Writing and personalization

​

What shapes Swan’s writing

When Swan drafts an email or LinkedIn message, it draws on three sources, in this order of context:

1. Org knowledge — your value proposition, competitive positioning, and ICP. This is the baseline context Swan uses for any GTM writing or reasoning, not just outreach. See org knowledge.
2. Sender-specific instructions — style and tone guidance tied to one connected sender (one person’s email or LinkedIn account), so messages from different senders can sound like different people.
3. Message examples — a store of real messages that worked well, retrieved for the specific situation at hand rather than applied blanket-wide.

Swan always pulls both the sender’s instructions and relevant examples before writing, and follows the instructions while matching the tone and structure of the examples.

​

Sender instructions

Sender instructions are free text tied to a specific sender, and they fully replace whatever was there before when updated — there’s no merging of old and new instructions. Before Swan changes them, it asks you for confirmation, since instructions apply to every future message sent from that sender, not just the current conversation. You can also set these directly rather than asking Swan to, from the Instructions action on a sender’s row at Outreach → Sequences → Senders; see senders and connected accounts.

​

Message examples

Examples are stored per organization and tagged as email or LinkedIn. They get added in two ways:

• Automatically, the first time a contact’s first message in a sequence is approved (with or without edits). Only the first message counts this way — approving a later follow-up step in the same sequence doesn’t add a new example, since a follow-up isn’t a fresh “here’s how I’d approach this kind of contact” example the way an opening message is.
• Manually, any time you tell Swan a message it wrote (or one you wrote) is a strong example worth remembering. This is worth doing deliberately — if you see a message land well, say so and ask Swan to save it, rather than assuming it happens on its own.

At write time, Swan retrieves the four examples most similar to the current situation — matched by the reasoning behind the outreach (why this contact is being reached out to), not just by contact type or industry name-matching. It doesn’t apply examples from unrelated situations just because they’re the newest in the store.

​

Why you should curate examples deliberately

The example store isn’t browsable — there’s no screen listing everything that’s been saved, and you can’t see how many examples exist for a given sender or situation. That means the only real quality control is what you put in: if you approve a mediocre message without comment, Swan may still save it as an example (for a first message) and it’ll shape future drafts. Actively saving your best messages, and asking Swan to update sender instructions when you’re unhappy with its style, is the main lever you have to steer its writing over time.

​

What this doesn’t include

Swan personalizes per contact using research and reasoning, not through structured A/B testing — there’s no built-in mechanism to split traffic between two message variants and compare results. If you want to compare approaches, the practical workaround is running two separate sequences and comparing their outcomes manually in analytics.

​

Common questions

If I edit a message before approving, does that count as feedback? Yes, for a contact’s first message in a sequence — your edited version is what gets saved as the example, not Swan’s original draft. Do sender instructions apply to LinkedIn and email both, or just one channel? They’re tied to the sender account, and a sender is either an email account or a LinkedIn account, so instructions apply to whichever channel that specific sender sends through. Can I ask Swan to write more like a specific past message? Yes — save that message as an example (manually, if it wasn’t already captured) and Swan will retrieve it for similar future situations. Will Swan use an example from a completely different context, like a different ICP? It shouldn’t — examples are matched by similarity to the current outreach reasoning, so an example saved for a very different type of contact is unlikely to surface for something unrelated. Does Swan tell me which examples it used for a given draft? Not currently as a visible citation — Swan runs a semantic search over your saved examples internally to inform the draft, and the step shows up only as a generic “Examples fetched” status, not a list of which examples were used.