Almost everything a creator does splits into three questions that get confused with each other. Will anyone want this? That is demand, and it exists before you make anything. Will the system show it to them? That is distribution, a machine with rules you can read. Are you allowed to do it this way? That is the law — disclosure, copyright and likeness — and it does not care how small your channel is. Craft sits inside the first two and is worth less than people expect without them. Use this guide to sort a problem into the right question, because the commonest creator mistake is answering one of them with effort aimed at another.
Each chapter opens with the short version. Tap one to read the detail.
The business decision that comes before any content
~2 min
Positioning is choosing who you are for and what they get, and it constrains everything downstream. A channel that cannot say who it is for cannot be found, recommended, or sponsored — three different systems all keyed to the same answer.
Before format, platform or schedule comes a business question: who is this for, and what do they get that they cannot get elsewhere?
Discovery needs the answer, because search and recommendation systems infer a topic from consistency. A channel covering six unrelated subjects gives them nothing stable to match.
Audience needs it, because a subscriber is predicting your next piece. If they cannot predict it, subscribing means nothing.
Monetization needs it, because sponsors buy access to a defined audience, not a headcount. A small, precisely defined audience is routinely worth more than a large, diffuse one; advertisers pay for relevance, and relevance is what positioning states.
Running a channel as a business also brings the ordinary obligations: a legal structure, records, tax on income, and contracts for anything involving payment. Creators often discover this when money first arrives, which is the worst time.
One structural decision belongs here too: every platform audience is rented. A follower count is access on terms the platform sets and changes. An email list reaches an inbox without an intermediary deciding, survives a suspension and moves between platforms, which is why experienced creators build one early rather than in a crisis.
Concretely, the owned channel comes with a short, checkable list from the FTC's own guide: no deceptive subject lines; disclose clearly and conspicuously that the message is an advertisement; include your valid physical postal address; explain how to opt out and honor the request within 10 business days. If a service sends for you, you cannot contract away the responsibility.
Rule: write down who the work is for and what they get, in one sentence — then build one channel you own, before you need it.
Demand exists before you make anything
~2 min
Search behavior is evidence of demand that already exists, and it is observable in advance. The alternative — making something and hoping it finds an audience — is the same work with the evidence collected afterwards.
The most expensive habit in content is making something first and asking whether anyone wanted it second.
Search data is demand made visible. What people type is a record of questions they already have, in their own words, in volumes you can compare.
The useful distinction is intent. Someone searching a how-to question wants an answer and will leave once they have it. Someone searching a comparison is deciding. Someone searching a name already knows what they want. Content that mismatches intent underperforms no matter how good it is, because it answers a question nobody asked.
The corollary most creators miss: an unanswered question is worth more than a well-answered one. Where existing results are thin, outdated or answering a different intent, there is room. Where the question is thoroughly covered by people with more authority, being marginally better is not a strategy.
Concretely: Google publishes the questions it wants you to ask of your own draft — does it provide original information, reporting, research or analysis; would you expect to see it in or referenced by a printed magazine, encyclopedia or book; does the title avoid exaggerating or being shocking; is there a byline where one would be expected? A piece that answers no to the first two is the well-answered question again, in your voice.
Rule: before making anything, find the question it answers and check who already answers it — if the answer is "well, and by everyone", pick a different question.
Packaging earns the click, structure earns the minute, and the numbers say which failed
~2 min
Titles and thumbnails decide whether anyone starts, and the opening decides whether they stay. They fail differently — a packaging problem shows as low click-through, a structure problem as a drop-off curve — and diagnosing the wrong one wastes the fix.
Packaging — the title, the thumbnail, the first line of a description — decides whether anyone begins. Its job is a promise specific enough to be worth acting on. Vague packaging fails not because it is dishonest but because it gives no reason to choose this over the next item.
Structure decides whether they stay, and the first moments carry the most weight. An opening that restates the title, introduces the creator or explains what is coming spends your viewer's patience before delivering anything. Opening inside the substance is the most reliable structural improvement.
The trap that ties them together: packaging that over-promises fixes the first number and destroys the second. A title that wins the click and is not delivered on produces exactly the retention shape that tells a recommendation system the piece was not worth showing.
Read the funnel in order, because each stage conditions the next. Impressions say the system offered the work. Click-through says the packaging worked on people who saw it. Retention says the content kept the promise. Return says the audience wants more.
That order makes the numbers diagnostic. Falling views with stable click-through means fewer impressions, a distribution question. Falling views with falling click-through means packaging. Good click-through with poor retention means the promise was not kept. One drop in views has three different causes, and only the sequence separates them.
Cross-platform totals double-count the same person and last-touch attribution ignores everything that led there, so treat any single-platform total as a lower bound.
Rule: locate the stage where the audience is lost before changing anything — the title and the first thirty seconds are different problems with different fixes, and a metric names where, not why.
Formats are different contracts with the audience
~2 min
Short-form, live and audio are not one skill delivered at different lengths. Each sets a different expectation about attention, interaction and completion, and content moved between them without adaptation usually fails on the expectation rather than the craft.
Short vertical video is browsed rather than chosen. Its viewer has not committed, so the first moment must earn the next, and completion matters more than duration. Its strength is reach among people who have never heard of you; its weakness is that reach converts to lasting audience poorly, because a viewer who never chose you has formed no relationship.
Live trades production for presence. Its value is that it is happening now and can respond — which means the interaction is the content, and a live stream that could have been recorded has given up its only advantage.
Audio is consumed while doing something else. That single fact drives everything: no visuals to carry meaning, so structure must be audible; attention is divided, so signposting matters more; and episodes are typically completed rather than sampled, which makes the middle worth more than in video.
Concretely: the platform's own numbers show the contract. A Short can run up to three minutes, but a Short over one minute carrying a Content ID claim is blocked, and licensed music covers at most 90 seconds of it — so a three-minute cut of a long video, set to a track, fails on rules before it fails on attention.
Rule: pick a format for the attention it assumes, and when moving a piece between formats, re-make it rather than re-cut it — the contract changed even though the topic did not.
Writing for a reader who is scanning
~2 min
Online readers scan before they read, and decide from the scan. Front-loaded conclusions, informative headings and short paragraphs are not stylistic preferences — they are what makes a page usable to someone deciding whether to read it at all.
Put the conclusion first. A reader arriving from a search has a question and is scanning for the answer. Building to a conclusion works when attention is committed; it fails when the reader is deciding, paragraph by paragraph, whether to stay.
Headings are navigation, not decoration. A reader scans headings to find their section, so a heading that says "Background" tells them nothing while one that states its own point lets them skip accurately.
Short paragraphs and plain words are about scanability, not about writing down to anyone. A dense block signals effort before it signals value, and specialist vocabulary used without explanation excludes exactly the reader who most needed the explanation.
Plain language is also an accuracy discipline: writing "we will decide by June" rather than "a determination will be made in due course" forces a commitment that the vaguer phrasing conceals. Much bureaucratic prose is vague because vagueness is safer, not because the subject is complex.
Rule: write the sentence that answers the reader's question, put it first, and make every heading say something — then check that the headings alone tell the story.
Recommendation systems, what they can read, and why accessibility is the same work
~2 min
A recommendation system predicts satisfaction from behavior, not quality. It also cannot see inside a video or an image, so metadata is how a machine learns what a piece is about — and captions, transcripts and alt text are that metadata as well as the accessible version.
The system optimizes for predicted satisfaction, measured by behavior. It cannot assess quality; it observes whether people who were shown something engaged with it and stayed. That is why a small but strongly engaged audience can outperform more views, and why misleading packaging is punished without anyone judging it.
The system cannot see or hear. Text is what it reads. Titles, descriptions, transcripts, captions and structured data are the machine-readable surface of a piece, and a video with no transcript is largely opaque to it. Here accessibility and distribution become the same work: captions serve viewers who need them and are the text a system indexes.
Structured data states explicitly, rather than leaving it inferred, what the page is, who made it, when, and what it contains, in a form built to be parsed.
Concretely: Google's feed of interest-based content needs no special tags or structured data. Being indexed and within policy makes a page eligible, though eligibility does not guarantee it appears. What it does specify is the image: at least 1200 px wide and more than 300,000 pixels in total.
Captions serve viewers who are deaf or hard of hearing, and everyone watching without sound, which on some platforms is most of the audience. Automatic captions are only a starting point: they mishandle names, technical terms and overlapping speech, and an uncorrected track can be worse than none when it confidently states the wrong word.
Transcripts make audio searchable, skimmable and quotable. Alt text describes images for screen-reader users; convey the function of an image, what the reader needs from it, rather than cataloging what it contains.
Rule: give the system the text it needs, treat behavior as its only measure of quality, and caption and transcribe everything, correcting the automatic output — you will have done the indexing work too.
Moderation, and the audience setting with legal weight
~2 min
Community management is a policy problem before it is a personality one. And one audience setting is not a preference at all: content directed to children triggers legal obligations about data collection that override how you would prefer to operate.
A comment section is a small platform, and running one raises the same questions a large one does: what is allowed, who decides, and what happens when the rule was applied wrongly.
Deciding those in advance is what separates moderation from mood. Rules written after an incident always look targeted at the person involved, and enforcing an unstated rule reads as arbitrary even when the decision is right.
Content for children is a legal category, not an audience preference. Where content is directed to children, rules governing the collection of personal information from them apply, and platforms implement this through an audience setting that changes what features are available — typically restricting personalized advertising and interactive features like comments.
Two things make this trap sharp. Mislabeling has consequences that fall on the creator, not the platform. And "directed to children" is assessed on the content itself — subject matter, presentation, apparent intended audience — rather than on who happens to be watching. A creator who says "my content is for adults" while making animated content about toys is describing an intention, not the test.
The rules differ by jurisdiction and are revised; check the current ones for the audience you reach.
Rule: publish your moderation rules before you need them, and treat the children's-audience setting as a legal determination about the content — not a preference about the audience.
Revenue models, and the standards attached to advertising money
~2 min
Advertising, direct audience payment, and selling something of your own are three different businesses with different failure modes. Advertising uniquely comes with content standards, because advertisers choose what they appear beside.
Advertising pays per unit of attention and scales with reach.
Direct audience support — memberships, subscriptions, tips — is far more stable per participant and depends on relationship rather than reach. A small percentage of a committed audience routinely outperforms advertising on a much larger one.
Selling something of your own — a product, a service, a course — has the highest revenue per customer and the highest obligations: fulfillment, support, refunds, consumer-protection rules.
Advertising alone brings content standards, and the reason is structural: advertisers are buying adjacency and can refuse it. So the platform enforces suitability rules covering violence, profanity, controversial subjects and more — and a piece can be entirely legal, entirely within community rules, and still not eligible for advertising. Those are three separate tests, and creators routinely conflate the second with the third.
Concretely, as the help pages read on 2026-09-03: ad revenue on the largest video platform opens at 1,000 subscribers plus either 4,000 qualified watch hours in the last 365 days or 10 million qualified Shorts views in the last 90; memberships open at 500 subscribers with three public uploads in 90 days. And the not-advertiser-friendly list runs to fourteen topics — inappropriate language, violence, adult content, shocking content, controversial issues, sensitive events and on — with the page's own caveat that context matters.
Rule: know which revenue model each piece of work serves, and never assume that "allowed on the platform" means "eligible for advertising" — they are different standards with different judges.
Disclosure: the rule is clarity, not the presence of a word
~2 min
A material connection between a creator and a brand must be disclosed clearly and conspicuously. The legal test is whether an ordinary viewer actually notices and understands it — so a disclosure that is technically present but easy to miss fails.
This is where creators most often believe they are compliant and are not.
The obligation attaches to a material connection: payment, free product, an affiliate commission, a family or employment relationship — anything that might affect how the audience weighs the endorsement. It is not limited to cash; a free product you kept is a material connection.
The standard is clear and conspicuous, judged by effect rather than form. A disclosure buried below the fold, hidden behind a "more" expander, placed at the end of a video, colored to blend into the background, or reduced to an ambiguous tag among twenty hashtags fails, because the test is whether an ordinary person actually notices and understands it.
So put the disclosure where the endorsement is, in the same medium: spoken in a video as well as written, near the top rather than at the end, in plain words rather than jargon.
Endorsements must also be honest regardless of disclosure. An endorsement should reflect genuine experience and honest opinion, and a claim about results must be typical or its atypicality made clear. Disclosure permits a paid relationship; it does not license an untrue claim.
Synthetic and AI-generated media raise the same question: does the audience understand what they are looking at? Where realistic content is generated or materially altered, saying so is the same duty of non-deception, and platforms increasingly require a label.
The specific guidance is revised periodically. The principles are stable; check current requirements before relying on a particular form of words.
Rule: ask whether an ordinary viewer would notice and understand the disclosure in the moment they form an opinion — if they might not, it does not count, however technically present it is.
Copyright, fair use, and what a license actually grants
~2 min
Protection is automatic on creation, so almost everything you find is already someone's. Fair use is a four-factor judgment rather than a rule you can satisfy by duration, and a Creative Commons license permits only what its terms say.
Copyright attaches automatically on fixation, the moment a work is recorded in tangible form. Nothing needs registering, marking or claiming, so anything you find is protected by default, and a missing notice means nothing.
Fair use is a defense weighed on four factors, case by case: the purpose and character of the use, including whether it is transformative or commercial; the nature of the work; the amount and substantiality used relative to the whole; and the effect on the potential market for the original.
So discard the myths: there is no safe number of seconds, no fixed word count, and crediting the author is not fair use, since attribution is a courtesy, not a license. A short clip can infringe; a long quotation can be fair.
Licenses are the reliable alternative because they grant permission explicitly. Creative Commons licenses vary: some bar commercial use, some require derivatives to carry the same license, and all require attribution in the specified form. "It's Creative Commons" names a family of permissions with materially different terms, not a permission.
Takedown regimes give rights-holders a notice procedure and the accused a counter-notice, with platform systems on top; false statements carry consequences either way.
Two further rights are not copyright and are cleared separately. Trademark protects marks that identify a source, and its test is consumer confusion: would an ordinary person think this comes from, or is approved by, the mark's owner? Naming a product is generally fine; risk rises sharply when a mark appears in a channel name, logo or merchandise. Likeness rights protect a person's image, name and voice from commercial use without permission, so filming in public settles nothing. What matters is the use, not the capture, and a release documents permission for a specified use.
Clearing the copyright in a photograph does not clear the rights of the person in it: two permissions, two owners, two decisions.
Rule: assume anything you did not make is protected, prefer a license to a fair-use argument, and ask the trademark and likeness questions separately from the copyright one.
Written by Keentune. We are not affiliated with or endorsed by the organizations whose documentation informs this guide, and any linked sources belong to their respective owners.
All exam, test, and product names and trademarks are the property of their respective owners and are used here for identification and reference only. Keentune is independent study practice — not affiliated with, authorized, or endorsed by any of these organizations.