The hard part is the channel switch. They wrote something on LinkedIn and your message arrives in their work inbox, so the context that made the comment answerable is missing. Write the email so it stands on its own without referring to where you found them.
Disclosure: Signl is ours. It finds these people, qualifies them and drafts the opening line; your own sequencer sends. Every example below is invented, with no reply rates attached, because we have not measured any across customers in a way we could reproduce. For the opener itself rather than the email, see how to message someone who commented.
Key facts
- An email is a different channel from the one the comment happened in, and that gap is the whole problem.
- Naming the post or the platform is what makes a well-sourced email feel like surveillance.
- The subject line has to work without the context. Most "personalised" subject lines do not.
- Three sentences and one question. Fifty to eighty words.
- This is still cold outreach, and the compliance questions are unchanged by the comment being public.
The context switch nobody designs for
On LinkedIn, a reply under someone's comment makes sense to them instantly. They are in the thread, they wrote the thing you are answering, and the connection between the two is visible on screen.
Move that same message to email and three things break at once. They are no longer in the conversation. They have no idea how you reached them. And the only way to restore the context is to tell them you were reading their comments, which is the one thing that reliably ends it.
So the email has to do something slightly odd: use everything you learned from the comment while referring to none of it. The topic, the stage, the vocabulary and the objection all come from what they wrote. The provenance stays out.
In practice this means writing the email as though you had worked out their situation from their role and company, and happened to be right. Which, in a sense, you did.
When email is the wrong choice
Email is one of three options and often the worst of them. Decide before you spend a lookup.
| What they wrote | Best channel | Why |
|---|---|---|
| A question you can answer | Public reply | They asked the room. Answering in the room is proportionate and visible to everyone else reading |
| A request for a recommendation | Public reply | Same, and the thread is where the shortlist forms |
| A comment on your own post | LinkedIn message | They engaged with you knowingly, so continuing there is natural |
| A stated problem, on someone else's post | No question to answer, no thread to join, and a message out of nowhere on LinkedIn is no less cold | |
| A complaint about an incumbent | Email, later | Useful, but wait. Replying within the hour on a competitor's thread is conspicuous |
| Praise, or a two-word comment | Neither | There is nothing to write about. Note the person and wait |
The fourth row is the one this article is about, and it is a large share of the useful engagement you will find.
Subject lines that survive the switch
A subject line cannot reference the comment, which rules out most of what gets called personalisation. What it can do is name the problem in the reader's own vocabulary.
| Their comment, roughly | Subject line | Why it works |
|---|---|---|
| Close takes nine days, reconciliation manual | reconciliation before close | Names the specific step, not the category |
| Tried an AI SDR, replies collapsed | the reply rate thing | Low-key, and it is their sentence |
| Three tools and none of them talk | the handoff between tools | Describes the gap, not a product |
| Nobody owns this internally | who owns this at yours? | A real question, answerable in one line |
| Rolled out X last quarter, mixed | quarter two, honestly? | Curious rather than commercial |
| We built our own and it rotted | the maintenance cliff | Names the thing they have not named yet |
Three rules behind those. Lower case reads as a person and capitals read as a campaign. Under five words survives a phone. And never put their company name in the subject, because every automated sequence does that and readers have learned what it means.
The three-sentence body
Sentence one: the problem, as they would describe it. Not your category. Their situation, specific enough that a reader without the problem would not recognise themselves.
Sentence two: something they did not know. A distinction, a cause, a number from your own work. This is the sentence that earns the reply, and the one most emails skip entirely.
Sentence three: a small question. Not a meeting. A question they can answer in one line from their phone, whose answer tells you whether to continue.
No link. No attachment. No calendar. Those belong in the second email, after they have replied, and including them in the first is what makes it a sequence.
Six worked examples
All invented.
1. The specific number
- Bad: "Hi Priya, I saw your comment on Acme's post about month-end close. We help finance teams automate reconciliation. Do you have 15 minutes this week?"
- Good: "Subject: reconciliation before close. Nine days to close with reconciliation still manual usually means two or three specific reconciliations are doing the damage rather than the whole process, most often bank and intercompany. The teams who fix it tend to do those two first and leave the rest. Which ones are yours?"
The bad version names the source in sentence one and asks for time in sentence three. The good version never mentions LinkedIn.
2. The failed attempt
- Bad: "Saw you tried an AI SDR and it didn't work out. Ours is different!"
- Good: "Subject: the reply rate thing. When an AI SDR drops reply rates it is usually volume rather than the writing: the same list gets worked harder, so the people who were never going to reply get three emails instead of one. Did yours change who you were contacting, or only how much?"
3. Dissatisfaction with an incumbent
- Bad: "I noticed you're unhappy with [competitor]. Here's why we're better."
- Good: "Subject: switching cost, honestly. The thing that keeps most teams on a tool they have outgrown is one specific dependency rather than the whole migration, usually historical data or an integration nobody wants to rebuild. Worth knowing which one it is before you look at anything else. Which is yours?"
4. The build-versus-buy moment
- Bad: "Building in-house usually costs more than buying. Can I send a comparison?"
- Good: "Subject: the maintenance cliff. Internal tools rarely fail on build, they fail when the person who built it changes team and nobody owns it. The question that predicts it is whether yours has a named owner or is technically everyone's. Which is it at the moment?"
5. The unowned problem
- Bad: "Is lead qualification a priority for your team this quarter?"
- Good: "Subject: who owns this at yours? Qualification tends to sit in the gap between marketing and sales, which is why it stays broken for years without anyone being wrong. If it is nobody's at the moment, that is usually the thing to fix before the tooling. Who would it land on?"
6. When not to send
- Their comment: "Great post, so true."
- Good: nothing. There is no material. If their role fits, write a role-based email and accept that it is cold, or wait for them to say something.
Volume, deliverability and the sequence question
This source produces fewer people than a list does, which changes how you send.
Send fewer, write each one. The entire advantage here is that the material is specific. Running it through a sequencer that swaps one variable throws that away, and the output reads worse than an honest template.
One follow-up, not five. They did not ask to hear from you. A single short follow-up a week later is proportionate; a five-step cadence turns a well-sourced email into spam with better research behind it.
Deliverability is a volume problem, and this motion is low volume. That is an advantage worth keeping. Sending 40 considered emails a week from a warmed domain is a different risk profile from 400.
Only email people who passed the fit check. Enriching everyone who commented and sorting afterwards means you paid to look up the job seekers, and it fills your sequencer with people who will never reply. The ordering argument is in qualify before you buy an email, and what a verified address should actually mean is in what verified email means.
Signl runs that ordering by default: classification decides whether an email lookup happens at all, so the plan's 500 verified work emails a month go to people who fit, with nothing charged per person. The drafted opener is checked against the comment, and a quoted fragment that does not appear verbatim is replaced with a plain line, because an invented quote is worse than no quote.
Limits
- A comment is not consent and not a lawful basis. The usual obligations apply: a lawful basis for processing, the information you owe people under Article 14, and an absolute right to object, which the UK ICO describes as having no exemptions (read 25 October 2026). Our position is in GDPR and B2B prospecting. This is not legal advice.
- Not everyone has a findable work email. Some people will only ever be reachable on the platform.
- The comment goes stale. A month later the situation may have changed and you will not know, which is an argument for writing the week you find them.
- Automated collection is contractually contested. LinkedIn's User Agreement section 8.2 prohibits scraping and bot access (read 25 October 2026), separately from the data protection question above.
