A hook is the opening promise that helps a viewer decide whether the next few seconds are relevant. It is not a guarantee of reach, retention, or sales. Choose a pattern because it makes the video’s question or evidence clearer, then make the body deliver what the opening led the viewer to expect.

The twelve patterns below are deliberately distinct. Each includes a good use, a poor use, and a truth/accessibility note. Adapt the words to the actual material; never borrow a dramatic promise that the footage cannot support.

1. The direct question

Open with a question the video can answer: “Which of these two layers blocks wind better in this test?” Use it when the audience has a specific decision to make and the answer will arrive in the body. Avoid broad questions such as “Want to change your life?” when the video only offers a small tip. Put the question in spoken words and captions so it is available without sound.

2. The visible result

Show the relevant result first, then explain the steps or conditions that produced it. This suits a craft demonstration, recipe, styling change, or before-and-after presentation where the result is real and inspectable. It is a poor fit when the result requires hours of context or looks like a clinical outcome. Label time, preparation, and editing; a reveal is not proof of causation.

3. The problem in progress

Begin at the moment a problem is visible: a tangled cable, a confusing setting, or a draft that will not export. Use it when the viewer can recognize the problem and the video provides a concrete fix. Do not manufacture damage or hide the easier cause to make the problem look urgent. Describe the visual state for viewers who cannot see it clearly.

4. The specific promise

State a bounded outcome and scope: “In this one-minute walkthrough, you will set up three caption styles.” Use it when the deliverable can genuinely fit the stated scope. Avoid promises of guaranteed results, universal suitability, or an exact performance outcome without evidence. Put limits—software version, audience, material, or test condition—in the opening or immediately after it.

5. The surprising comparison

Place two relevant options side by side: “The same edit with room tone and without it.” This works when the comparison uses matched conditions and the distinction matters to a choice. It fails when framing, lighting, duration, or treatment changes between options. Label the variables and do not convert one person’s preference into an objective winner.

6. The mistake and correction

Show a common error, then correct it. Use this for teachable workflows where the correction is safe and reproducible. Do not stage a dangerous failure, ridicule a participant, or imply that every other method is wrong. Keep the incorrect state brief but understandable, and caption the action so viewers can follow without audio.

7. The counterintuitive observation

Lead with an observation that challenges a familiar assumption: “More sharpening can make small text less readable.” Use it when the body can explain the condition and support the nuance. Avoid contrarian wording that simply reverses a fact for attention. Separate an editorial recommendation from a claim about health, money, safety, or universal performance.

8. The process promise

Invite the viewer into a visible sequence: “Watch this loop go from clicky to clean.” It suits a transformation in which the intermediate decisions matter. It is unsuitable when the process is mostly hidden or the edit will skip the only evidence. Show enough of the steps to make the promise honest, and mark resets, time jumps, presets, or professional intervention.

9. The quoted question

Start with a real audience question from a brief, comment, or support request, then answer it. Use it when the question is representative and permission/privacy are handled. Do not invent a fake customer or expose a private commenter to create intimacy. Keep the answer specific and link the wording to the relevant source or demonstration.

10. The sensory detail

Open with a sound, texture, movement, or close-up that makes the subject tangible: a zipper, a room tone change, or a brush crossing fabric. Use it when the sensory cue is central to the lesson and can be described in captions or narration. Avoid strobe, flashing, extreme volume, or a close crop that implies detail the camera cannot resolve. Give viewers enough stillness to interpret the cue.

11. The time-bound context

Anchor the video in a real deadline or moment: “Before today’s client handoff, check these three files.” Use it when the timing changes the viewer’s decision. Do not fabricate scarcity, false urgency, or an expiring offer. State the actual date, time zone, platform version, or scope when those facts matter, and update or remove the hook when they expire.

12. The invitation to choose

Offer a genuine choice: “Which color reads better under this light?” Use it when the viewer’s response is part of the format and both options are represented fairly. Avoid using a poll to conceal a required disclosure, a safety decision, or a claim that needs evidence. Describe the options in text and moderate responses that reveal personal information or target a participant.

Match pattern to the body

Before recording, write the hook, the next sentence, the evidence shot, the limitation, and the payoff. If the hook says “compare,” include a controlled comparison. If it says “how,” show the method. If it says “today,” confirm the timing. A hook that is technically accurate but leaves out a material condition can still mislead by implication.

Use the first shot to orient the viewer, not only to decorate the feed. Keep the speaker, crop, sound, text, and visual subject consistent enough that the hook does not feel like a different video. W3C media guidance recommends planning captions, transcripts, and descriptions of important visual information from the start. Write a caption or integrated narration for what the opening shows, including relevant sound and on-screen text.

Choose hooks responsibly for commercial content

If the video is paid, gifted, affiliate-linked, or made under an employment relationship, include a clear disclosure where the endorsement occurs. The FTC advises that disclosures be conspicuous and understandable; a vague “collab” or a disclosure hidden after “more” may not be enough. Build the disclosure into the script and caption of every variant. A hook must not make a stronger product or health claim than the approved body supports.

Clear music, stock footage, logos, participant consent, and location permissions for each opening. A short hook does not create an automatic right to use someone else’s work. Keep original files, license terms, and approvals in the project record.

Run a controlled hook test

Use a modular master: three hook tracks, one shared body, one shared payoff, and one shared disclosure. Record the hooks in the same session with the same speaker, camera, light, audio, captions, music, duration, and call to action. Change one intentional variable—the pattern—not the evidence or claim.

Choose a test path deliberately:

  • A native concurrent experiment is strongest when the platform supports the content type. YouTube’s current A/B tool can compare up to three titles and thumbnails on eligible long-form videos, but not Shorts; do not describe that feature as a hook test.
  • A sequential organic test can compare separate posts, but audience mix, timing, recommendation systems, and account momentum differ. Rotate order across comparable projects and call the result directional.
  • A paid split test can hold audience, placements, budget, dates, and objective steady, but paid findings do not automatically transfer to organic distribution.

Write the hypothesis and primary measure before publishing. YouTube Analytics Advanced Mode supports comparisons across videos, groups, metrics, and periods; retention, average view duration, and watch time answer different questions. Do not select a “winner” from raw views alone. Annotate changes in audience, traffic source, moderation, spend, caption, sound availability, or upload time.

When variants run at different times, avoid causal language. Say “Pattern 5 had the stronger early-retention signal under these conditions,” not “Pattern 5 caused growth.” If the platform reports the result as inconclusive, keep it inconclusive. The next action may be to clarify the promise, improve the body, or run a genuinely concurrent test.

Accessibility and safety gate

Check every hook with sound off and captions on. Captions should include speech and relevant non-speech audio; a descriptive transcript or integrated narration should convey visual information that the hook requires. Keep text readable and avoid flashing or motion that obscures the subject. Provide a calmer version when a decorative effect is unnecessary, and never make a safety instruction depend on a rapid visual.

Do not use fear, humiliation, medical alarm, dangerous stunts, or manipulative urgency as a substitute for relevance. Obtain consent for identifiable people, private comments, locations, and personal conditions. If the hook involves health, safety, financial, or legal claims, route wording through a qualified reviewer and retain the supporting source.

A hook review card

For each pattern, record:

  • exact spoken and on-screen wording;
  • visual and audio action in the opening;
  • body evidence and payoff;
  • material limitation or qualification;
  • disclosure, caption, rights, and consent status;
  • destination, version, and test condition; and
  • observation, uncertainty, and next test.

Release only when all twelve candidates have been judged against the same brief, the chosen hook delivers its promised answer, accessibility and disclosures survive the crop, rights are recorded, and the result is described no more strongly than the evidence allows.

Sources and scope

These sources support the platform, accessibility, and disclosure boundaries. The twelve pattern definitions, use cases, test workflow, and review card are Angry Media editorial synthesis. This article is educational, not legal, medical, statistical, or platform-policy advice. Source review completed 2026-09-15; scheduled publication 2026-11-10.