The video library your site cannot see — 4,725 videos it does not know it has

Your archive holds 5,712 published articles with a video that actually plays on the page. Your site knows about 987 of them. The other 4,725 carry a working video and are filed as ordinary text articles. Nothing about them is broken — the video plays for a reader who opens the article. What they cannot do is be found as video. Every part of your site that asks "which of my articles are videos" — the category video player, the author pages, a future video index, a feed, a sitemap — asks that question of the marker, and for these 4,725 the marker is not set. The gap splits by where the video is hosted: 3,420 YouTube, 781 Facebook, 504 Spotlightr, 12 Vimeo. That accounts for 4,717 of them. The remaining 8 carry a video on none of those four, and part of this item is finding out what they are rather than rounding them away. The work is not the identification. That is already done, and it is not charged for here. The work is saving those lists so they can be re-checked, reconciling them against what your site currently believes, putting the result in front of you, and setting the marker only on what you approve — with a record of every change and a way to undo all of it.

The first thing this item produces is the survey data itself, written down and re-checkable. It is listed first because every other item in this document reads from it, and because a list that cannot be re-run is not evidence.

Your site currently marks 1,002 articles as video. 987 of them really are. Fifteen are not — they carry the marker and contain no video of any kind. Fifteen is a small number and this is a small item, priced accordingly. It is separate from the item above for one reason: everything downstream of the marker treats it as fact. A video index built on it will show fifteen entries that open onto an article with nothing to play, and it will keep doing so until someone looks at the fifteen. They are not assumed wrong. Each is opened and read, because a marker can be correct about an article whose video was removed later, and that is a different problem with a different answer. The finding is recorded per article either way.

These fifteen are the reason the 987 figure in this document is stated as 987 and not as 1,002. The two numbers differ by exactly this item.

This is the cheapest real fix available to the Facebook problem, and it needs no new engineering at all. 914 of your video articles are Facebook-hosted. A review of 664 of them found that for 315, a YouTube version of the very same segment was already live on your own channel — and for the most recent batch, the YouTube version had gone up the same day as the Facebook one. Nothing was ever missing. Two copies were published and the article was pointed at the wrong one. A Facebook-hosted article cannot play inside your category video player or your author pages, because that player reads YouTube and only YouTube. An article pointed at the YouTube twin plays there immediately, with no change to any template. So 315 of these are a swap, not a build. A further 102 are probable matches that need confirming individually — a title can match while the segment does not. And 250 of the 914 have never been checked for a twin at all, because the earlier review only covered the articles filed under your video categories. Those 250 are searched here. Nothing is deleted. The Facebook video stays in your records; the article stops leading with it.

This item stands on its own and also reduces whichever Facebook path you take below. Segments with no YouTube version anywhere are not invented here — they are reported as having none, which is what the two items after this one are about.

This is the cheaper of the two ways to answer the Facebook question, and it is the one already implied by your own not-yet-live video page, which defaults to exactly this. Today a Facebook-hosted article in a video listing falls back to a plain link. That is not a decision anyone made. It is what happens when the player is handed something it cannot read, and it looks like it — a bare link sitting where every neighbouring card has a picture and a play button. This item makes that fallback deliberate. A Facebook-hosted article gets a proper card with a picture and a clear outward label, consistent in both places your site lists video, so a reader understands they are being sent to Facebook rather than finding a card that failed. The rule is then written into the template, so the next person who touches it changes it on purpose rather than by accident. It buys no video playback. That is the item below, and it is the alternative to this one rather than an addition to it.

Taking the item below instead of this one replaces it; the two are alternatives and only one is needed. This item is the one included in the total printed on this document.

Your video player is built for widescreen, and almost all of your library is widescreen. The exceptions are few enough to name. 71 YouTube articles and 92 Facebook articles are not widescreen. Within those, 48 YouTube and around 47 Facebook are confirmed upright — phone-shot verticals and Shorts. Those are the ones that sit badly in a widescreen frame, and they are the concrete starting list here. For the upright YouTube segments the first place to look is your own channel. Grant Miller in particular has published widescreen versions of his own segments alongside the upright ones, so a proportion of this list is not a hunt but a match against material you already own. That check runs before anything is treated as unsolved, and the ones with no widescreen version anywhere are reported as such rather than quietly left out. Separately, around 195 articles store no dimensions at all — older shortcode-style embeds and articles where the video reference sits outside the article body. Their shape cannot be read from your records and has to be read from each video. That is the same kind of pass your own team ran once before on a smaller batch, so the method is known rather than invented.

The upright Facebook segments that the swap item above resolves are not counted again here. If that item is not taken, this one covers them at the same price.

At least one Facebook video sampled during the survey is dead on Facebook's own end — the article is fine, the embed is fine, and Facebook returns "Video Unavailable." Nothing on your site can fix that. What is worth knowing is how many there are. The equivalent exercise has already been run against part of your YouTube library and found nine dead videos, one of them the newest upload by your own editor, which meant it was the first card a reader saw. The same thing is presumably true somewhere in the Facebook set and nobody has looked. Facebook cannot be asked the way YouTube was. YouTube answers a simple public query about whether a video exists; Facebook closed that route years ago and now requires a developer account for it. So the check here reads the video's own page and detects the unavailable state directly, which is slower per video and has to be paced so Facebook does not start refusing. That probe is most of what this item builds, and it is why the price does not simply scale with the 914. This finds a content problem rather than fixing one. A dead video needs re-uploading, which is yours to do and is not priced here. The deliverable is knowing which ones and how many.

Run after the swap item above, this shrinks: an article already pointed at a live YouTube version is no longer exposed to a dead Facebook one. The price is held the same either way.

The survey behind this document is a snapshot. Articles published after it will drift the same way, and the natural moment to catch them is when a category is already being worked on. So from here on, each category consolidated under the taxonomy work already in progress also gets swept for articles carrying video that is not marked as video, and anything found is added to the same record this document's first item creates. This is not charged. Once the saved lists from the first item exist, the sweep is a few extra checks on work already being done and billed, and it would be dishonest to price it as though it were a separate visit. It is written down here rather than left as an understanding, so it is part of what that work delivers rather than a favour that depends on who is doing it. The one condition is the first item in this document. Without those saved lists there is nothing to check a category against, and the sweep becomes its own piece of work each time rather than a few extra checks.

Included with the category consolidation work already underway, for as long as that work continues. It is stated here so it is a commitment rather than an assumption.