AI email copy in Eloqua and Marketo

Both platforms write email now

Within two years generative AI went from a subject-line experiment to a button in the editor of both major enterprise marketing automation platforms. For an individual marketer that is a faster first draft. For the Marketing Ops team that owns the brand voice, the legal wording and the consistency of everything that leaves the instance, it is a new source of copy that nobody reviewed before it was typed into a template. Whether it helps or hurts depends on what controls it comes with so it is worth being precise about what each platform offers.

Generative AI in Eloqua

Eloqua’s generative AI features are part of Advanced Intelligence which Oracle made available to all customers in June 2025. They aren’t on by default: Oracle’s release notes say you “must request to have Gen AI enabled” through a service request and once it is on, a Use Generative AI Features action permission decides which security groups can use it.

Subject lines

Subject Line Generation sits in the Design Editor’s email settings as an AI Assist button above the subject line. The user enters an email topic and picks a language and a purpose. Oracle defines the purposes as promotional, transactional or neutral. Five suggestions come back and any of them can be copied into the subject line with one click. It replaced the older predictive Subject Line Optimization in the 25C release.

Content

Content Generation works on text components. In an email or landing page the user clicks the component, clicks AI Assist and chooses Generate for new text or Create Variation which is “allowed only if you have existing text in the content block” and rewrites it. The inputs are a topic, a purpose, a tone which Oracle illustrates with “a Professional tone will generate content that sounds objective, and courteous,” a formality that “ranges from Formal to Casual” and a language. The same feature appears in the Redwood dynamic and shared content editors which were in controlled availability when the feature shipped and in the SMS editor for customers with the SMS add-on.

After generation four buttons adjust the result: Lengthen and Shorten change it by about ten percent, Rewrite rephrases it and Bulletize turns it into bullet points. Use This replaces whatever was in the block. Supported languages for both subject lines and content are Arabic, English, French, German, Italian, Portuguese and Spanish, added in 25C.

What is left to the user

Oracle’s pages don’t name the model, describing it as configured for the task, and their data guidance is a note on the dynamic content editor: “be sure not to include personal or sensitive information.” The documentation describes no administrative setting for tone, length, vocabulary or standing instructions. Each of those is chosen by the user, per prompt, from the options Oracle offers.

The AI Assistant in the Marketo Email Designer

Marketo’s equivalent is the AI Assistant for Email Designer, part of the new Email Designer introduced in January 2025. Adobe describes it as using “Adobe’s generative AI technology and prompt library along with Firefly for image generation.”

Turning it on

It is “not enabled by default.” An administrator must first agree to Adobe’s core generative AI terms and the supplemental terms for the Email Designer, then contact the Adobe account team, then grant the Access AI Assistant permission to a role under Access Design Studio and assign that role to each user along with a workspace where relevant. Until all of that is done the AI buttons don’t appear.

What it generates

More than Eloqua’s. Subject lines and preheaders come as several samples with refinement and feedback. Content can be generated for a selected section. An entire email can be drafted from a selected template, though not when designing from scratch. Images can be generated from a description, using Firefly by default or Gemini 2.5 as an alternative, with controls for content type, color, lighting and composition.

The controls

Each prompt can be shaped by a free-text prompt with a prompt library to start from, a buying group such as practitioner or decision maker, a marketing journey stage such as discover, evaluate or commit, a communication strategy such as urgent or social proof, a language and tone, whether emojis are allowed, reference content to personalize against and uploaded brand assets. A community post from October 2025 reported the assistant as English-only at that point which may have changed since.

It is a richer set of dials than Eloqua’s and they are still dials on the marketer’s side of the screen. Brand assets can be uploaded as references which helps the model match a look. The documentation doesn’t describe a way for an administrator to fix the tone for everyone, cap the length of a draft or rule out a word.

Why per-prompt controls aren’t enough

Imagine the central team has done everything right. The brand voice is documented: confident, never pushy. Legal has a list of words that must never appear in customer email: guarantee, best-in-class, any claim about results. Every template zone has a length it was designed for. The house style is US English and never mentions pricing in an announcement.

Now a hundred marketers across regions have an AI Assist button. Each one picks a tone from a drop-down, or doesn’t. Each one gets a draft that may or may not say “guaranteed results,” may or may not run to twice the length the hero zone can hold, may or may not be in the right English. The careful ones paste the brand guide into every prompt. Most won’t because the whole point of the button was to save time. The brand voice took years to make consistent when one team wrote everything. Now it varies by who pressed the button and what mood the model was in.

Per-prompt controls help the individual marketer get a better draft. They don’t give Ops a way to make the same rules apply to every draft which is the thing Ops is accountable for. In both platforms the only enforcement left is the one that existed before AI: a human reviewer reading the copy, in Eloqua through campaign approvals where those apply, in Marketo through whoever holds the Approve Email permission. AI makes the copy faster to produce and no faster to review.

There is also a quieter cost. When drafting is nearly free, teams draft more and the result is more email to the same contacts. The post on AI content and contact fatigue goes into why the limit on what goes out matters more as the cost of writing it falls.

AI drafting with house rules in Campaign Launchpad

Campaign Launchpad added AI content generation on the principle that already governs the rest of the product: Ops decides once what is allowed and the software applies it every time. Marketers generate copy with AI inside Launchpad’s editor. Before any of them types a prompt, Ops has written the rules:

• Tone. A description of how the brand sounds. Every draft starts in that voice, whoever asked for it.
• Character limits. A cap on how long a draft can run so copy fits the zone it is written for rather than being trimmed by hand afterwards.
• Blocked words. The list legal would strike, from competitor names to claims about results. They stay out of every draft.
• System prompts. Standing instructions, such as “write in US English, don’t mention pricing or promise results,” added to every prompt automatically so nobody has to remember to paste them in.

The marketer’s prompt can be as short as “write a short intro for our June 12 webinar on supply chain planning.” Launchpad adds the rules, sends the request and returns a draft that is already in the brand’s voice, under the limit and free of the blocked words, into a zone of a locked template where it can’t disturb the layout. Approval, where the wizard requires it, reviews content that has already passed the house rules rather than content that might say anything.

Generative AI for email copy in Oracle Eloqua, Adobe Marketo Engage and Campaign Launchpad

Capability	Eloqua	Marketo	Campaign Launchpad
Where it runs	Design Editor for emails and landing pages, Redwood dynamic and shared content editors, SMS editor	Email Designer	The email and landing page editor inside each campaign wizard
What it writes	Subject lines and text content, new or as a variation of existing text	Subject lines, preheaders, section content, whole emails from a template, images	Copy for the editable zones of a locked template
Controls chosen per prompt	Topic, purpose, tone, formality, language	Prompt, buying group, journey stage, communication strategy, tone, language, emojis, reference content	The marketer’s prompt
Rules set once by Ops	None documented	Brand assets can be uploaded as references	Tone, character limits, blocked words, system prompts, applied to every prompt
Model and provider	Oracle-hosted and not named in the documentation	Adobe’s generative AI, Firefly or Gemini 2.5 for images	Anthropic or OpenAI, your choice
Whose agreement covers the data	Oracle’s	Adobe’s generative AI terms	Your own agreement with the model vendor
How it is turned on	Service request to Oracle, then the Use Generative AI Features permission	Accept Adobe’s terms, contact the account team, grant Access AI Assistant to a role	Ops adds an API key and writes the rules

Two more differences matter to the people who have to sign off on AI in the first place. Launchpad connects to Anthropic or OpenAI with your organization’s own API key so you choose the model and requests go out under the agreement you already have with that vendor. The data retention and privacy terms your security team reviewed are the ones that apply, rather than a marketing platform’s generative AI terms accepted in an admin screen. And because the same rules apply whichever platform the campaign is built in, a company running Eloqua and Marketo gets one brand voice from AI across both.

Oracle and Adobe will keep adding to their assistants and it is worth watching what they add. The question for Ops stays the same whichever model is writing: who sets the rules and do they apply to everyone.