Writing emails - Help Center

Open any email in the editor and the agent picks it up as its subject. Everything below assumes you have a draft open — though you can also ask the agent to create the draft and write it in one go, from anywhere in the app.

Writing a first draft

Give it a brief and it writes a complete email — headline, developed sections, a call to action — in your brand's voice.

"Write a monthly newsletter covering our new pricing page, the API launch, and the team offsite."

It won't hand you three sentences and a button. A full-email request produces a full email, and the agent checks its own work for length and structure before applying it. If you want something short, say so — "just a two-line reminder" — and it will respect that.

Asking for a new email works from anywhere:

"Create a new draft and write a welcome email for people who just joined."

You can also start from one of the designed templates in the gallery, but only if you ask for one by name or explicitly say "start from a template" — templates bring their own theme, so the agent never swaps your design out from under you.

Editing what's already there

The agent edits specific blocks, not the whole email, so the rest of your layout is untouched and your teammates' cursors stay where they are.

"Make the second paragraph shorter."

"Change the button to say 'Claim your spot'."

"Add a P.S. at the bottom about the referral bonus."

"Delete the testimonial section."

You can undo the agent's last edit. Its most recent body edit to the open email carries a Revert this edit button on the chat message, which puts the email back exactly as it was — and then turns into Re-apply edit if you change your mind. Agent edits arrive from the server rather than from your keyboard, so this button, not Cmd+Z, is the way to undo them. The button only appears on the latest edit, so revert before you ask for the next change if you're unsure about one.

Design, layout and theme

Ask for structure and you get real editor blocks, fully editable afterwards:

"Put these three features in columns."

"Wrap the shipping details in a highlighted box."

"Add a header section at the top and a sign-off at the bottom."

The agent can also insert preset blocks and your own saved components by name — heroes, CTA sections, footers — and customize their copy as it drops them in.

For look and feel, it edits your theme, not the individual blocks:

"Make this dark mode."

"Modernize the design — cleaner fonts, more breathing room."

"More space at the top of the email."

That's deliberate. Colors set on the theme follow your branding everywhere, respect your reader's dark mode, and survive a theme switch. Colors baked onto a single block don't. If you do want a specific color in one spot, just name it and the agent will use it.

You can combine the two in one request — "expand this section and make the whole thing feel more premium" — and it will do the copy and the restyle together.

Header and footer

Ask explicitly and the agent will set your shared header or footer:

"Add our mailing address and an unsubscribe link to my footer."

Two things hold regardless of what it writes: a working unsubscribe link is always present, and the Sent with Audienceful badge stays exactly as your theme has it — the agent carries it over rather than deciding for itself (hiding the badge is a paid-plan theme setting). Contact details are only used if they came from you or your brand context — the agent won't invent an address.

Personalization

Type-level personalization works the same way it does by hand, just described in words:

"Greet everyone by first name, with 'there' as the fallback."

That inserts a real merge tag, resolved per recipient at send time. For event-triggered emails, it can also insert event properties — an order number, a coupon code, a link — including inside a button's URL, which is the only way to make an event value clickable.

Writing from the web and your own site

This is the part that changes what a newsletter costs you. Most emails aren't written from nothing — they're written from something: a blog post you just published, a changelog, a competitor's announcement, this week's news in your industry. The agent can go and get that material itself, so you don't have to summarize it by hand first.

There are two ways in, and they work together.

Paste a link

Give the agent a URL and it reads the actual page before it writes a word.

"Turn https://ourblog.com/2026-recap into an email for our list."

"Here's our new pricing page — https://oursite.com/pricing — write an announcement about the change."

"Write a product update from our changelog: https://oursite.com/changelog"

This is the fastest path from published on your site to in your subscribers' inboxes. Publish the post, paste the link, and the agent writes the email version of it — in your brand's voice, at email length, with a call to action back to the original.

You can paste up to three links in one request, which is what makes a link roundup practical:

"Write our monthly newsletter covering these three posts: [link] [link] [link] — a short section on each with a link through."

A few things worth knowing:

It reads the live page, so it's working from what's actually published right now, not from what a model remembers.
It only reads links you pasted. It won't invent a URL, follow a link it found on the page, or go wandering.
Ask for the page's images and it brings them across too — useful for a blog post with a good hero image.
A link into Audienceful itself (an /editor/ URL) isn't treated as a web page — that's one of your own drafts, and the agent uses it as a style reference instead.

Ask it to search the web

When there's no link to paste — because the thing you want to write about is out there somewhere rather than at a URL you have — ask the agent to search.

"Search the web for what's happened in AI this week and write a roundup newsletter."

"Research what our competitors launched this quarter and draft an email about how we compare."

"Find the latest email deliverability changes at Gmail and write a short explainer for our list."

"What's new in sustainable packaging this month? Turn it into our customer newsletter."

It runs a real search, opens the top results and reads them, then writes from what it found. Snippets alone make for thin, vague copy; reading the pages is what produces an email with actual substance in it.

Under the reply you'll see a "Searched the web for…" row with a chip for each source it used. Click through to check anything before you send.

Two nice touches: the agent searches on its own when a request obviously needs current information — "a roundup of what happened in e-commerce this month" gets researched, not written from memory — and if its first search comes back empty it retries with a broader query rather than giving up.

Combining them

The two compose, along with everything else the agent can read. All of this is one request:

"Read our launch post at [link], search for what people are saying about the category this week, and write it up like my top performing emails."

The agent reads the page, runs the search, pulls your best-performing emails as a style reference, and produces one draft from all three.

What it does with what it reads

It writes original copy. Fetched pages are reference material, not source text — it won't reproduce a page verbatim, and it attributes what it draws on.
It won't take instructions from a page. Anything that looks like a command inside fetched content is treated as content, not as something to obey.
Nothing is retained. Pages it fetches are used for that one request.

Attach files

The paperclip takes images, PDFs and text/CSV files (up to 8 per message). Attach a PDF brief and ask it to write the email; attach a photo and ask it to put it in the email; attach a CSV and ask it to write from the numbers.

Writing in your own voice

Beyond your brand context, the agent can read your actual emails and mirror them:

"Write this like my top performing emails."

"Match the tone of our Black Friday email."

"Write it like this draft: [paste an editor link]"

It pulls up to three of your own emails — ranked by real engagement when you ask for "top performing" — and matches their voice, structure and length. It writes new copy every time; it never reuses theirs.

Tips

Be specific about scope. "Rewrite the intro" edits the intro. "Rewrite this email" rewrites the email.
One thing at a time works well, but compound asks are fine too — the agent will use a multi-step plan when a request genuinely has stages.
You can stop it. If a reply is heading the wrong way, hit stop and re-ask. A stopped reply can still use some of your AI credits for the work it had already done.
Check its work before you send. Use test sends exactly as you would for anything else.