LinkedIn Conversation Ads: How to Design the Message Flow

Most of the LinkedIn conversation ads that land in my inbox were not designed. They were written. Somebody lifted the copy off a landing page, pasted it into the first message layer, added three buttons that all point at the same demo request, and set the campaign live. Nothing in the format stopped them, so nobody stopped.

That is more or less the whole problem. A LinkedIn conversation ad isn't an ad with buttons stuck on the bottom. It's a branching message tree, and the tree is the creative. Where the branches go, how many you build, and what waits after each tap will settle the result long before anyone starts arguing about the headline.

I'm Louise Perrin, Executive Assistant at Moriah, a LinkedIn agency and a LinkedIn Certified Marketing Partner. When our team maps a client's conversation ad, the work looks less like writing an ad and more like scripting a conversation. What do we open with? What do we expect the reader to say back? And what do we say next, depending on which answer we get? Here's that process, in the order we actually run it.

What LinkedIn Conversation Ads Actually Are

LinkedIn conversation ads arrive in a member's inbox as a message from a person, not from a company page. The member opens it, reads it, and picks from a set of call-to-action buttons underneath. Each button can lead somewhere different: another message in the thread, a landing page, or a native lead form.

So the reader isn't drifting past your creative. They're making a choice, then another one, and the route they take through your tree tells you something about what they came for. Nothing else LinkedIn sells does that. For a wider view of where it sits, our overview of the types of LinkedIn ads covers the full format list and what each one is built to do.

Worth saying plainly, though: this format sits at the bottom of the funnel. It works when the audience already has some reason to recognize the sender. Sent cold to a list that has never heard of you, it reads exactly like what it is.

Start With the Last Step, Not the First

Before writing a single line of the opening message, decide what single action you want at the end of the tree. Not three things. One.

That end point is usually one of four:

• A booked call on a calendar
• A completed lead form
• A downloaded asset the reader will actually open
• A reply in the thread that a salesperson takes over

Every branch has to arrive somewhere. If you can't name where one ends, it shouldn't exist. Skip this and the format punishes you for it, because a tree with a dead end costs exactly the same per send as one that works.

Once you know the ending, work backwards. What question gets someone there? What message earns them answering it? Building in that order, ending first and opening last, produces a tighter flow than writing left to right.

Scripting the Opening Message

The first message has one job: make the second tap likely. It doesn't have to sell, explain the product, or list capabilities. It has to be worth the ten seconds it takes to read.

Four elements belong in it:

1. A first line that names the reader's situation, not your company. The preview line is what gets it opened. Start with "At [company] we help businesses transform" and it's closed already.
2. One sentence of context. Who you are, and why this message is landing in their inbox. One sentence, not a paragraph.
3. One offer. A benchmark, a short assessment, twenty minutes with someone who works in their sector. Not a menu.
4. A question the buttons can answer. Most drafts are missing this piece. Close on a statement and the buttons underneath feel arbitrary. Close on a question and they read as the answer set.

LinkedIn will let you write up to 8,000 characters in a message layer. Treat that as the ceiling of a room you have no intention of filling. Our strongest openings run four to six short lines. Anyone who wanted the long version will tap the button that hands them the long version.

One more point about the opening. The message sends from a named individual's profile, and that person has to approve being the sender, so pick someone whose profile stands up to a click. Plenty of recipients check the sender before they even look at the buttons. That's one of several reasons paid inbox campaigns do better when the same executives are also visible through personal branding content.

How Many Branches Should the Tree Have?

LinkedIn allows five call-to-action buttons per message and fifty across the whole conversation. Almost nobody should use five. A "Not Interested" button is added automatically to the first message of a new conversation ad and counts against that limit, so the practical ceiling on your own first layer is four.

Build two or three. Here's why.

Every button you add splits the audience, and splits your data with it. Two branches against a modest audience and you'll know inside a week which one people prefer. Five branches gets you five thin streams, none of them reaching a number you can act on, and you end up guessing anyway.

Two or three also matches how people read an inbox message. A choice between two clear options is instant. A choice between five is a form, and a form in a message thread gets closed.

The second reason to keep the tree narrow is about volume. LinkedIn caps how often the same member can be sent a sponsored message from the same advertiser, at roughly one send every 45 days. You don't get to keep messaging the same list until something works. Two sends is close to your entire allowance with that audience in a quarter, which argues for a couple of well-considered branches over a spread of half-written ones.

Make the Branches Genuinely Different

The failure I run into most often is three buttons that are the same button wearing different labels. "Learn more," "See how it works," and "Get the details" are one branch typed three times.

Real branches split on something meaningful:

• Intent. Ready now versus researching for later.
• Role. The person who will use it versus the person who signs for it.
• Problem. Two different reasons someone in this audience would care.

Split on one of those and the tap itself turns into information. You learn what the person came for, and the next message can answer it.

Button labels are capped at 25 characters, which is less room than it sounds. Write them in the reader's words, not yours. "Show me the numbers" beats "Explore our solutions."

Where Each Button Should Lead

Every branch resolves to one of three destinations, and picking the wrong one is where a good opening message goes to die.

Another message layer is the default, and it's the right call more often than people think. The reader stays in the inbox, where they already were, and you get to answer the choice they just made. Anything you can handle in a few lines belongs here.

A native lead form is the second option. The reader taps and gets a form that pre-fills from their LinkedIn profile. Almost no friction, which is why LinkedIn Lead Gen Forms usually produce a lower cost per lead than sending someone to a landing page. Use it when what you want is the contact details and the follow-up happens elsewhere.

Then there's the off-platform page, the most expensive branch in the tree. It asks the reader to leave somewhere comfortable and load something they didn't ask to load. Save it for the cases where the destination genuinely can't live in a message: a booking calendar, a real product demo, a document worth the trip.

The rule we hold to is simple. Don't send anyone off LinkedIn on the first tap. That tap just told you something, and the reasonable next move is to acknowledge it, not to hand the reader to a page that has no idea what they chose.

Writing the Second Layer

The second layer is where conversation ads earn their premium. It's also the layer that gets the least attention.

It has to open by answering the choice the reader just made. Someone who tapped "We're evaluating options now" can't get the same general pitch a "Just researching" tap would get. That's the entire point of building a tree. If both branches run identical copy, you've built one message with extra steps and paid for the privilege.

Keep second layers shorter than the first. Two or three lines, then the resolving action. By now the reader has spent two taps on you and is owed whatever they came for.

One button here, two at the outside. The tree is meant to narrow. A second layer that opens back out into four fresh options reads as evasion.

Where People Drop Off, and What Each Drop Means

A conversation ad leaks at four points. Knowing which one is leaking tells you what to fix, which is more than most campaign reports manage.

They never open it. The problem is the sender and the preview line, not the message. A thin profile plus a first line about your company doesn't get opened. Change the sender before you touch the copy.

They open it and tap nothing. Either the offer isn't worth a tap, or the closing line wasn't a question. This is the most common leak and the easiest one to fix. Rewrite the last two lines before anything else.

They tap the first button and stop at the second layer. The second message didn't respond to the choice they made. Nine times out of ten, both branches are running the same copy. Split them properly.

They tap through and never complete the action. The destination broke the promise. A branch labeled "Show me pricing" that lands on a generic homepage loses almost everyone, and no amount of message polish repairs that.

Read them in that order. Rewriting second-layer copy when the sender profile is the real problem costs you a week you don't get back, and on a format where frequency limits stop you re-sending to the same list, that week is expensive.

A Conversation Ad Example, Scripted End to End

Here's the kind of tree we'd script for an industrial services company whose objective is booked calls with operations directors. Two branches, three layers, one ending.

Layer 1, from the Managing Director's profile:

Most operations directors I speak with are running last year's supplier list against this year's costs.

>

I'm John, Managing Director at [company]. We put together a short benchmark on what comparable operations are paying this year, broken out by region.

>

Useful to you, or is cost not the pressure right now?

Buttons: `Send the benchmark` and `Our issue is capacity`

Layer 2, benchmark branch:

Here it is. Page three has the regional breakdown, which is the part most people go to first.

>

If your numbers look off against it, I'm happy to walk through why.

Buttons: `Get the benchmark` (native lead form) and `Book 20 minutes` (calendar)

Layer 2, capacity branch:

That comes up more than cost does at the moment. We've handled it two ways with operations of your size, and one of them is considerably cheaper than the obvious answer.

>

Worth twenty minutes?

Buttons: `Book 20 minutes` (calendar)

Notice what the second branch doesn't do. It doesn't send the benchmark, because the person just told you cost isn't the pressure. It answers what they actually said. That's the difference between a conversation ad and a message ad with decorations, and it's available to anyone willing to write the second layer twice.

The Specifications, Once the Flow Is Built

Specs belong at the end, after the thinking. For reference, when you sit down to build the tree in Campaign Manager:

• Message text: up to 8,000 characters per layer, including spaces and punctuation, with up to 10 emojis per layer.
• CTA buttons: up to five per message and fifty across the full conversation. The "Not Interested" button is added automatically to the first message of a new conversation ad.
• Button text: up to 25 characters, including spaces and punctuation.
• Banner image: 300 x 250 pixels maximum, JPG or PNG, under 2 MB. It shows on desktop only, so the flow has to work without it.
• Sender: a named LinkedIn member who has approved the sender request.
• Pricing: charged per message delivered rather than per click, which is why a tree with dead branches costs the same as one without.
• Audience note: members in the European Economic Area and Switzerland only receive sponsored messaging if they have opted in to see it in their inbox. Plan audience sizes for those regions accordingly.

For how this format prices out against the rest of the platform, our breakdown of LinkedIn Ads costs has the per-format numbers.

Why the Best Flow in the World Still Underperforms Alone

Let me be blunt about the limit of everything above. A beautifully designed conversation ad, sent to an audience that has never encountered your company, will underperform, and the tree won't be the reason.

The format asks for something feed ads don't: a moment of recognition. The recipient reads a sender name and decides in roughly a second whether this is a message or an interruption. That decision gets made on ground you prepared, or failed to prepare, before the campaign ever ran.

Which is why LinkedIn Ads are one focus area out of three at Moriah, never a service you buy on its own. We run executive personal branding, targeted outreach, and ads in parallel, pointed at a single business objective. Content published from a personal profile performs roughly 5 to 10 times better than the same content from a company page, so when a client's leaders are visible and posting, the sender name on a conversation ad is already familiar. Targeted outreach has opened conversations inside the same accounts. The ad lands as the third touch instead of the first.

That is how the three work together, and the mix shifts with the objective: when a client's buyers are not yet posting, we lean harder on targeted outreach. Whatever the business objective, there is a version of the LinkedIn strategy that answers it. On its own, a conversation ad is a cold message with a budget behind it. Inside the engine, it's the point where a warm audience gets an easy way to say yes.

Frequently Asked Questions

What are LinkedIn conversation ads? LinkedIn conversation ads are a sponsored messaging format that delivers a branching message tree to a member's LinkedIn inbox from a named sender. The recipient chooses from call-to-action buttons, and each button can lead to a different follow-up message, a lead form, or a website.

How are conversation ads different from message ads? A message ad is a single message with one call to action. A conversation ad branches: the recipient picks a button and gets a different next message depending on what they chose. That branching is the whole reason to use the format, and it's wasted if every branch lands in the same place.

How many CTA buttons can a conversation ad have? LinkedIn allows up to five call-to-action buttons per message and up to fifty across the whole conversation. A "Not Interested" button is added automatically to the first message of a new conversation ad and counts toward the limit on that layer.

How many branches should I actually build? Two or three on the first layer. Any more and you split your audience into streams too thin to read, plus a long list of options in an inbox message reads like a form. Narrow the tree as it goes deeper rather than opening it out.

How long should the first message be? Four to six short lines. The technical limit is 8,000 characters per layer, but that's a ceiling, not a target. The first message only has to earn the next tap, and anyone who wants detail will tap the button that provides it.

What should each CTA button lead to? Another message layer for anything you can handle in a few lines, a native lead form when what you want is contact details, and an off-platform page only when the destination genuinely can't live inside the thread. Try not to send people off LinkedIn on the first tap.

Where do people drop off in a conversation ad? At four points: never opening it, opening without tapping, stopping at the second layer, and abandoning after they click through. Each one points to a different fix, respectively the sender, the offer, the branch copy, and the destination page.

How are LinkedIn conversation ads priced? They're charged per message delivered rather than per click, so every send costs the same whether the recipient engages or not. That pricing model is the practical argument for designing the tree carefully before you launch.

Can I use conversation ads to reach an audience in Europe? Members in the European Economic Area and Switzerland only receive sponsored messaging if they've opted in to see these ads in their inbox, so the reachable audience in those regions is smaller than your targeting estimate might suggest. Check audience sizes before you commit budget.

Do conversation ads work for cold audiences? Rarely on their own. The format leans on the recipient recognizing the sender, so it does best against audiences already warmed by executive content and targeted outreach. Sent cold with nothing behind it, it reads as an interruption and gets closed.