Most title advice is about getting the click. Curiosity gaps, numbers, strong verbs, specific promises. All of it works to some extent.
What it mostly omits is that the title also determines what the viewer expects, and a viewer whose expectation is not met leaves. Which means an effective title and a good title are not the same thing.
The two failure modes
A title can fail by being unappealing, which everyone worries about, and by being inaccurate, which fewer people do.
The inaccurate version is more damaging because the cost is invisible. The click happened. The metric that everyone watches improved. The viewer left after ninety seconds feeling misled, and that shows up somewhere else entirely, later.
I did this deliberately once as a test with a title that implied more than the video delivered. Best click-through of the quarter, worst average view duration, and the video underperformed overall by a wide margin.
What the title needs to establish
Three things, and most titles handle one.
What the subject is. This is the part everyone gets right.
What kind of treatment it gets. A test, an explanation, an opinion, a story, a tutorial. A viewer looking for instructions and receiving an essay is disappointed regardless of the essay's quality.
And roughly what scale. A title that sounds comprehensive on a topic covered in six minutes sets up a disappointment. A modest title on a thorough treatment undersells it.
The specificity trade
Specific titles get fewer clicks and better retention. General titles get more clicks and worse retention.
This shows up consistently in my own data and is worth understanding rather than just picking a side.
A general title is a wide net. It catches people with adjacent interests, most of whom leave. A specific title catches fewer people, nearly all of whom wanted exactly this.
Which is better depends on what you are optimising for. For growth in a subject where you want the right audience rather than a large one, specific wins clearly. For something intended to reach broadly, general has a case.
My own default has moved toward specific over several years, mostly because the audience quality difference compounds and the view count difference does not.
The curiosity gap and its cost
The technique of withholding the key information so the viewer must click to find out.
It works. It is also a debt, because the viewer arrives expecting the answer and every second before they get it is a second they are considering leaving.
The version that works is a title that raises a question and a video that answers it early, then explains. The version that fails is a title that raises a question and a video that withholds the answer for eight minutes to maximise watch time, which viewers detect quickly and resent.
Titles that age
Something worth thinking about for anything intended to last.
A title referencing the current year dates immediately and looks stale within months. A title referencing a current event is meaningless once the event passes.
A title stating a persistent question does not date. Those are the videos still accumulating views three years later, and their titles are generally the plainest ones I have written.
The exception is genuinely time-bound content where the date is useful information, in which case it belongs there.
The relationship with the thumbnail
Which I have written about elsewhere and which belongs here too.
The title and the thumbnail are one unit and should not duplicate each other. If the thumbnail says something and the title says the same thing, one of them is wasted.
The arrangement that works for me is the title stating the subject plainly and the thumbnail showing something specific that raises a question about it.
Changing titles later
Worth doing and worth doing carefully.
A video that underperformed with one title sometimes performs with another, and there is no cost to trying if the video is not currently doing well.
What is riskier is changing the title of something performing well, since whatever is happening is working and the change may stop it. I have done this twice and regretted it once.
The safer version is changing titles on older work that has settled into a low steady rate, where there is little to lose.
The rule I use now
Write the title before making the video, and then check at the end whether the finished thing delivers it.
If it does not, change one of them. Usually the title, occasionally the video.
The discipline of writing it first also catches vague ideas early. A video whose title cannot be written in a sentence is a video without a clear subject, and that is better discovered before shooting than after.
Length, and where it gets cut
A mechanical point that matters and is easy to forget. Titles are truncated differently in different places — search results, the home page, the sidebar, notifications, and on a phone versus a desktop.
The safe assumption is that only the first forty or so characters are reliably visible everywhere.
Which means the important part goes first. A title whose subject appears at the end is a title that reads as nothing in half the places it appears.
I check this by looking at my own videos on a phone, in the places they actually show up, rather than in the upload interface where the whole thing is always visible.