The Dead Embed Report: A Third of Our Founding Era Is Already Gone

We have been posting music videos since 2009. Somewhere along the way the archive got big enough to be a question instead of a blog: if a video blog runs long enough, how much of it is still there?

So we checked. Every embed, every post, all the way back. The answer is more interesting than “some links break,” because the breakage isn’t random. The internet has a memory, and it is forgetting on a schedule.

The stat, if you’re here to steal it: of the 2,135 YouTube music videos we’ve embedded since 2009, 1.4% are now dead, but that average lies. Link rot is savagely uneven by age: 33% of our 2009 posts and 41% of our 2010 posts now have a dead video. 2011–2016 lost 32% of its original videos too; it only looks clean because we’d already swapped in re-uploads. For 2026 it’s 0.7%. The oldest, biggest, most-licensed stuff rots first.

The headline number is a lie of averages

We scanned 2,585 posts holding 2,135 unique videos. 30 of those videos are dead: pulled, privatized, terminated, or vanished into the gray “this video is no longer available” box. That kills 28 of the 1,992 posts that ever carried an embed: a 1.4% death rate.

Sounds fine. Reassuring, even. The internet remembers 98.6% of what we showed it.

Except that average is hiding the actual story, the way an average always does. Split it by year and the floor drops out:

About that “already patched” column. The first version of this report put 2011 through 2016 at a clean 0%. That was wrong. A 2011 post on our site was playing a video uploaded in 2017, which is a neat trick. Before we ever ran the audit, a repair script had swapped live re-uploads into 22 of those 68 posts, and every original it replaced is still dead. 2011 didn’t survive. We just painted over it.

Read that again. The stuff we posted this year is basically immortal: 14 posts out of nearly nineteen hundred lost their video, and half of those are already patched. The stuff we posted at the founding, when this was a blog about an EDM scene that hadn’t blown up yet, is a third to over forty percent gone. Not degraded. Gone.

That gap is the report. Link rot doesn’t tick along at a steady 1.4%. It compounds with age, and it has a taste.

What dies first

Look at which posts went dark and a pattern shows up immediately. The 2009–2010 graveyard is wall-to-wall with the biggest names of the moment: Lady Gaga (four separate posts, all dead), Swedish House Mafia, deadmau5, Pryda, Afrojack, Tiësto remixes, Sebastien Leger. The exact instant dance music went mainstream is the most rotted stretch in our entire history.

None of that is chance. The bigger and more commercial the moment, the more hands eventually reach for the takedown button: labels consolidating rights, official channels replacing fan uploads, artists scrubbing the messy early footage, management cleaning house. The amateur upload of a 2010 Swedish House Mafia Ibiza clip was always going to lose a fight with a record label. And 2011 didn’t get a pass. The videos we’d already had to swap out of 2011 and 2012 posts included Afrojack, Chris Lake, Iggy Azalea, deadmau5, Skrillex, björk. Same pattern, a year or two later.

The internet doesn’t preserve what mattered most. It preserves what mattered least. Cultural importance is a risk factor.

The one that hurts

For the full bleak-comedy effect: one of the dead posts from 2010 is titled “Coldplay Sucks, But Lady Gaga doesn’t.” Our own thesis statement, the entire reason this site has the name it has, set to a Lady Gaga video, and the video is dead. The argument outlived its evidence. There is no neater illustration of the whole problem: the take is permanent, the proof rotted off.

Why this isn’t just our problem

Every “best of 2010” listicle, every nostalgic Reddit thread, every Wikipedia citation pointing at a YouTube embed from that era is sitting on the same fault line we are. Most of them just haven’t run the check. The web treats a 15-year-old link and a 15-minute-old link as identically trustworthy: same blue underline, same confident embed, right up until you click and find the hole. We only know our 2009–2010 numbers because we own the archive and went looking. Multiply that across a decade of music blogs, many of which are themselves already dead, and the “permanent record” of an entire genre’s birth is quietly turning into Swiss cheese.

What we’re doing about it

We’re not going to pretend we can fix YouTube. But we can stop pretending nothing’s wrong. A script (dead-embed-check.js) now runs against the archive every month, flags newly-dead embeds, and patches the post: swap in a surviving upload, or pull it instead of serving you a broken black box. That patching is exactly what made 2011 look clean, so this report now counts patched posts as losses, not survivors. (The 2009–2010 Flash-era embeds sit outside the checker’s reach, which is why they’re still dead.) An archive that watches itself rot and tells the truth about it is the most we can promise. It’s more than most of the internet does.

Methodology (for the skeptics and the press)

Every published post was parsed for YouTube embed IDs, from 2009 Flash <object> embeds to modern iframes; each unique ID was checked against YouTube’s oEmbed endpoint, which returns a clean signal for removed/private/terminated videos, and every failure was confirmed against the video’s watch page. “Dead” = the video no longer resolves to playable content. Videos with embedding switched off count as alive, and so do four old videos that are blocked by a copyright claim but not deleted. “Already patched” = posts where a repair pass swapped a dead video for a live re-upload; each replaced original was re-checked and is still dead. Counts are posts and unique videos, not raw embeds (a few tracks were posted twice). Years are original post dates. Snapshot date: 2026-10-13, n = 2,135 unique videos across 2,585 posts. (Our first run, on 2026-06-08, counted 1,357 videos across 1,690 posts and missed the patched posts, which is where the bogus 0% for 2011–2016 came from.) The method is simple enough to run on your own archive, and we’re happy to share the raw per-post data if you ask. We’d genuinely like to know whether your founding era rotted as hard as ours did.

Got an archive worth checking, or writing about how the internet forgets? We’ll share the data and happily compare notes: find us on Bluesky.