Benchmark report
Creative Benchmark · 2026

What the lowest-cost B2B advertisers put in their ads

14,000+ ads • 40 B2B advertisers • LinkedIn · Meta · Google · YouTube • July to November 2026

Anyone can look up a competitor's ads. What no ad library tells you is which of those ads are worth copying. So we compared the ads of B2B advertisers getting the lowest cost per result with the ads of advertisers in the bottom third.

About this benchmark

Who is in it
40 B2B advertisers: 20 leaders with among the lowest cost per result for the way they sell, and 20 from the bottom third, also based on cost per result. Both groups cover ABM, Demand Gen and PLG companies.
The data
Every ad they ran on LinkedIn, Meta, Google and YouTube from July 5 to November 8, 2026, more than 14,000 in all. We read the copy, looked at every image and watched the videos, then compared the two groups feature by feature.
Leaders
20 advertisers with among the lowest cost per result, each ranked against advertisers that sell the same way on the same platform.
Bottom third
20 advertisers that weren't in the top third on any platform. We took the biggest spenders, so each ran plenty of ads to study.

Every figure below reads the same way: "41% vs 25%" means 41% of a typical leader's ads on that platform have the feature, against 25% of a typical bottom-third advertiser's. Each advertiser counts equally, so one advertiser with hundreds of ads can't swing a result.

This report is a companion to The State of B2B Paid Advertising 2026, which found that how you sell predicts where you spend. Here we look at what goes into the ads, and the same pattern shows up: the best offer depends on your GTM motion.

80% vs 58%
of Google ads had run 60+ days. On Google, leaders leave their ads up.
14% vs 32%
of Meta ads had run 60+ days. On Meta, leaders replace them.
49 vs 34
different headlines per 100 LinkedIn ads. Leaders write more versions.

Finding 1

1. On Google, the leaders' ads stay up. On Meta, they come down.

How long an ad has been running is the one hint of performance that every ad library shows. It turns out to mean opposite things on the two biggest platforms.

On Google, 80% of the leaders' ads had run for more than 60 days, against 58% for the bottom third. Their typical ad had been running for 268 days, against 105. On Meta, it flips: only 14% of the leaders' ads ran past 60 days, against 32%, and their typical Meta ad came down after 24 days, against 45.

Both gaps hold in ABM, Demand Gen and PLG alike. One likely explanation is how people meet the ads. A search ad appears to someone new each time they search, so a good one can keep working for months. A feed ad reaches the same people again and again, so it wears out, and the leaders refresh theirs sooner.

InfographicShare of ads that ran 60+ days
Google
A long-running ad is a good sign
80%
58%
MotionLeadersBottom third
ABM83%66%
Demand Gen83%58%
PLG70%51%
Typical ad had run 268 days for leaders, 105 for the bottom third.
Meta
A long-running ad is a bad sign
14%
32%
MotionLeadersBottom third
ABM15%38%
Demand Gen15%41%
PLG7%15%
Typical ad came down after 24 days for leaders, 45 for the bottom third.
LeadersBottom third
Google search and display ads, and Meta ads including paused ones, that ran between July 5 and November 8, 2026. Each advertiser's own share, averaged.
In Google's Ads Transparency Center, an ad that has run for months is probably working. In Meta's Ad Library, it's more likely an ad nobody has replaced.

Finding 2

2. Leaders write more headlines

Most leaders don't put their money behind one ad. They write many versions. On LinkedIn they ran 49 different headlines per 100 ads, against 34 for the bottom third, and the gap holds in every motion: ABM 48 vs 39, Demand Gen 48 vs 32, PLG 55 vs 28.

At the extreme, one Demand Gen leader ran 162 different headlines across 240 LinkedIn ads, and two others ran 137 and 103. A few leaders do the opposite and put a third or more of their LinkedIn ads behind a single message, such as a brand line, a playbook or a product launch. Either way, it looks like a choice rather than a default.

On Google the same habit takes another form: one headline for each thing people search for. One leader writes a headline per product, another one per role its product covers, another one per tax situation. Two of them ran 96 different headlines in 112 ads and 85 in 106.

Example ads from advertisers outside the benchmark. They show the pattern; they aren't the leaders' own ads.

Vanta ad: all-type design on deep purple pitching a SOC 2 checklist
VantaOne plain design that can carry many headlines.
monday.com ad: a project list with the slipping project flagged in red
monday.comA familiar screen with one problem flagged.

Finding 3

3. Leaders ask the buyer a question

On LinkedIn, 41% of the leaders' ads ask the reader a question, against 25% for the bottom third. The gap is widest in Demand Gen, at 58% vs 24%, and it's there in ABM (33% vs 27%) but not in PLG.

The question often goes on the image itself: 15% of the leaders' LinkedIn images vs 3%. They style the image as a search history of the buyer's worries or a to-do list, ask whether AI assistants recommend a competitor instead of you, or ask HR teams, in their own language, whether they're lost among employee benefits. On YouTube, ABM leaders open with a question in 54% of their ads, against none in the bottom third.

The questions that work are ones the buyer is already asking themselves, not quiz questions about the product.

Example ads from advertisers outside the benchmark. They show the pattern; they aren't the leaders' own ads.

Semrush LinkedIn ad styled as a chatbot prompt typing Is my site AI-ready?
Semrush · LinkedInA chatbot prompt typing "Is my site AI-ready?"
Ahrefs Meta ad asking about competitors showing up in AI answers
Ahrefs · MetaWhy do competitors show up in AI answers when you don't?
Factorial Meta ad asking HR in Spanish whether payroll changes are complete
Factorial · MetaThe question HR asks before payroll closes, in the reader's language.

Finding 4

4. Leaders open on the buyer's problem, not a face

The leaders' Meta videos open on the buyer's problem in 34% of cases, against 12% for the bottom third, and on something unexpected that stops the scroll in 20%, against 10%. The bottom third more often opens on a person: someone talking to camera appears in 63% of its Meta videos, against 46% of the leaders'. The bottom third also shows its product more (78% vs 61%).

Still images follow suit. On Meta, the image is a photo of a person in 21% of the leaders' ads vs 38%, and a product screenshot in 14% vs 28%. On LinkedIn, the buyer's problem opens the image in 9% of the leaders' ads, against 4%.

YouTube is the exception. There the leaders show their brand within the first three seconds (62% vs 45%, in every motion), likely because a viewer can skip a YouTube ad after five seconds.

InfographicHow Meta videos open, and what's in them
Opens on the buyer's problem
34%
12%
Opens on something unexpected
20%
10%
Someone talking to camera
46%
63%
Shows the product
61%
78%
LeadersBottom third
Share of each advertiser's Meta videos, averaged. 144 unique leader videos and 118 from the bottom third.

Example ads from advertisers outside the benchmark. They show the pattern; they aren't the leaders' own ads.

HeyGen Meta ad: an invoice of video production costs against a monthly subscription
HeyGen · MetaThe buyer's bill as the ad: $41,600 of production costs against $24 a month.
Semrush LinkedIn ad: a notes-app checklist of what marketers actually do
Semrush · LinkedInA to-do list that shows the job is bigger than people think.
Salesforce LinkedIn video still: hands and an object cracking under pressure
Salesforce · LinkedIn videoOpens on the pain as a picture, with the words in motion.

Finding 5

5. The offer that works depends on how you sell

The State of B2B Paid Advertising found that GTM motion predicts channel mix better than industry does. The same is true of the offer inside the ad. Each motion's leaders have their own, and it barely shows up in the other two.

ABM
Something to read
A guide, report or survey in 31% of LinkedIn ads vs 13%. No gap in Demand Gen or PLG.
One leader ran a single playbook in 86 LinkedIn ads. Another laid its trend reports out like a magazine.
HubSpot Meta ad offering a free report with a Download button
HubSpot · Meta
Lattice LinkedIn ad: a survey report laid out as a newspaper front page
Lattice · LinkedIn
Demand Gen
A gift for taking a meeting
A gift card or similar in 20% of LinkedIn ads vs 3%, and in 15% of Meta ads vs none.
The best versions name who qualifies, such as firms of a certain size, which keeps out people who only want the gift.
Huntress LinkedIn ad offering a gift card as thanks for a demo
Huntress · LinkedIn
Abnormal LinkedIn ad offering free AirPods for booking a demo
Abnormal · LinkedIn
PLG
"Free", where people search
Free or a trial in 46% of Google ads vs 18%. The PLG bottom third pushes free on LinkedIn instead (24% of ads vs 8%).
On LinkedIn, PLG leaders play with familiar formats: parodies, instructions, notifications.
Synthesia Google search ad: start today, no sales call
Synthesia · Google
Spendesk Meta ad parodying a museum exhibit of receipts
Spendesk · Meta

Example ads from advertisers outside the benchmark. They show the pattern; they aren't the leaders' own ads.

Demand Gen leaders also send more LinkedIn message ads, the kind that land in the reader's inbox from a named person: 14% of their LinkedIn ads vs 5%. Theirs run about 78 words, against 93, and three leaders put the gift on the button itself.

Asking for a demo isn't the difference. Leaders and the bottom third ask outright for a demo or meeting about equally often: 15% vs 13% of LinkedIn ads, 18% vs 21% on Meta, 10% vs 12% on Google. What differs is what they offer in return.

Finding 6

6. Several popular moves are more common in the bottom third

Some of the most repeated advice in B2B advertising is to run more video, show your customers and post from your people's profiles. In this data, the bottom third does each of those more than the leaders do.

InfographicWhere the bottom third does more (LinkedIn)
Video, as a share of ads · lower for leaders in every motion
12%
21%
Names a customer on the image
11%
24%
Thought Leader ads (posted from a person's profile)
7%
15%
A product screenshot is the image
6%
13%
Company news or an announcement opens the image
4%
10%
LeadersBottom third
Share of each advertiser's LinkedIn ads, averaged. 3,030 leader ads and 3,695 from the bottom third.

That doesn't mean these formats fail. It means they aren't what separates the advertisers with the lowest costs from the rest, so they're poor things to copy just because a competitor runs them. When the leaders do use video on LinkedIn, it's mostly animated text and visuals with captions, and less often scripted live action (7% of their videos vs 20%).

No difference between the groups

  • Asking outright for a demo or meeting, on every platform.
  • A price or discount in Google ads (2% vs 2%).
  • A percentage or multiplier in the LinkedIn copy, such as "72%" or "3x" (12% vs 14%).
  • LinkedIn document ads (4% on both sides) and Meta's dynamic creative (32% vs 35%).
  • Personalizing LinkedIn message ads with the reader's job title or company (46% vs 49%), and a softer second button such as "Tell me more" (33% vs 37%).
  • Where the ads send people: the type of landing page, pages built for the campaign, and booking tools.

Finding 7

7. Thought Leader ads: leaders use fewer, and not team photos

Thought Leader ads are LinkedIn ads posted from a person's profile instead of the company page. They make up 7% of the leaders' LinkedIn ads, against 15% for the bottom third, lower in every motion. Only one leader posts 30% or more of its LinkedIn ads this way; four bottom-third advertisers do.

The bigger difference is what's in them. Among advertisers that run five or more, the bottom third illustrates them with photos of the team, the office or an offsite (45% of images, against 18% for the leaders), and with clips from podcasts and TV interviews (36% of videos, against 5%). The leaders post designed work, such as research posters or a billboard campaign, or film the person making their point straight to camera (20% of videos, against 2%).

Who posts, how they write and what they ask for are the same on both sides. A CEO or founder posts about half of them, the posts run about 165 words, and a demo request appears in 1%.

An example ad from an advertiser outside the benchmark. It shows the pattern; it isn't a leader's own ad.

Personio Thought Leader ad: an executive outdoors asking the cost of doing nothing
Personio · LinkedIn Thought Leader adAn executive making one point to camera.

Finding 8

8. On Google, leaders lean on search, awards and rivals' names

Search makes up 77% of the leaders' Google ads, against 66%, and YouTube a smaller share (14% vs 23%). Inside those search ads, ABM and PLG leaders mention awards, G2 ratings or "#1" more often (11% vs 6% overall), and name a competitor or say "alternative" more often (13% vs 9%). Their competitor ads ask about a rival's weak spot, such as fees adding up or an email tool you've outgrown, and send the click to a page about that competitor.

A few leaders also buy the cheapest clicks there are: searches for their own name typed wrong, or for the name they had before a rebrand.

Example ads from advertisers outside the benchmark. They show the pattern; they aren't the leaders' own ads.

Postman Google search ad: SmartBear alternative
Postman · GoogleNames the rival and gives its own numbers.
beehiiv Google search ad comparing itself with a rival
beehiiv · GoogleA head-to-head comparison in the headline.
Mailchimp YouTube ad still: a costumed character with the brand shown first
Mailchimp · YouTubeBrand first, then the joke, then the product.

Putting it to work

How to read a competitor's ad library

LinkedIn's Ad Library, Meta's Ad Library and Google's Ads Transparency Center are free, and most B2B marketers have scrolled through a competitor's ads at some point. The findings above suggest five rules for reading them.

Read how long an ad has run by platform.A Google ad that has run for months is probably working. A Meta ad that has run for months is more likely one nobody has replaced.
Look at the range, not one ad.Leaders run many versions of the same idea. Count the headlines and note which ideas keep coming back in new versions.
Copy the offer that fits how you sell.A report for ABM, a gift for a meeting in Demand Gen, "free" in search for PLG. A competitor that sells differently is the wrong model, even if it's your closest rival.
Don't copy something just because everyone does it.Video, customer logos and Thought Leader ads were all more common in the bottom third, and both groups asked for a demo about equally often.
Remember what the library can't show.No ad library shows spend, who saw the ad or whether it worked. Treat anything you borrow as a test, and check the result.

6 steps from ad library to campaign plan

Step 1
Pull their ads
LinkedIn, Meta and Google libraries. Note who they target, what they promise and what they ask for. That's your message map.
Step 2
Find the gap
Which buyer problem do their ads underplay? Check it against your sales calls, and write it as one hypothesis.
Step 3
Qualify the accounts
Find companies that likely use them, then cut the list to your ICP. Using a competitor doesn't mean they're unhappy with it.
Step 4
Check the audience
Build it, see who's actually in it (title, size, region, and who isn't your buyer), then write the ads.
Step 5
Launch one test
One change, one control, a fixed budget. Change the message; keep the audience, offer and budget.
Step 6
Review after two weeks
Ask what's working and what to change: keep, change or stop. Two weeks is a first read, not a verdict.

What you learn sends you back to step 2 (a new angle) or step 4 (a new audience). In Primer, steps 1 to 4 can happen in one conversation: ask for a company's current LinkedIn, Meta and Google ads, find the companies that likely use it, and build and check the audience. Step 6 is asking what to change.


Lookup

Every gap we measured

Every feature where the leaders and the bottom third differ by at least five points, in at least two of the three motions (or by a wide margin in one). Choose a motion to compare only advertisers that sell the same way.

LeadersBottom third

Within one motion, each figure averages only a few advertisers a side, so read single rows with care. A dash means too few advertisers to compare.


Methodology and important caveats

  • The benchmark. We looked across hundreds of B2B advertisers. Leaders were ranked within the same GTM motion and platform, among advertisers that spent $5,000 or more there, on results from August 5 to November 2, 2026. A result is a LinkedIn lead form submission, a Google conversion adjusted for the share of converters that fit the advertiser's ideal customer profile where we could measure it, or a Meta conversion credited to a click on an ad set that counts only lead events. Each leader needed 20 or more results (100 or more for PLG, where sign-ups are cheap). We removed advertisers whose costs looked too low for what they counted, whose leads matched none of their own ICP, or who sell to consumers.
  • The groups. We classified the leaders and the bottom-third advertisers into ABM, Demand Gen and PLG groups. The bottom third are the biggest spenders in each group with no platform in the top third. GTM motion comes from each company's website, set by hand where the website was ambiguous.
  • The ads. Every ad that ran from July 5 to November 8, 2026, from an ad-library data service and the platforms' own public libraries: 3,030 leader and 3,695 bottom-third LinkedIn ads, 390 and 756 Meta ads (paused ones included), and 3,246 and 3,582 Google ads, with YouTube videos watched for their first 90 seconds. For some of the largest advertisers we have only the most recent part of the window.
  • How ads were read. Copy was matched against keyword lists, and the matches spot-checked by hand. A model looked at every image and video against a fixed checklist (what it shows, how it opens, what it offers, whether a person or the product appears), and we checked its answers against the frames. The bottom third's images and videos were sampled, up to 40 images and 15 videos per advertiser and platform.
  • Each advertiser counts equally. Figures are each advertiser's own share, averaged. An advertiser counts on a platform only if it ran 10 or more ads there.
  • What it can't say. We can't see results for individual ads, so nothing here says which ad worked. With 20 advertisers a side, and fewer within each motion, treat each gap as an idea to test. Gaps that hold in all three motions (how long ads run, headline variety, LinkedIn video) are the strongest.
  • The example ads. To keep the benchmark advertisers anonymous, every ad shown comes from an advertiser outside the benchmark, chosen because it makes the same point as an ad the leaders ran. The figures come only from the leaders and the bottom third.

An ad library shows you what a competitor says. This benchmark is a way to judge which of those things are worth copying, and a reminder that the only way to know is to test it on your own buyers.

PRIMER · Creative Benchmark 2026
Ads from July 5 to November 8, 2026 · 40 B2B advertisers
Primer brand logo glyph in white

Stop wasting ad spend. Start proving impact.

Create a free account and start testing.