Key takeaways
- The description box is four unrelated jobs sharing one text field — orient a viewer, give the machines something to classify, carry your legal disclosures, and route traffic — and most bad descriptions are a case of one job quietly eating the other three.
- Only the opening lines are read by anybody. Write the first sentence as if it is the entire description, because for the overwhelming majority of viewers it is.
- YouTube's own documentation is blunter about this field than the SEO industry is. It says tags play a minimal role in discovery, and that packing extra tags into the description is a policy violation, not a clever trick.
- The hashtag cap everyone quotes is out of date. YouTube's Help Centre now puts the cut-off at more than 60 hashtags, at which point every hashtag on the video is ignored — and only the first three ever appear above the title.
- Links in Shorts descriptions have been non-clickable since August 2023, and description links on long-form carry no SEO value to Google. They are a traffic tool, not a ranking one.
- If you set no translation, viewers in other languages may be shown a machine-translated version of your description that you never wrote and cannot preview in Studio.
Of all the fields on an upload page, the description box gets the strangest treatment. It is simultaneously the thing creators are told is critical for search and the thing they fill with the same forty lines of social handles they pasted in 2021. Both habits come from the same place: nobody has ever worked out what the box is actually for, so it gets treated as either a ranking lever or a dumping ground, and usually as both at once.
The honest answer is less dramatic and more useful. The description is four separate jobs sharing a single field, and the four jobs want different things. One of them has a reader. Three of them do not. The reason so many descriptions are bad is not that creators write them badly; it is that they write one job's version and let it crowd out the rest.
This is what each of the four jobs needs, what YouTube's own documentation says about the field (which is considerably less flattering than the advice industry built on top of it), where the hard limits actually sit in 2026, and a structure that does all four without turning into a wall of keywords.
The four jobs sharing one field
Separating them out is most of the work. Once you can see which job a line is doing, it becomes obvious which lines are earning their place.
| Job | Who it serves | What it needs |
|---|---|---|
| Orientation | A viewer who clicked and wants to know they are in the right place | One or two plain sentences, at the very top, in human language |
| Classification | YouTube search, suggested, Google, and the AI surfaces that summarise before anyone clicks | The real subject named in natural prose, plus a transcript-like layer of context |
| Disclosure | Regulators, and YouTube's own policies | Specific, unmissable statements about sponsorship and affiliate relationships |
| Routing | The small share of viewers who scroll looking for a link | A short, stable, scannable block below everything else |
The conflicts are immediate. Orientation wants the top two lines; routing also wants the top two lines, because that is where links get clicked. Classification wants length; orientation wants brevity. Disclosure wants prominence; nobody wants to lead with a sponsorship notice.
The resolution is an ordering, not a compromise. Orientation and classification share the top, because the top is the only part with a guaranteed audience. Disclosure sits immediately below, where it is prominent without being the opening line. Routing goes last, because the person who wants your link will scroll for it and the person who does not will never see it either way.
How much of it anyone actually sees
The field caps at 5,000 characters. Strictly it is 5,000 bytes — YouTube's Data API documents the field that way — which only becomes a practical difference once you write outside the Latin alphabet, where accented characters, emoji and non-Latin scripts each consume more than one byte and the ceiling arrives earlier than the character count suggests.
What matters far more is the visible window before the "...more" fold. Here the honest position is that YouTube does not publish a number. The figures in circulation — roughly 150 to 160 characters on desktop, closer to 100 on a phone — come from third parties measuring the interface, not from documentation, and they move whenever the watch page is redesigned. They also shift with your own text, because line breaks and hashtags consume the available lines faster than prose does.
Treating that as a budget is the wrong lesson. The right one is structural: there is a short window with a reader and a long tail without one, the boundary is not under your control, and anything load-bearing has to sit above it. In practice that means one sentence. Write the first sentence so that it would work as the whole description, and everything downstream becomes a bonus rather than a dependency.
The test that settles most arguments
Open your last upload on a phone, without expanding the description, and read what is visible. If what you see is three social handles and half a hashtag, the field is not doing any of its four jobs — it is doing the fourth one badly, at the expense of the other three.
What YouTube actually says the field is for
This is where most description advice quietly parts company with the platform. YouTube's Help Centre page on adding tags is unusually direct: tags "can be useful if the content of your video is commonly misspelled", and "otherwise, tags play a minimal role in your video's discovery". That is the platform's own assessment of the field the entire tag-research industry is built on, and it reframes the description by implication. If the dedicated keyword field is near-worthless, a description written as a second dedicated keyword field inherits the same problem.
What the description does contribute is relevance — the question of whether your video belongs in the candidate set for a query at all. YouTube has never published weights, and anyone who gives you a percentage is guessing. The defensible version is the shape rather than the size of the effect: metadata helps determine what your video is about, and viewer behaviour determines whether it deserves to rank for it. A description can get you considered. It cannot make a video that nobody finishes outrank one that they do.
That has become more interesting, not less, as the surfaces that read your text before anyone watches have multiplied. Summaries, conversational answers and AI-assembled result pages all work from text, and your description is some of the only text you write deliberately. The short version is that a description written for a machine to classify and a description written for a machine to summarise are closer to each other than either is to a keyword dump.
The line between keywords and a policy violation
The same tags page draws a boundary that a surprising number of channels cross without noticing. "Adding excessive tags to your video description is against our policies on spam, deceptive practices, and scams." Not discouraged. Not ineffective. Against policy, in the section of the rulebook that ends in removals and strikes.
The behaviour that triggers it is familiar: a readable description, then a horizontal rule, then sixty comma-separated phrases nobody was ever meant to read. It is a holdover from an era when it may have worked, and the only thing it reliably does now is place your channel on the wrong side of a spam policy for a benefit the platform says is minimal.
The related trap is misleading metadata rather than excessive metadata. Titles, descriptions, tags and thumbnails are all covered by the same expectation: that they represent the video. A description promising something the video does not deliver is a packaging problem that has become a policy problem, and it is judged on the gap between what you promised and what you delivered rather than on the words themselves.
Keywords in a description are fine. They are fine in the way they are fine in a well-written article: the subject is named because the piece is about the subject. The difference between optimisation and stuffing is whether a human reading the sentence aloud would notice anything strange.
Hashtags: the number everyone quotes is the old one
Hashtag advice is the most reliably out-of-date material in this corner of the internet. The widely repeated rule is that more than 15 hashtags causes YouTube to ignore all of them. That wording is from an older version of the Help Centre. The current page on finding videos using hashtags puts the threshold at 60: go past it, and every hashtag on that video or playlist is ignored.
Two things follow, and they point in opposite directions. First, the practical ceiling is far higher than almost anybody acts on. Second, it does not matter much, because the display rule is unchanged and brutal: only the first three hashtags from your description ever surface above the video title, and a hashtag placed in the title itself takes that slot instead. Hashtags four through sixty are a classification signal at best.
| Rule | What it means in practice |
|---|---|
| More than 60 hashtags ignores all of them | The real cap, per YouTube's current documentation. Far above anything worth using. |
| First three appear above the title | The only hashtags with a visible position. Choose them as display copy, not as keywords. |
| A hashtag in the title overrides the description's three | Spending title characters on a hashtag costs you both the characters and the description slot. |
| Irrelevant or misleading hashtags | Grounds for removal of the video or playlist, per the same page. |
| Over-tagging | May result in removal from your uploads or from search. |
Three to five, chosen because they are genuinely the categories your video belongs to, is a defensible default. The first three are effectively a second subtitle under your packaging, so they should read like one.
Timestamps do two different jobs
A list of timestamps in a description is doing something more specific than organising your video, and it is worth knowing which of two mechanisms you are feeding.
Inside YouTube, correctly formatted timestamps become chapters — the segmented progress bar, with labels, that viewers can jump between. The formatting requirements are strict enough that most failed attempts are formatting failures rather than eligibility ones, and we laid them out in full in the guide to chapters and key moments, so there is no sense repeating them here.
Outside YouTube, the same lines can feed Google's key moments: the timestamped entries that appear beneath a video in search results and jump a searcher directly into a segment. Google's own video structured data documentation describes a related arrangement for pages you host yourself, where a description is a recommended rather than required property and markup influences what Google shows instead of what it guesses. For a YouTube-hosted video, your timestamps are the closest thing you have to that control.
The point is that timestamps are the one part of a description that is unambiguously functional. They do not need keyword consideration. They need to be accurate, because a chapter label that lies about what is in the segment is a retention problem wearing a navigation costume.
Links: clickable, gated, and worth nothing to Google
Three constraints on the routing job, each of which catches people out.
Format. A URL becomes a link only if it carries the full protocol prefix. A bare domain stays as plain text, which is the explanation for most "my link is not clickable" confusion.
Shorts. Links in Shorts descriptions and Shorts comments stopped being clickable in August 2023. YouTube announced it as an anti-spam measure after Shorts' growth made clickable links an attractive vector for scams, and it applied to creators' own links as well as commenters'. YouTube's guidance on sharing links with your audiences still describes Shorts descriptions and comments as non-clickable, and the partial replacement was a more prominent set of links on the channel profile near the subscribe button. If your Shorts strategy ends in a description link, it does not end anywhere.
SEO value. There is none. Description links are nofollowed, which is to say they pass no ranking credit to the destination. They are worth having for traffic and worth nothing for authority, and any service selling YouTube description links as backlinks is selling you a nofollow link at a dofollow price.
Clickable external links on long-form also sit behind YouTube's advanced-features tier rather than being available to a brand-new channel by default. Access to that tier is granted on channel history or identity verification rather than automatically, which is one of several reasons it is worth clearing early.
The disclosure layer, and why it belongs high
If a video is sponsored, includes a paid product placement, or carries affiliate links, the description is one of the places that has to say so — alongside, not instead of, YouTube's own paid-promotion checkbox and whatever you say on camera. The rules here are specific enough that guessing is a bad idea, and we worked through what each mechanism actually requires in the guide to branded content disclosure.
What is worth saying in the context of writing the box is positional. A disclosure buried below forty lines of links is technically present and practically invisible, and "technically present" is not the standard being applied. Put it immediately after your opening sentences, before the timestamps, in plain language. It costs you two lines of a field whose lower two thirds nobody reads anyway.
Upload defaults, and the problem they create
Studio's upload defaults — Settings, then Upload defaults — pre-fill the description on every new upload. For anything you repeat, this is straightforwardly correct: the disclosure boilerplate, the link block, the channel line. Typing them fresh each time is how they end up inconsistent or missing.
The failure mode is subtler than forgetting to customise. A default description is, by construction, generic — and because it is already in the field when you start, it occupies the top of the box. The video-specific sentence then gets appended underneath, which puts the only part of the description with a guaranteed reader below the only part that is identical across your entire catalogue.
The fix is to build the default with the gap at the top. Leave the first lines empty, or hold a placeholder you have to delete, so that the structure forces the specific sentence into the position that matters. If you would rather start from a draft than a blank field, our free YouTube description generator produces that first-sentence-plus-structure shape, which is the part worth having before you get to the boilerplate.
Shorts descriptions are a different field doing a different job
The same 5,000 characters are available on a Short, and almost none of the reasoning transfers.
Links do not work, as above. The field is barely surfaced in a feed built for vertical swiping, where the visible fragment is shorter than on a watch page and the viewer's next action is a swipe rather than a scroll. What remains is the classification job: the text is still how the system works out what the Short is about, which matters because a Short is distributed to people who did not search for it and the first audience it is tested on has to be the right one.
So the useful Shorts description is short, names the subject in plain words, and carries three or four hashtags. One sentence and a few tags. The elaborate structure that earns its place on a twelve-minute video is wasted effort on a thirty-second one.
The description you did not write
One mechanism most creators never think about: your description exists in languages you do not speak, whether or not you put it there.
YouTube's Help Centre documents how to translate your own video titles and descriptions — per language, from the Subtitles or Languages tab in Studio, with an option to set a thumbnail per language alongside the text. Creator documentation consistently reports the corollary: where you have supplied nothing for a viewer's language, YouTube may display a machine translation of your title and description that you did not write, cannot preview in Studio, and have no control over the wording of.
For a lot of channels that is fine, and for some it is actively useful. It stops being fine the moment your description does work that a machine translation will mangle — a specific product claim, a disclosure, a joke in the opening sentence that carries the whole promise. If an international share of your audience is meaningful, the translations you supply yourself are the ones that say what you meant, and the same logic extends to the frame: there is a per-language thumbnail slot sitting in the same interface, which we covered in auto-dubbing and localised thumbnails.
A structure that does all four jobs
Put together, the ordering falls out of the constraints rather than from taste.
| Position | Contents | Why it is here |
|---|---|---|
| Lines 1–2 | One or two sentences naming what the video is and what the viewer gets, in the words a person would use | The only guaranteed-visible region. Serves orientation and classification simultaneously. |
| Line 3 | Disclosure, if there is one. Sponsorship, paid placement, affiliate links. | Prominent without leading. Above the fold where it counts. |
| Next block | Timestamps, correctly formatted, accurately labelled | Functional. Feeds chapters inside YouTube and key moments outside it. |
| Next block | Two or three sentences of genuine context: what is covered, what is not, who it is for, what to watch next | The classification layer, written as prose rather than as a keyword list. |
| Next block | The link block. Stable, labelled, short. | Routing, for the minority who scroll looking for it. |
| Last line | Three to five hashtags | The first three show above the title. Choose them as display copy. |
Three to four hundred words of that is plenty. The 5,000-character ceiling is not a target, and the marginal value of characters 1,200 through 5,000 is very close to nothing unless you are transcribing — which is a legitimate use, though captions and a transcript do the same job better and are read by more of the system. The companion piece on the field above the box is how to write YouTube titles, and the full search picture sits in the YouTube SEO guide.
A ten-minute audit on your last ten videos
- Read the first sentence of each, on a phone, unexpanded. How many of the ten tell a stranger what the video is? That number is your real description quality score, and on most channels it is low.
- Find and delete every keyword block. The comma-separated list below the horizontal rule. YouTube's documentation calls that a spam-policy violation and values the benefit at approximately nothing.
- Count your hashtags and look at the first three. Those three are the only ones with a visible position. If they are generic — one of them is probably your niche and nothing else — replace them with the categories the video genuinely sits in.
- Check your disclosures are above the fold. Any sponsored or affiliate video where the disclosure sits under the link block needs moving, not rewriting.
- Click your own links. Especially on Shorts, where they have not been clickable since 2023, and on anything older than a rebrand.
- Open your upload defaults. If the template starts with generic text rather than an empty first line, restructure it so the specific sentence lands at the top.
- Check one video in a language you do not publish in. If what appears is a machine translation and that video carries claims or disclosures, add the translation yourself.
None of this is glamorous and all of it is cheap. A description pass on a back catalogue of fifty videos is an afternoon, it carries no risk of resetting anything the algorithm has learned, and it fixes a field that most channels have been treating as either a ranking machine or a filing cabinet for links.
What a description cannot do
Here is the part the optimisation advice tends to leave out. A description is read after the click, by machines before the click, and by almost nobody in between. It has never persuaded a person to watch a video. The two things that do that work sit above it, in the hundred characters of the title and the 1,280 pixels of the thumbnail, and they are competing for a decision that gets made in well under a second against every other option on the same screen.
Which is why a channel with a weak description and strong packaging grows, and a channel with an immaculate 1,200-word description and a thumbnail nobody can read at phone size does not. Get the description right because it is cheap, takes an afternoon, keeps you the right side of a spam policy and gives the summarising surfaces something accurate to work from. Do not expect it to move a number that is decided before anyone has seen it.
The decision it is downstream of is the one worth most of your attention, and that is the problem Thumblore exists to solve: a frame a stranger can read instantly, in a style your channel can hold across every upload.
Then write one honest sentence at the top of the box, put your timestamps underneath it, and stop pasting the keyword list. That is the whole job.