LinkedIn Document Ads: 6,577 Real Examples, With Observed Campaign Durations

One creative per template, drawn from 440 companies across the industries below. Captions are each ad's own intro text, cleaned of trailing calls-to-action and truncated to a caption length; none are AdPeekly's own copy. Each card also carries the observed duration: how long AdPeekly watched that ad stay live, not an estimate. These 40 are a sample of 6,577 distinct document creatives in the full library.

Of the 432 companies where AdPeekly holds 200 or more published ads, 331 (76.6%) have run at least one document ad. Among the 440 companies running any document ad at all, the median one puts 4.21% of its ads into the format, the 90th percentile is 22%, and 121 of those 440 go past 10%.

The 200-ad denominator is deliberate, for the same reason it is on every format page: adoption rises with how many ads AdPeekly holds for a company, so a figure computed across every tracked company would measure our own coverage as much as advertiser behaviour. Scope note: this describes the companies AdPeekly curates, not B2B advertisers as a whole.

Cybersecurity leads document ad volume, 1,135 creatives across 43 companies, well ahead of any other tracked vertical.

Across 2,877 document creatives with a reported start date (273 advertisers), the median is 19 days, against a 19-day dataset median and carousel's 22-day figure. AdPeekly has not run a within-company paired test on this comparison the way it has for carousel: document is a fraction of the site-wide dated population, so a close or distant match to the aggregate could be company-mix overlap as easily as a real format effect. This is an observed number, not a tested claim.

The distribution's spread: the 25th percentile is 7 observed days, the 75th is 50, and the 90th is 112.

How to read these durations. They cover only creatives where LinkedIn reported a campaign start date, 2,877 of 6,577, a 56.3% missing-start-date rate. For the rest, the only measurable span is how long AdPeekly has been watching, which describes our scraping schedule rather than the advertiser's decision, so those are excluded rather than mixed in. Full detail on the methodology page. For this same distribution across the whole dataset, broken down by industry and funding stage rather than just document, see how long B2B LinkedIn campaigns run.

LinkedIn publishes the permitted limits for file size and page count and is the authority on them. Document ads carry no headline field in AdPeekly's data, zero coverage on 6,577 representative creatives, so this section anchors on the field every document ad has instead: intro text (primary_text in the dataset), the copy LinkedIn shows above the document preview in the feed.

LinkedIn's own guidance recommends keeping intro text under roughly 150 characters to avoid the feed's "...see more" truncation, with a published 600-character technical maximum. Across 6,577 document creatives, the median length is 203 characters and the 90th percentile is 608, effectively at that 600-character ceiling. Only 35.7% (2,347) stay under the 150-character recommendation: most document ads in this dataset are written to run past the truncation point, not to avoid it.

Video ads and carousel ads get the same treatment: real examples, adoption, and observed campaign durations. Or start from the full LinkedIn ads hub for the format mix across every ad AdPeekly tracks.

What surprised me here was the intro text, not the format itself. Document ads carry zero headline coverage in this dataset, so primary_text is the only caption field advertisers actually fill in, and the 90th percentile length lands at 608 characters, a hair over LinkedIn's own 600-character technical ceiling for intro text. That is not noise: it reads as advertisers writing to the actual limit, not the 150-character recommendation LinkedIn publishes to avoid the feed's "...see more" truncation, which only about a third of document ads respect. Pulling the sample for the gallery captions was its own lesson in why a straight primary_text quote would not have worked: real ad copy this size carries embedded line breaks, emoji, dangling CTA fragments and, in a few cases, LinkedIn's own feed truncation baked into the scraped text. What is on the page is the cleaned version of that, and the rule for cleaning it is documented in the generator, not applied by hand.

Daniele Viero, founder of AdPeekly