B2B LinkedIn Ad Examples: What Makes Them Work

The best B2B LinkedIn ads rarely win on production value. They win because the format matches the offer, the copy fits inside LinkedIn's own truncation limits, and the first few words of the headline make a specific claim instead of a vague one. Swipe-file posts show you the screenshot. This is the teardown: the recurring pattern behind each format, the exact specs LinkedIn documents for it, and how to judge whether an example is even worth copying for your situation.

What separates a good B2B LinkedIn ad from a generic one

Pull apart enough high-performing LinkedIn ads and the same four habits show up, regardless of format:

• One claim, not three. The ad states a single, specific outcome ("cut onboarding time to 9 days") instead of a list of features.
• The hook sits inside the truncation point. LinkedIn cuts off intro text and headlines at a specific character count before "see more." Ads written for that limit read like a complete thought; ads that ignore it get chopped mid-sentence.
• The format matches the commitment being asked. A cold, top-of-funnel audience gets a single image or a thought leader post. A warmer, retargeted audience can handle a gated document or a conversation ad.
• A human signal is present. A real face, a named person, or an actual product screenshot reads as more credible than stock photography or an abstract gradient, a pattern the evidence below supports with caveats.

Examples by format: the pattern and the LinkedIn specs behind it

Each LinkedIn ad format has its own technical ceiling and its own winning pattern. Here is what LinkedIn's own documentation recommends for each, and the creative pattern that tends to perform inside those limits.

Single image ads

The pattern: one image, one headline that states the outcome, and intro text that reads like something a person would actually post, not a brand committee's approved copy. According to LinkedIn's own single image ad specs, LinkedIn recommends a 70-character headline and 150-character introductory text, with a hard maximum well beyond that before the text gets hidden behind "see more." Images should be JPG, PNG, or GIF under 5 MB, at 1.91:1 (1200x628) for landscape, 1:1 (1200x1200) for square, or 4:5 for vertical mobile placements, which LinkedIn recommends specifically to avoid the borders added to 1:1.91 vertical images.

The mistake to avoid: writing 150 characters as a teaser instead of the actual pitch. If the hook isn't inside that window, most scrollers never see the rest.

Document ads

The pattern: page one states the problem or the number, not your logo. Because members preview the first pages before deciding to open the full document, the opening page has to earn the next swipe on its own. According to LinkedIn's document ad specifications, files can't exceed 100 MB or 300 pages, you can upload up to five documents per ad, and lead-generation document ads must be PDFs. Headline and intro text follow the same 70 and 150 character recommendations as single image ads, with 200 and 600 character hard limits.

Document ads work best as something genuinely useful on its own, a checklist, a benchmark, a framework, not a teaser for a longer gated asset.

Carousel ads

The pattern: each card earns the next swipe and the last card repeats the single call to action, instead of spreading five different asks across ten cards. Per LinkedIn's carousel ad specifications, carousel ads run 2 to 10 cards, with introductory text recommended at 150 characters (255 character maximum) and a card headline that truncates after two lines. Cards are built at a 1:1 ratio.

The common failure: using all 10 slots because you can, which dilutes the one idea the format is supposed to build toward.

Thought leader ads

The pattern: it looks like a personal LinkedIn post because it is one. Thought Leader Ads let you sponsor an existing organic post from an employee instead of building new creative, using the single-image or video ad format behind the scenes. In LinkedIn's own Thought Leader Ads two-pager, LinkedIn reports a 1.7x higher click-through rate and a 1.6x higher engagement rate versus other single-image ad campaigns, based on an early pilot group tested between February and April 2023. LinkedIn flags that figure as directional, since early-adopter bias can inflate pilot results.

LinkedIn's own best practice is to sponsor posts less than 30 days old and revisit the content strategy quarterly, since the format depends on the post still feeling current.

Video ads

The pattern: the hook lands in the first few seconds, captions run by default, and the video makes sense with the sound off. Per LinkedIn's video ad specifications, LinkedIn accepts MP4 files 75 KB to 500 MB, 3 seconds to 30 minutes long, at 4:5, 9:16, 16:9, or 1:1 ratios, with the same 70-character headline and 150-character intro text recommendations as single image ads. According to Factors.ai's guide to LinkedIn ads for B2B, most LinkedIn video is watched on mobile with the sound off, which is why captions matter more than production polish.

Conversation ads

The pattern: a short, branching message with one offer per path, saved for a genuinely high-value ask rather than a routine demo request. LinkedIn's conversation ad specifications cap each message in the flow at 8,000 characters, including the intro message and every follow-up, limit each message to 5 CTA buttons (50 per conversation), and automatically add a required "Not Interested" button to every new conversation ad. Factors.ai's guide notes that conversation ads tend to perform well the first time an audience sees one and lose effectiveness with repeat exposure, so treating the format as an occasional, segment-specific play rather than an always-on channel tends to hold up better.

Do ads with people's faces actually perform better on LinkedIn?

The honest answer: faces correlate with better engagement in the data available, but there's no large, controlled, platform-wide study proving it as a universal rule, and the gain shows up unevenly across metrics. That Random Agency's Random Labs analysis tested its own 2024 LinkedIn posts, splitting content with faces from content without. Posts with faces averaged 17.28 engagement versus 11.63 without, and 5.44 likes versus 3.70, both statistically significant at the 95% confidence level. Clicks were higher with faces (11.44 versus 7.75) but the difference wasn't statistically significant, while impressions were significantly higher with faces too.

Two caveats matter here. That test measured organic posts from one agency's own page, not paid ads, and the sample sizes were modest (at least 25 posts per group). Treat it as a directional signal, not a guarantee for your account.

Factors.ai's B2B guide makes a related, more informal claim: images with real people, rather than stock photos, tend to outperform abstract visuals or plain product screenshots in their experience running campaigns. That lines up with why Thought Leader Ads work: the "image" is a real employee's profile photo attached to their actual name, which is a stronger human signal than any stock photo could be.

How to turn one winning concept into 10+ testable variations

A single idea that works becomes a real test only when you change more than the background color. Vary these levers independently and you get variations fast:

1. Headline angle (3 to 4 versions): the same claim framed as a stat, a question, and a direct statement.
2. Proof point (2 to 3 versions): swap the specific number, customer type, or outcome you lead with.
3. CTA verb (2 to 3 versions): "Download" versus "See the framework" versus "Get the checklist."
4. Visual treatment (2 to 3 versions): a product screenshot, a real person, and a simple stat callout, holding the copy constant.
5. Format conversion (2 to 3 versions): the same message as a single image, a 3-card carousel, and a short document.

Three headline angles times two visual treatments times two formats is already 12 distinct ads from one underlying idea, without touching targeting. The bottleneck is almost never the concept, it's producing enough on-brand variations to actually fill that matrix. That's the gap a tool like producing ad creative without a designer is meant to close, and it's the same reason a brand kit built for generating AI ad variants saves time: once your fonts, logo, and templates are set up as reusable layers, multiplying one concept into a dozen on-brand versions stops being the bottleneck.

How to judge an ad example before you copy it

A screenshot in a swipe file never shows you the three things that actually determine whether it will work for you:

• Offer fit. A gated ebook aimed at a cold, prospecting audience behaves differently than the same offer retargeted at people who already visited your pricing page. If you can't tell which audience an example was aimed at, don't assume it will transfer.
• Funnel stage. A thought leader post building awareness and a conversation ad pushing a demo request are solving different problems. Copying a bottom-of-funnel pattern into a top-of-funnel slot (or the reverse) is the most common way a "great ad" underperforms when copied.
• Audience fit. An example built for VP-level, six-figure deals won't automatically translate to a self-serve, individual-contributor audience. Check whether the CTA matches the commitment level: a lead gen form fits a warmer ask, a website visit CTA fits a colder one.

If execution capacity, not ideas, is what's slowing your LinkedIn program down, that's the specific problem Tofu Ads Agent's LinkedIn Ads workflow is built around: it reads your CRM's closed deals to build the audience and funnel plan, then produces enough on-brand creative variations across single image, document, carousel, and video formats to actually run a test, before staging campaigns into your own ad account for a person to review and turn on.

Frequently asked questions

Sources

1. Single Image Ads · LinkedIn
2. Document ads advertising specifications · LinkedIn
3. Carousel Ads · LinkedIn
4. Thought Leader Ads two-pager · LinkedIn
5. Video Ads Specifications · LinkedIn
6. Conversation ads advertising specifications · LinkedIn
7. Random Labs: Do Faces Boost Social Media Post Performance? · That Random Agency
8. LinkedIn ads for B2B in 2026, with data from $150M in spend · Factors.ai