Video SEO, taught from the crawler's point of view
A crawler does not press play. It reads what the page wrote down about the video, and most pages write down almost nothing. The five fields that matter, the embed trap nearly everyone falls into, and how to look at exactly what the engine sees before it does.
Most advice about video SEO is a list of tips: put the keyword in the title, write a long description, add tags. The tips are not wrong, but they are cargo cult until you understand the thing they are trying to influence. So this guide starts one level down, with what a search engine's crawler actually experiences when it arrives at a page with a video on it. Once you can picture that, every tip either makes sense or falls away.
What a crawler sees when it sees a video
A crawler is a program that fetches a page, reads the text and the markup, follows the links, and moves on. It does not press play. It does not watch. It cannot hear you. When it reaches a video, what it gets is whatever the page chose to write down about that video: possibly a title, possibly a thumbnail address, possibly a duration, and, if the page was built carelessly, nothing at all but a rectangle it cannot see into. The twelve minutes of expertise you recorded is, to the crawler, exactly as informative as a black square.
Everything in video SEO is therefore one question asked over and over: what has the page written down about this video, in a form the crawler can read?
A crawler does not press play. It reads what the page wrote down about the video.
The five things worth writing down
Search engines publish the fields they want for a video, in a machine-readable format that sits alongside the visible page. Five of them do nearly all the work. StreamAgent's watch page sends all five automatically for every findable video; the point of naming them is so you know what to look for on any platform.
- Title and description. The plain-language answer to “what is this?” The title should say what the video answers, not what it is called internally. The description is a paragraph, not a keyword list.
- Thumbnail. The image the engine shows next to your result. Without a declared thumbnail the video is often not eligible for the video result at all.
- Duration and publish date. Two facts the engine displays and, more importantly, uses to decide whether the video is a real answer or a clip.
- Transcript. The words. This is the difference between a crawler knowing your video exists and a crawler knowing what it says. It is also the field most platforms quietly leave out.
- Key moments. Named points in the video with a clock time, the way a chapter list works. Engines can show these as jump links under the result, and they let an engine send someone to the part that answers them instead of to the beginning.
Give every video a page of its own
A video embedded halfway down a long page shares that page's title, description, and address with everything else on it. A search engine files one topic per page. So a video that should be findable on its own needs its own page: one address, one title, one set of machine-readable facts. In StreamAgent every published video gets a watch page on your subdomain, or your custom domain when one is connected, and that page is the one the engine indexes. The embed on your site is a window onto it, not a replacement.
Solve the embed problem
Here is a trap almost everyone falls into. You embed the video on your own site with an iframe. Your page ranks. But the crawler reading your page cannot see inside the iframe, so on your page the video is the black square again, and the structured data that would have made it a video result lives on a page the crawler was never sent to. The fix is a small companion piece in the embed code itself that writes the video's facts into the host page, so the crawler reading your page sees the video and its transcript right there. StreamAgent's embed includes this by default; some platforms make you find and add a separate snippet. Either way, check it exists.
Tell the engine your videos exist
A crawler discovers pages by following links, which is slow and unreliable for a page nothing links to yet. A video sitemap is a file on your domain that lists every findable video with its title, description, thumbnail, and length, so the engine learns about a new video the next time it reads the file instead of when it happens to stumble on the page. StreamAgent maintains this file on your subdomain and your custom domain. Your channel page does the linking job for the crawlers that follow links, listing every watch page and transcript page.
Publish the words
The transcript is where video SEO stops being about metadata and starts being about content. A readable transcript page, with each line carrying a timestamp that opens the video at that second, gives the engine the full text of what you said and a way to link to any part of it. It also carries the key moments, each with a clock time and a one-line description. Together those are how an engine answers “do they cover pricing?” with “yes, at 2:30” and a link that lands on 2:30. If your platform does not publish this page, no amount of title tuning gets you there.
Decide, video by video, what should be findable
Not every video wants to be found. A client walkthrough, an internal update, a password-protected proposal: these should play where you share them and nowhere else. The control should be per video, and it should be honest about what it does. In StreamAgent each video's Discoverability tab has one master switch, “Findable in search & AI.” Off means link-only: the video still plays anywhere you embed or share it, the page carries a do-not-index signal, and it drops out of the sitemap and the structured data. Under Advanced you can split search engines from AI assistants and set them separately. Password-protected and archived videos are never listed regardless. And Settings lets you choose whether new uploads start findable or link-only, with a bulk action for the existing library, so a privacy-minded workspace does not have to flip switches one at a time.
Look at what the engine will see, before it does
The tips-list version of video SEO has you edit fields and hope. The better habit is to look at exactly what the engine receives. The “See how it looks in Google” preview on the Discoverability tab builds a mock result from the real structured data the watch page sends: thumbnail, title, length, publish date, key-moment chips, with a view of the raw data underneath. When a video is not eligible, it says why in a sentence instead of faking a result: listing turned off, password-protected, not yet published, still processing. Read that preview for every video you care about. If it looks right there, it is right.
What comes after being found
Being findable is the first duty of a video, not the last. Someone who lands on a watch page from a search result is a stranger with a question, and a player that only plays lets them leave with the answer and nothing else. The rest of the work, asking who they are, scoring the answer, booking the call, happens inside the player after the play button. Search brought them. The video has to do the rest.
Write down what the crawler cannot see. Then check what it saw.
One platform, seven layers
This article covered one slice. The machine ships whole: record it, get it found, arm it, test it, rank the leads, train the ads, and ask your AI how it is going.
Common questions
- What does a search engine actually see when it crawls a video?
- Only what the page has written down about it in a form the crawler can read: typically a title, a description, a thumbnail address, a duration, a publish date, and, on well-built pages, the transcript and a list of key moments. The crawler does not play or watch the video. Without those written-down fields the video is a rectangle it cannot see into.
- Why does an embedded video not show up in Google video results?
- A crawler reading your page cannot see inside an iframe, so the video's structured data lives on a page the crawler was never sent to. The fix is a companion piece in the embed code that writes the video's facts, including the transcript, into the host page itself. StreamAgent's embed includes this by default; some platforms require a separate snippet.
- Does a video need its own page to rank?
- For a video that should be findable on its own, yes. A search engine files one topic per address, so a video embedded halfway down a long page shares that page's title and description with everything else on it. A dedicated watch page gives the video one address, one title, and one set of machine-readable facts, while the embed on your site acts as a window onto it.
- How do I keep some videos out of search results?
- Use a per-video control that is honest about what it does. In StreamAgent each video's Discoverability tab has a "Findable in search & AI" switch; off means link-only, so the video still plays anywhere it is embedded or shared but carries a do-not-index signal and drops out of the sitemap and structured data. Password-protected and archived videos are never listed regardless of the switch.