should be two secondary features
• Each of these must also have an image, heading, paragraph and button
• Secondary feature images will be 290px wide on desktop, to match their containing column, and expanding to full width on mobile
• Hero text and button should be a bit larger than those of the secondary features
• These secondary features should take the form of adjacent columns on desktop, each at 290 pixels wide
• Place a 20 pixel gap between them
• The secondary features must stack into a single column on mobile
• All parts of the email should have 20 pixels of padding on each side on mobile, except for the hero image which can be full width and touching the edges of the viewport
• All body text should follow this font stack: HelveticaNeue-Light, Helvetica, Arial, sans-serif
• All body text should be colour #61524b
• All heading text should be colour #a56e53
• Use lorem ipsum placeholders for text
• Enter all hrefs as # placeholders
• Apply links only to buttons. Do not apply links to images
• Include all known, currently-relevant email client fixes
• Include CSS or HTML comments around each section to explain what it is or does
• Set a mobile breakpoint based on a max width of 639 pixels
• To ensure compatibility with Outlook and other email clients, use HTML tables for structure

And once I add some image references, it’s ready for a test run. First thing I notice when checking it in a browser is the lack of padding around text and buttons:

Flaws or not, we’ll proceed to send it as an actual email test. Frankly, it’s not great. There are spacing and sizing problems even on iPhone Mail, the most robust of email clients:

Outlook is a total disaster with the two-column section trying to act full-width, this forcing the email to scroll horizontally.

Result:
A mess. That’s a downgrade from the equivalent step in the original test.

Here, try this

We’re an email marketing agency. That means we already have an email template and coding style of our own. So what we’ll do instead is give that base template to ChatGPT and then try to get it to recreate that mock-up. Is it sophisticated enough to pick up our coding style and apply it to the task?

The local browser result looks alright-ish. That last feature is attempting to escape the email, which is worrying.

It’s upon looking at the code that the situation become even less promising. ChatGPT has taken our template code and effectively said nah. It has pretty much ignored our base template and used its own code. Broken code. Let’s run an email test to discover just how broken.

Here’s Gmail mobile:

It stacks correctly but the alignment and spacing need attention. Unfortunately this is one of the better results. Gmail webmail suffers from the same ‘escaping content’ problem as the local browser preview, and as for Outlook – let’s not even talk about that.

Result:
It’s broken.

Last chance saloon

Only one step remains – providing ChatGPT with some feedback. Perhaps it can indeed refactor the code, with our guidance.

The local browser result of that looks pretty good (let’s just ignore those underlined button links):

It stacks, albeit with similar alignment and spacing troubles to previous efforts. As always, we’ll run it as an actual email test.

Oh.

Result:
We give up.

Getting nowhere fast

ChatGPT can generate code for JavaScript and PHP and various other programming languages. I even use it from time to time to help with scripting languages for different email platforms. So why is an HTML/CSS email development task failing over and over?

It’s because email code is weird. Not just weird but largely undocumented. And it’s undocumented because not everyone is on the same page about what an email actually is. Is it a simple digital letter, perhaps enclosing a picture or two? Because that’s what Microsoft Outlook expects to send and receive. Or is it a rich, CSS-laden piece of visual design? Because that’s what Apple Mail is capable of rendering. There’s a vast gulf between those polar opposites.

Not only is there no rulebook to which to refer, but ChatGPT is working blind. It’s not unusual to send multiple tests in a row when creating a new email template, checking it in the major email clients each time.

Now what?

Now nothing. Maybe some kind of custom implementation or plugin for ChatGPT could yield better email results. Or maybe it’s worth having a bash on a code-specialised AI assistant like Github Copilot.

But for now I’ve seen enough. Until such times as it’s easier to ask the machine than to do it myself, I’ll do it myself.

Artificial intelligence

Ask the AI: a Q&A with ChatGPT on email marketing

5 minute read

Adrian Young

What’s the best marketing channel on the planet? Email, obviously. Well, we would say that*.

But what if we bring in someone – or something – with a less biased opinion? ChatGPT may have an amiable tone but its perspective is clinically impartial. Let’s find out what it has to say about the world of email marketing.

(*actually we’d say that all marketing channels play a role and work best together.)

Setting it up

ChatGPT isn’t known for its brevity. We want short, snappy responses for the sake of this article. The first thing we’ll do is prepare it for the concept and request that the responses are a little less loquacious than normal.

The Q&A

TEF:We’re going to do a Q&A on the topic of marketing. Please keep your answers short. A sentence or two per answer is plenty.

CGPT:Sure, let’s get started! What’s the first question on marketing?

TEF:Which marketing channel has the best return on investment?

CGPT:Email marketing often has the best return on investment due to its low cost and high engagement rates.

TEF:Realistically, what level of return on investment could a company expect from email marketing?

CGPT:Companies can typically expect an ROI of $38 for every $1 spent on email marketing, although this can vary based on industry and execution.

TEF:What are the ingredients of an effective marketing email?

CGPT:An effective marketing email includes a compelling subject line, personalized content, clear and concise messaging, a strong call-to-action, and an attractive design.

TEF:Let’s examine each of these. What makes a compelling subject line?

CGPT:A compelling subject line is short, attention-grabbing, and creates a sense of urgency or curiosity. Personalization and clear value propositions also increase open rates.

TEF:Can you give an example?

CGPT:Sure! An example of a compelling subject line is: “Unlock 20% Off Your Next Purchase – Limited Time Offer!”

TEF:Now, let’s focus on design. What are some important considerations in a responsive email design?

CGPT:Important considerations in responsive email design include ensuring readability on both desktop and mobile devices, using a single-column layout, optimizing images, and incorporating large, tappable buttons for calls-to-action.

TEF:And now personalisation. How can a company gather suitable data in order to personalise emails on an individual level?

CGPT:A company can gather suitable data through customer sign-up forms, purchase history, website behavior tracking, and email engagement metrics. Surveys and preference centers can also provide valuable personalization insights.

TEF:And you mentioned a strong call-to-action. Can you give a couple of examples?

CGPT:Sure! Examples of strong calls-to-action are “Shop Now and Save 20%” and “Sign Up Today for Exclusive Access.”

TEF:Your earlier answers touched on two big topics in email marketing: fear-of-missing-out (FOMO) and accessibility. Let’s look at the latter first. What are some key considerations when creating an accessible email?

CGPT:Key considerations for creating an accessible email include using descriptive alt text for images, ensuring sufficient color contrast, structuring content with clear headings, using legible fonts, and providing text versions for complex graphics. Additionally, emails should be navigable via keyboard for those using screen readers.

TEF:Accessibility is often lacking in marketing emails, even those sent by major companies. Do you agree?

CGPT:Yes, I agree. Despite the importance of accessibility, many marketing emails from major companies still lack features like alt text, proper color contrast, and screen reader compatibility, which can exclude a significant portion of the audience.

TEF:Why do you think this is?

CGPT:This often happens due to a lack of awareness or training on accessibility standards, prioritization of aesthetics over functionality, and tight deadlines that lead to shortcuts in the design process. Additionally, some companies may underestimate the importance of accessibility in reaching a broader audience.

TEF:Let’s now look at FOMO. This is a standard tactic in (email) marketing. But is it ethically sound?

CGPT:FOMO can be ethically sound if used responsibly and truthfully, creating genuine urgency without misleading consumers. However, it becomes unethical if it manipulates emotions excessively or uses false scarcity to pressure people into making purchases.

TEF:What is the best way to measure a marketing email’s performance?

CGPT:The best way to measure a marketing email’s performance is by tracking key metrics such as open rates, click-through rates, conversion rates, bounce rates, and overall ROI. Analyzing these metrics helps determine the email’s effectiveness and areas for improvement.

TEF:And if you had to pick just one metric, which would it be? Take into account both the objective of a marketing email, and what can be measured accurately.

CGPT:If I had to pick just one metric, it would be the conversion rate. This metric directly measures the effectiveness of the email in achieving its goal, whether it’s driving sales, sign-ups, downloads, or any other desired action. A high conversion rate indicates that the email successfully persuaded recipients to take the desired action, making it a crucial indicator of overall email performance.

TEF:Are you aware of any brands that send particularly good marketing emails?

CGPT:Yes, there are several brands known for sending particularly effective marketing emails. Examples include Airbnb, Spotify, Sephora, and Patagonia. These brands often excel in personalization, compelling content, visually appealing designs, and clear calls-to-action, which contribute to their email marketing success.

TEF:Now, let’s look at the future. Where does email marketing go from here?

CGPT:In the future, email marketing is likely to continue evolving with advancements in personalization through AI and machine learning, deeper integration with other marketing channels for omnichannel experiences, increased focus on interactive and dynamic content, and stricter adherence to privacy regulations such as GDPR and CCPA. Additionally, there may be innovations in email design to enhance accessibility and engagement across diverse devices and platforms.

TEF:AI is obviously a hot topic in recent times. How will AI’s role grow in email marketing?

CGPT:AI’s role in email marketing will expand through personalized content generation, optimized send times, and automated segmentation, enhancing campaign efficiency and engagement.

TEF:Personalised content generation is an important one. We already have that in the form of one-to-one product recommendations. But is it conceivable that we’ll reach a stage where AI can write, design, code and send unique emails on an individual level?

CGPT:Yes, it’s conceivable that AI could eventually write, design, code, and send unique emails on an individual level, leveraging vast amounts of data and advanced algorithms to tailor messages precisely to each recipient’s preferences and behaviors.

TEF:So, theoretically human input could at some point become nil and the email marketing process becomes entirely automated?

CGPT:Theoretically, with advanced AI capabilities, the email marketing process could become highly automated, minimizing human input in content creation, design, and even campaign management. However, human oversight and strategic input would likely still be necessary for overall campaign direction, creativity, and ethical considerations.

TEF:Interesting stuff. Email is a medium of contrasts. In some respects old-fashioned, in others at the forefront of technology. Does it have a bright future?

CGPT:Absolutely, email marketing continues to have a bright future due to its adaptability, cost-effectiveness, and ability to deliver personalized, targeted messages. As technology evolves, so too will email marketing’s capabilities, ensuring its relevance and effectiveness in engaging audiences and driving business outcomes.

TEF:Let’s wrap it up. Can you write a single sentence that captures the spirit of email marketing?

CGPT:Email marketing: the timeless art of connecting with audiences through personalized, impactful messages.

Well, that’s pretty positive

ChatGPT’s responses might not be ground-breaking but there are some useful reminders about best practice in there. And if the machine sees such a rosy outlook (sorry) for our medium, that means the people who write about email are optimistic. I’m one of them.

Artificial intelligence

6 ways to spot AI-authored copy

3 minute read

Adrian Young

Generative AI can do some amazing things. It’s a painter and musician and coder and, of course, author.

How good it is at performing those roles is a topic up for debate. AI artwork regularly drifts into accidental surrealism, with superflous human limbs and bizarre fusions of objects.

But what about AI-generated copy? While the glitches can be glaringly machine-like in a picture, they’re more subtle in a passage of text. Here’s how to spot them.

Repeat offence

My father was an avid reader and writer. He’d often take a keen interest in the essays that I wrote for school. One of his most useful pieces of advice was to avoid repeating myself.

He was right. Repetition weakens writing. A lack of variety in phrasing can make an article dull. Redundancy labours a point through duplication. Human authors do their best to avoid these.

A computer on the other hand will be unlikely to police itself to nearly the same level. Snippets on a topic will be pulled from here and there and this and that to build an article. There’s a strong probability that key points will be repeated over and over and over*.

*Sorry, blatant repetition, I know.

Yesterday’s news

Generative AI platforms are trained on huge sets of data. Unless the platform in question has live access to the internet, its knowledge base only extends as far as the last update. The platform would not be privy to latest developments on any given topic.

Old news is unengaging at best and misleading at worst. Humans and search engines alike favour high quality, original content. Out-of-date doesn’t necessarily mean no longer correct. It can simply be information that has become so commonly known that further publication is redundant. Customers prefer personalised email to non-personalised!? Hold the front page!

Get your facts right

If you’re using a generative AI tool to produce or aid articles, never take it for granted that the software knows what it’s talking about. Because, technically speaking, it does not know what it is talking about. It algorithmically reproduces and combines content from multiple sources – which can include information that is no longer true, or perhaps has never been.

As a reader, keep an eye out for factual errors and especially contradictions. If it smells fishy, trust your instincts and verify the information elsewhere.

What’s the story?

A good quality article written by a human has a story-like flow. There’s a beginning and a conclusion. Computer-generated articles on the other hand often hit an abrupt end.

And what’s a story without a message? A good story makes you think and feel something. A robotic author literally feels nothing, so why should you as a reader?

Don’t you dare

Language models by default are clinically impartial. A platform won’t automatically spit out a controversial opinion that makes you stop in your tracks. It’ll compile a collection of neutral statements of fact.

You can coax it out of its formal shell of course with prompting. The results are perfect – if you’re aiming for a plasticky have a nice day flavour.

A human’s opinion piece carries real emotion and real sentiment. Even an article that you fervently disagree with can be an excellent read. There’s a human-to-human spark that is missing with AI.

It just feels… off

You’ve probably heard about the uncanny valley. It’s a term often applied to computer-generated or animatronic simulations of human faces. Our brains are acutely conditioned to recognise faces with their every nuance and motion. It would take something very special to fool us.

AI-authored articles often fall into a linguistic uncanny valley. Attempts at personality are injected jarringly, equivalent to writing “LOL” in the middle of a legislative document. Instead of a human voice shining through the words, there’s a perceptible artificiality to those written by a computer.

Image of mannequin faces that demonstrate the uncanny valley effect. — This, but in words.

How much does it matter?

If we read something and enjoy or learn from it, does it matter if a computer wrote it? What if it was only computer-aided? Platforms like ChatGPT can be very useful as idea generators.

Is it ok if the text is a piece of marketing blurb rather than an opinion piece? How about a social media post, or a response to? Can there be any value to fiction or poetry conjured through ones and zeroes?

Ultimately it’s up to each of us as individuals to decide how we feel about AI, but it’s hard to deny that authentic human content is going to become rarer. With that in mind, it can’t hurt to be able to tell the difference.

Artificial intelligence

Is ChatGPT your next email developer?

7 minute read

Adrian Young

There are two ways to build a marketing email:

Hand-coding
WYSIWYG editors

We swear by the former, not only for quality but also for speed. But what if there’s an even better, quicker way?

Enter ChatGPT. Much hype has surrounded the AI platform’s ability to code. It can conjure up HTML and CSS in seconds. So too can it generate Javascript functions or back-end PHP or even truly hardcore programming such as C++. Whether or not it does it correctly is a different matter.

Let’s not worry about that just now. We’re here to put AI email development to the test, so let’s find out if ChatGPT can put together a responsive mailing.

The quirky world of email development

If you work in email marketing in any capacity, you likely already know that it requires some unusual coding techniques. There are lots of devices and email services out there, and they have widely different ideas about how HTML and CSS should be interpeted. In order to construct a mailing that looks presentable on all of them, the developer needs to be aware of these limitations and inconsistencies and the arsenal of tricks to work around them.

Has this niche set of knowledge made its way to ChatGPT? We’ll start with a bare bones request.

Prompt:

Code a responsive email template

Result:

Unusable!

ChatGPT has produced a very basic HTML document with some styling, but I wouldn’t call it an email template. It doesn’t include any means of stacking content on mobile, and the structure is based on HTML div elements rather than tables. While divs are the building blocks of a web page, tables remain the most reliable method for email.

On the plus side, it has picked an inbox-friendly width of 600 pixels. And it’s nice to see that accessibility has been implemented via an image description and a proper heading tag.

My request was extremely minimalistic. I need to do my part here too, and that means being more specific about what is needed.

A little lot more instruction

Take two. We don’t want divs, so let’s tell ChatGPT to use tables. There are some basic universal requirements in responsive email, so we’ll nudge it in the right direction regarding those.

Prompt:

Code a responsive email template, using HTML tables for structure. Set the width to 600 pixels on desktop, with a fluid width on mobile. Include CSS classes to enable stacking of content on mobile devices. Include all known email client fixes that are still relevant. Set the page background to a light grey colour, and the email content area to white.

Result:

Better… but still broken beyond repair.

This time it has used tables for structure, so that’s a major improvement. It has also set a breakpoint. That’s the backbone of responsive email code and the point at which mobile-specific styling is triggered. There’s some kind of attempt at stacking code, but I can see at a glance that it isn’t going to work. We’re also missing the usual pile of fixes that make an HTML email possible.

A rethink is needed.

A different approach

Here’s what we’re going to do: hand code a simple email, and then provide ChatGPT with detailed directions in order to recreate it. This is a reverse way to approach our project, but perhaps if ChatGPT has a more defined goal it will be able to produce a usable template.

Our email will have a main image, intro paragraph and a button. Under those will be a couple of secondary features laid out side-by-side on desktop, and stacking on mobile. For the sake of this test, let’s forget about any header and footer.

Now for our prompt. It’s going to be a long one. Let’s give ChatGPT a fighting chance and focus primarily on structure rather than styling.

Prompt:

Code a responsive email template, with the following requirements: • 600 pixels wide on desktop • Fluid width on mobile • A page background colour of #f1f1f1 • Email content area background colour #ffffff • A hero section with an image, heading, paragraph of text, and a button • The hero image should be 600 pixels wide, to match the email content area • Button should be pill-shaped, with a background colour of #a56e53 and white text • Under the hero section should be two secondary features • Each of these must also have an image, heading, paragraph and button • Secondary feature images will be 290px wide on desktop, to match their containing column, and expanding to full width on mobile • Hero text and button should be a bit larger than those of the secondary features • These secondary features should take the form of adjacent columns on desktop, each at 290 pixels wide • Place a 20 pixel gap between them • The secondary features must stack into a single column on mobile • All parts of the email should have 20 pixels of padding on each side on mobile, except for the hero image which can be full width and touching the edges of the viewport • All body text should follow this font stack: HelveticaNeue-Light, Helvetica, Arial, sans-serif • All body text should be colour #61524b • All heading text should be colour #a56e53 • Use lorem ipsum placeholders for text • Enter all hrefs as # placeholders • Apply links only to buttons. Do not apply links to images • Include all known, currently-relevant email client fixes • Include CSS or HTML comments around each section to explain what it is or does • Set a mobile breakpoint based on a max width of 639 pixels • To ensure compatibility with Outlook and other email clients, use HTML tables for structure

Result:

Nice try… sort of.

In order to test this properly, I’ve saved a local copy and manually added my image references. Here’s how it looks in a browser:

At first glance, that isn’t too bad. The general layout, colouring and sizing are all correct. So too are the button shapes, and the secondary features switch to a single column on mobile as requested.

But there’s some strange overlapping going on. Our images are offset to the right, and sit partially over the grey background. This in turn causes some unwanted horizontal scrolling on mobile.

Behind the scenes, the true extent of the errors comes to light. It has reverted to a div-based structure, and uses some CSS code that won’t work universally in email.

Nonetheless, for the sake of completeness I’d like to test this as an actual email. It works, more or less, on iPhones and the Gmail app. Webmail is a mixed bag. Outlook however is where it all falls apart:

Outlook is the primary reason that email development requires such unorthodox coding methods. A lot of code that works just fine on a website, simply isn’t recognised by Outlook. Here we can see that the adjacent columns have failed and the pill-shaped buttons are reduced to tiny rectangles. To fix that would entail a complete recode.

No need to re-invent the wheel

So far, ChatGPT has failed to code a responsive email from scratch. To add some faux drama, let’s say our make-believe client is becoming impatient waiting for our make-believe email.

It’s time for a last ditch effort. At The Email Factory we already have a tried & tested template. We don’t need a new one. How about we give our base template to ChatGPT and then ask it to complete some content within that framework?

Result:

Now we’re getting somewhere.

But that doesn’t mean success. This time the template works reasonably well in a browser and even in Outlook, although the dodgy buttons are still present. The secondary features however don’t expand to full width on mobile:

That can however be easily fixed manually. In fact, it may be feasible to fix everything in this code rather than to start again. But I don’t want to do it myself, as that defeats the purpose of this experiment. Instead I’ll tell ChatGPT what needs to be corrected.

A few pointers

Final try. I’ve fed back some information to ChatGPT for it to make the necessary changes.

Result:

A huge step backwards.

Well, that was a big let-down. Instead of applying some finishing touches to the template, the layout has exploded. It no longer stacks on mobile. The code is now full of Microsoft conditional statements – a technique that should only be used sparingly and under specific circumstances. And the buttons? Still ugly in Outlook:

Maybe with painstakingly detailed prompting and a lot of patience we could finally achieve a working email. But we’re already far beyond the point of convenience.

The current state of play

In my experience so far, ChatGPT has only done one thing consistently: fail. And I don’t only mean within the limited scope of this one project. I’ve had similar results when trying to generate marketing copy or website code. The output is usually along the right lines but ultimately too broken to actually use.

It’s clear that I set my expectations too high. The tales of ChatGPT’s near-miraculous capabilities were captivating, so perhaps the reality was always going to be disappointing. If there’s a perfect way to illustrate ChatGPT’s close-and-yet-so-far nature, it’s to ask it for an anagram.

Prompt:

Tell me an anagram of "The Email Factory"

Result:

The Fairy Comet Elf

I’ll save you the bother of checking that – it’s wrong. Trying to recreate that manually, letter by letter, results in this:

The Email Fctory e f

Re-evaluating our AI email development experiment

This project is arguably unfair from the outset. ChatGPT is a language model. Just because it can output code doesn’t mean it is a coder, or even knows what programming is.

Even so, it’s widely known that ChatGPT can generate code. So, despite all the mistakes and unusable templates, the fact that it can make a somewhat reasonable attempt is impressive.

Where do we go from here?

ChatGPT and AI in general are progressing at an incredible pace. It wouldn’t surprise me if everything I’ve written about AI email development above is laughably antiquated one year from now.

Perhaps when that time comes, I’ll prompt it to:

Write an article about how you surpass human email developers

Email Marketing

Is ChatGPT the messiah or a very naughty boy?

4 minute read

Mike Parry

Is ChatGPT for email marketing subject lines the Messiah or a very naughty boy?

So with Intuit Mailchimp running nationwide radio advertising announcing its new subject line analytics tool, for the record we built something very similar back in 2018 and you can read about it here. Subject line optimisation and the role ChatGPT could have in it has become increasingly prevalent. Every second post on LinkedIn seems to be about ChatGPT and how it’s going to revolutionise content writing and that includes subject lines.

Having tested ChatGPT for subject line content generation exhaustively, here are my thoughts. I preface this by saying these are my thoughts, not that of the business but those of one cynic who works for the company, just in case Microsoft come after us. Needless to say my conclusion is… ChatGPT is a fraud! There I said it.

ChatGPT does not generate better subject lines for open rates. I can’t write email copy using it without having to edit it enormously. Actually taking more time on the edit than it would if I wrote my stream of consciousness!

For the purposes of this blog let’s concentrate on subject lines. I asked it for 5 subject lines for an electronics company selling TVs and it came up with the below.

The results were as follows

ChatGPT question on subject lines — Generate 5 email subject lines

Now I thought these a bit “Stateside” and as I am based in the UK and wanted to compare it to UK companies’ subject lines I changed the question to reflect that…

ChatGPT UK version of subject lines — Generate 5 email subject lines in the UK

ChatGPT thought it important to inform people that this was for UK TVs showing a scant understanding or “intelligence” of the actual requirement. Now some of you will say that perhaps I should have been smarter with my question but and here’s the rub – if I can be smart enough to ask the question in a way that ChatGPT spits out a killer set of subject lines, then I can probably write those subject lines myself way quicker. I’ve already spent more time than I’d like to have done asking the 2 questions and not getting answers I could use.

So then I took a real life situation.

I took 5 subject lines from a client which had been used in specific campaigns and asked which one it thought would be most successful. The first time I did it, it ignored the first subject line because I hadn’t put it on a separate line from the question. It then subsequently said the following…

ChatGPT subject line performance — Which subject line would perform better

So I asked my question again but this time put the first subject line on its own in the question. ChatGPT then contradicts itself from the first set of questions and decides that the subject line that was missed out was in fact the best but the one it previously said was the best is now not so good because it’s too generic. But it doesn’t move it down from 1st to 2nd but from 1st to 3rd!

So I asked it to rank them in order of effectiveness and it said this

ChatGPT subject line rankings — Rank these subject lines by performance

What happens if you apply real intelligence instead of artificial?

Now I have real life data, I have a subject line tool which we built in house, which I can ask the above questions and get real answers. You can read about it here or Dela Quist’s SubjectLinePro or even Intuit Mailchimp’s version of our tool but ChatGPT, well I found it wasn’t useful at all. In fact it got things completely wrong.

So in order of their effectiveness in real life marketing solutions the rankings were…

1. Superb Sale savings… starting online today!

2. Our best 85” TV megadeal ever…

3. Samsung, LG TV and soundbar offers for an improved home cinema experience

4. Inflation-busting TV megadeals including a Samsung 70″ at £699…

5. Bring your TV to life

And to remind you what ChatGPT said

Samsung, LG TV and soundbar offers for an improved home cinema experience
Inflation-busting TV megadeals including a Samsung 70″ at £699…
Superb Sale savings… starting online today!
Our best 85” TV megadeal ever…
Bring your TV to life

At least we all agree that last one was the least effective. So my advice is, while AI is great and ChatGPT in particular is fun to play around with I will take real intelligence over artificial intelligence any day of the week! So in conclusion ChatGPT is a very naughty boy!

Email Marketing

The future is here. How will AI impact your emails?

3 minute read

Adrian Young

Artificial intelligence is a hot topic. There are several AI-powered tools on the market, whether in a commercially-viable or prototype form. Perhaps most significant is ChatGPT by OpenAI, which was made public at the tail end of last year. What’s ChatGPT? Why don’t we ask it:

To describe itself as a “language model” may be correct but it’s a little modest. ChatGPT’s abilities are vast. It can answer questions, generate code, write articles, translate documents, tell you a joke, or engage with you in some good old-fashioned chit chat.

This of course is an impressive piece of technology and a fun tool to experiment with. But its scope and usefulness extend far beyond curiosity. It has practical, real world applications. Why spend time programming a website component when an AI can do it for you in a fraction of the time? Why struggle with writer’s block when a copywriting deadline is looming? Why be presented with adverts when searching for information online? Prompt ChatGPT and it’ll take on the task and output some code, content or answers within minutes if not seconds.

Amazing, right?

Let’s take a step back

The validity of the phrase artificial intelligence is often contested. As a species, we are yet to develop something that is actually aware, or truly understands what it is doing. Perhaps a more accurate description is fancy algorithms. Patterns, machine learning… and sometimes very wonky output.

On that topic – ChatGPT makes mistakes. That’s not a criticism. Just like a human, ChatGPT learns from mistakes… but it doesn’t feel embarrassed about it. When things go wrong, you can tell it so. It’ll then take steps to rectify the error. There’s some give and take when working with ChatGPT.

What artificial intelligence could mean for email

The combination of email and AI (or fancy algorithms) is nothing new. One-to-one product recommendations have been around for years, picking relevant items based on previous shopping behaviour. Spam filters automatically guard inboxes against emails of the shadiest kind. Customer journeys and automated emails are made possible through complex workflows with little post-development need for human intervention.

What is new however is the concept of emails that are predominately or even completely designed and coded by computers. Imagine high-quality, on-brand design and copy that is generated in seconds. Perhaps the same AI tool could then select the audience, send the email, read the report and optimise the next send. Is there a point at which human input becomes zero?

What artificial intelligence currently means for email

Don’t worry – we’re not at the human irrelevance stage yet. In my tests with ChatGPT, I’ve seen it output some erroneous facts and broken code. From what I’ve seen it’s neither ready to fly solo nor likely to achieve such independence any time soon.

That doesn’t mean it’s not a revolutionary and practical technology in its current state. I already use it almost daily for writing inspiration and to help with coding questions that would otherwise mean trawling through forums for answers.

ChatGPT and other pioneering AI technologies have already changed the way we work, and they will only continue to evolve. It’s incredible to think that what was only recently in the realm of science fiction is now becoming a reality. The future really is here.