How AI Is Changing Email Marketing

“AI in email marketing” has become one of those phrases that means everything and nothing at the same time.

For one marketer, it means using ChatGPT to write subject lines. For another, it means predictive segmentation inside their CRM. For someone else, it means the ESP automatically choosing send times, building variants, and personalising blocks at scale.

And that’s the first truth we need to anchor:

There isn’t one “right” way AI shows up in email marketing, and you DON'T NEED TO either. 

There are lots of ways. Some are genuinely useful, some are not, some are actively dangerous if your foundations are shaky.

This blog is about what’s actually changing, what’s being overpromised, what spam filters and inbox providers are doing, where AI can genuinely help email teams and where it absolutely cannot replace critical thinking.

What “AI in email marketing” actually means (because everyone’s using the term differently)

When we say “AI”, most people picture a chatbot writing copy. But AI in email marketing is broader than content generation. In practice, it usually falls into these buckets:

1) Generative AI for content

This is the obvious one: tools that generate subject lines, body copy, variants, CTAs, previews, and ideas.

The upside is speed and momentum. The downside is sameness and lazinessif you're not careful. If your email programme/function already lacks clear strategy, audience insight, and message relevance, generative content doesn’t fix that. It just helps you publish more of the same.

2) AI applied to data and audience understanding

This is where things get more interesting: enrichment, modelling, scoring, predictive intent, lifecycle classification, and better segmentation based on patterns humans struggle to see quickly.

HubSpot’s Breeze is one example of a platform positioning AI as a layer across the CRM experience (including data capabilities and automation support).

3) AI decisioning and orchestration

This is AI making choices about who gets what, when, based on rules, signals, and predicted outcomes. It’s not “write me an email.” It’s “decide what matters next.”

This is where tools like:

• Hightouch (data activation/personalisation & targeting)
• Braze (including its Project Catalyst initiatives)
• Iterable (with Nova positioning around AI capabilities)

 …tend to sit.

4) AI inside the inbox (the shift you're not ready for)

This is the part people keep ignoring: email isn’t only changing on the sender side, it's changing on the reader side.

Google has been rolling out Gemini for Workspace features that affect how people process information inside Gmail (including summarisation and assistance experiences).

Whether every feature is available in every region and account type yet, the direction is obvious: the inbox is becoming an AI mediated interface. And that changes how your emails get noticed, interpreted, prioritised, and remembered.

The biggest lie in the AI-email conversation: “AI will fix performance”

It won’t, it won't, it really, really won't. 

AI can speed up production, optimise distribution, and even help you model intent. But email performance still depends on fundamentals:

• Deliverability (are you landing where people can actually see you?)
• Audience expectations (did they opt in for this?)
• Relevance (does it feel for them, right now?)
• Timing context (is this message arriving when it makes sense?)
• Data quality (are your segments real or imaginary?)
• Journey cohesion (are you sending the right thing, not just a thing?)

If those conditions are wrong, AI doesn’t rescue you. It accelerates your decline — because it helps you do the wrong things faster.

That’s why the line I want you to remember is:

You get out what you put in.

If your strategy is weak and your data is messy and systems are rubbish, AI will “optimise” you into a more efficient (or not really) version of this. 

How marketers are using AI right now (and where it helps vs where it backfires)

AI is being used as a pressure release valve

Let’s talk about what’s actually happening inside teams.

A huge amount of “AI adoption” in email marketing isn’t coming from a strategic roadmap. It’s coming from a Monday morning Slack message that says:

“Can we get an email out today?”
 “We need to hit the number.”
 “We’ve got to promote this.”
 “Sales wants something in inbox by 2pm.”

And if you’re the person responsible for email, you already know what that creates:

• You’re under pressure to produce output at speed
• You’re inheriting a channel with inconsistent data
• You’re working with half-built journeys
• You’re juggling internal demands that have nothing to do with subscriber expectations
• And you’re getting judged on short-term spikes like email is a paid ad

So of course people reach for tools.

Not because they’re lazy. Not because they’re stupid.

Because they’re trying to survive a system that treats email like a megaphone and a vending machine at the same time.

The “AI shortcut” trend I’m seeing everywhere

Here’s the pattern I keep seeing:

AI is not the strategy layer, it’s the execution layer

If you want to use AI without making your email programme worse, put it in the right place.

Until AI can suddenly contextualise humans and is wayyyy better, it needs to leave strategy to the experts. 

AI is brilliant at:

• drafting
• restructuring
• generating variants
• summarising performance
• clustering feedback
• supporting segmentation work
• mapping content to journey stages

AI is not brilliant at:

• deciding what your brand should say
• defining audience truth
• understanding what permission you’ve earned
• resolving collisions between marketing, sales, service, product comms
• knowing the real-world deliverability state of your programme
• making trade-offs between volume and trust

If you’re under pressure and using AI to cope: here’s what to do 

If you are in “get emails out” mode right now, I’m not going to tell you to stop using tools.

I’m going to tell you to add three guardrails so you don’t accidentally torch your list:

1. Add an “intent check” before you generate anything

• Who is this for?
• What do they need right now?
• What’s the job of this email?

If you can’t answer those, AI will happily generate nonsense at scale.

2. Build a minimum viable exclusions rule

• Suppress new subscribers from generic blasts for 7–14 days
• Suppress active sales-cycle leads from broad “nurture”
• Suppress open support tickets from promos
• Throttle disengaged cohorts

This is how you stop AI output colliding with real human experience.

3. Use AI to reduce volume, not increase it

 The best use of AI isn’t “help me send more”.

It’s:

• What should we stop sending?
• Where is fatigue showing up?
• Which segments are cooling off?
• What journeys are leaking attention?
• What messages are causing deletes/unsubs?

The AI capability that actually excites me: intent prediction and journey mapping

If there’s one place I want AI to grow up fast in email marketing, it’s here:

Advanced intent prediction

Not “this person opened three times.” Real intent.

AI can help detect patterns like:

• Sudden spike in high-intent site behaviour (pricing views, comparison pages)
• Repeated consumption of a single topic category
• Signals of churn risk (drop in engagement + product usage decline)
• Movement from passive reader → active evaluator

When AI is applied here, it can do what humans struggle to do at scale: connect disparate signals into a probable state.

Better journey mapping (what should happen next)

This is the dream: less “everyone gets the same nurture” and more “the next message reflects what’s most likely true for this person right now.”

That’s where automated decisioning becomes powerful:

• What to send next
• What not to send
• When to pause
• When to escalate
• When to switch tone or cadence
• When to suppress because it’s the wrong moment

This is where your point matters most:

AI should help you stop doing what isn’t working, not help you do more.

Because the biggest hidden lever in email is exclusion and restraint.

The inbox rewards relevance and timing, not volume.

Automated decisioning: what it looks like in the real world

• A subscriber just entered a welcome flow → suppress promos for 14 days.
• A customer has an open support ticket → suppress sales messaging.
• A prospect is in an active sales cycle → switch marketing to enablement content or pause.
• Outlook placement is unstable → throttle, segment, and remediate with low-risk value sends.
• A person is showing “attention debt” → reduce cadence, offer preference centre, or snooze.

Tools and platforms that sit closer to this orchestration layer include the ones you named (and similar categories): Hightouch, Braze, Iterable.

The point isn’t “buy these tools.” The point is understanding what good looks like: AI assisting decision-making, not replacing strategy.

AI and CRM data: enrichment is helpful, but it’s not magic

In B2B especially, AI/data enrichment is one of the most obvious value-adds — because it can reduce manual effort and improve routing, personalisation, and relevance.

HubSpot’s AI positioning with Breeze is one example of where CRMs are pushing toward AI-enabled workflows and data usefulness.

But here’s the warning I want you to keep loud:

Enrichment does not equal insight.

Knowing someone’s job title doesn’t automatically tell you what they care about. And if your messaging is still one-size-fits-all, more data just becomes more clutter.

Strategic data is data that changes what you send, when you send it, or who you exclude. If it doesn’t change behaviour, it’s trivia.

AI and spam filters: yes, the filters are smarter (and they’ve been using ML for a long time)

A lot of marketers talk about AI as if it’s new. But inbox filtering has been driven by machine learning and pattern detection for years.

The important bit isn’t “spam filters use AI” as a headline. The important bit is what that means operationally:

• Filters learn from user behaviour (deletes, ignores, spam complaints, moving mail).
• Filters learn from patterns across senders (shared infrastructure signals, content patterns, volumes).
• Filters adapt to new tactics quickly (because they’re trained to spot abuse).

You can see Google’s general approach to automated risk analysis and abuse detection across its ecosystem in products like reCAPTCHA, which frames modern detection as risk analysis rather than simple rules.
 (Email filtering is not reCAPTCHA, but it’s the same direction of travel: pattern detection + behavioural signals + adaptive models.)

This is why “AI-generated spam” isn’t just an ethical issue — it’s a deliverability issue.

The more the internet fills with synthetic, repetitive messaging, the more filters will get aggressive at identifying patterns that look like automation at scale.

Are AI written emails more likely to go to spam?

It’s unlikely that filters are simply punishing “AI content” because it’s AI.

It’s far more likely that AI content correlates with lower engagement, because it’s generic, overproduced, and doesn’t sound like a relationship.

And lower engagement (plus negative signals) is what damages inbox placement over time.

So if you’re seeing “AI-ish emails” land worse, it’s probably not because the filter detected a specific phrase.

It’s because the programme is training filters that your mail is ignorable.

That’s why your stance is right:

People are trying to use AI to replace critical thinking and strategy, and it won’t work.

It can’t. Not in email. Not in a channel where trust and expectation setting matter.

In one recent study I did, we found AI-generated emails were more likely to land in spam by 20% and subscriber time to disengaged got 5x faster when all content was generated & sent from AI tools - coincidence? I don't think so. 

AI inside Gmail means people won’t experience email the way they used to

This is the part marketers need to sit with.

Google has been expanding Gemini capabilities across Workspace experiences.

That direction includes the kinds of features that change inbox behaviour: summarisation, prioritisation cues, “help me write” assistance, and surfaces that reduce how much a user has to read line-by-line.

So what happens when your reader’s inbox becomes a layer of AI interpretation?

• People may rely on summaries instead of reading.
• People may see “what matters” first, based on signals you don’t control.
• The inbox may compress or reframe your message before the human touches it.

This doesn’t mean email marketing is dead. It means the craft changes.

Clarity, structure, and relevance become even more important because you’re not only writing for a human brain — you’re writing for a machine layer that helps decide what the human sees.

So… is AI in email marketing “good enough”?

Sometimes.

AI is good enough when:

• Your data is clean and structured
• your journeys are intentional
• You know what impact email is meant to drive
• You have clear lifecycle logic and exclusions
• You’re measuring outcomes beyond opens/clicks

• You’re using it to ship more volume
• You don’t know why people are on your list
• Your segments are messy
• You haven’t validated the deliverability reality
• You’re trying to shortcut strategy

Or in other words:

If you don’t have an optimal strategy now, poor data, and you don’t understand modern email… don’t try to “AI optimise” it. Nail the foundations first.

What I actually recommend and what I’d avoid

Use AI for:

Be cautious with:

The real future of AI in email: less creation, more intelligence

The future isn’t “AI writes all emails.”

The future is:

• AI helps you understand intent faster
• AI helps you suppress the wrong sends
• AI helps you personalise based on signals that matter
• AI helps you map journeys that reflect real human behaviour
• AI helps you prove impact in business terms (not vanity metrics)
• AI helps you spend more time thinking and less time formatting

It’ll make email less noisy, less spammy, and more aligned with the conditions that make engagement possible.