Guide · Psychology
Write a Video Title That Earns the Click — with AI, Without the Bait
A number in your title can lift click-through by 20 to 30%, "7" beats "Several" every time, and the top titles run about eight words. But clicks are the easy half. Dan Ariely's experiments turned up something stranger: a label given before you taste changes how good the thing actually tastes — the same beer, the same coffee, rated better or worse on nothing but what you were told first. Your title is exactly that label. It doesn't only decide whether someone clicks; it pre-sets how good your video feels once they do. Over-promise, and a perfectly good video lands as a letdown. AI can spin you fifty titles in a breath — what it can't do is pick the one that wins the click and keeps the promise. This guide uses Ariely to write that one.
Before you start
- A video that's actually done, or close to it. You can't promise honestly what you haven't watched — the title has to match the thing, not your hopes for it.
- An AI writing tool to generate variants — ChatGPT, Claude, or similar. You'll ask for many and judge them yourself, not take the first.
- Three real competitor titles from your niche, copied as they appear in the feed. You'll judge yours next to them, because that's where it actually competes.
- One honest sentence of what the viewer gets by watching. If you can't write that sentence, no title will save you — the problem is upstream.
Write a promise, not a label
Ariely's first finding is that people have no built-in meter for value — we judge almost everything by comparison, against whatever sits next to it. A label describes your content: "My morning routine." A promise offers a payoff worth choosing over the next title in the feed: "The six-minute morning that killed my 3pm crash." Same video, different job. Prompt the AI plainly: "turn this topic into ten titles, each a specific promise of what the viewer walks away with — not a description of what the video contains." Then cut every one that merely names the subject.
Why a promise winsA title that only labels the content gives the scrolling eye no reason to pick you over the title right below yours. The viewer isn't asking "what is this?" — they're asking "what do I get, versus the next thing I could watch instead?" Answer that question, in the title.
Drop in one concrete anchor
The first number a person sees sets the frame for everything after — Ariely calls it anchoring, and title data backs it hard: a specific number lifts click-through by twenty to thirty percent, and it reads as more credible than a vague claim. "I saved money" is a shrug; "I saved $4,200 in 30 days" is a frame. Ask the AI to inject one true number, timeframe, or named thing into each title, and to delete every soft quantifier — "a lot," "some," "a while." A specific anchor tells the viewer exactly what they're weighing.
The anchor must be trueA number that the video doesn't honor is a false promise dressed as precision — and Ariely's next lesson is precisely why that costs you more than a vague title ever would. Specific, yes. Invented, never.
Open a curiosity gap — and leave it open
Curiosity fires in a narrow band: you have to know enough to feel a gap, but not enough to close it yourself. "You won't believe what happened" is too empty to open any gap at all — it names nothing. Spelling out the whole answer in the title leaves no reason to press play. The sweet spot names the subject and the stakes while withholding the turn. Ask the AI for titles that say what this is about and why it matters, then stop one beat before the payoff — and keep it near eight words, in the 45-to-55-character range where titles read cleanest in a feed.
Set an expectation the video beats
Here is the lesson almost no one applies to titles. Ariely gave people the same food and drink with different labels, and the better label made the same thing genuinely taste better — information delivered before an experience reshapes the experience itself. Your title is that information. So it isn't only a magnet for the click; it is the dial that sets how good your video is allowed to feel before anyone presses play. Promise more than you deliver and a fine video registers as a disappointment, and that early exit quietly tells the algorithm to stop handing you out. Aim the title one notch below what the video actually gives, so it clears its own bar and lands as a win.
Generate fifteen, judge against the feed
Ask the AI for fifteen to twenty variants, then judge them the way a viewer does — never alone, always shoulder to shoulder with the titles that will surround yours. Ariely's relativity again: you can't read absolute worth, only the clearly better option in a side-by-side. Paste your best title beneath the three real competitor titles you copied, and ask yourself which one your thumb goes to. Have the AI also write the deliberately boring version of your title, so the contrast is visible on the page. Cut anything that shines alone but blurs into the crowd.
Judge it small and surroundedYour title never appears the way you're reading it now, large and by itself. Look at it at feed size, stacked against rivals — that's the only test that matches where it competes, and it will quietly kill the clever title that needed a quiet room to work.
Test one variable — judge by watch-time, not just clicks
Change exactly one thing between two versions — the number, the gap, the verb — and let them run. YouTube's native title test pits two titles against the same video for channels past a thousand subscribers; below that, swap one in and watch. Now apply Ariely's lesson as a rule: a title that wins clicks but sinks your average view duration sets an expectation the video couldn't keep. It was bait, the bounce poisons everything downstream, and the platform reads it as a reason to bury you. Judge every title by clicks and by retention together — the right one lifts both, because the promise held.
Why both numbersClicks alone reward the lie; a clickbait title can spike them and wreck the watch-time underneath. The honest title is the one where the click and the minutes after it move in the same direction — that's the signal that your label matched the bottle.
Run it on a real video. You filmed yourself cutting your grocery bill without coupons. Step one, a promise, not a label: not "How I grocery shop" but "I cut my grocery bill in half — and never clipped a coupon." Step two, the anchor: "half" becomes "$310 a month," concrete and true. Step three, the gap: you keep "without a single coupon" because it names the surprise and withholds the how. Step four, the expectation: your method is genuinely solid, so you resist "$310 a month, almost effortlessly" — the "effortless" is a notch too high, and you cut it so the video can over-deliver. Step five, you generate eighteen variants and lay your favorite under three real competitor titles; it still pulls the eye, so it survives. Step six, you A/B the number against a version with the curiosity gap up front, and keep the one that won clicks without costing watch-time. Twenty minutes, and the part AI couldn't decide — which promise your video could actually keep — stayed yours.
Check your work
- The title makes a specific promise of what the viewer gets — not just a label of the topic.
- It carries one true, concrete anchor: a number, a timeframe, or a named thing.
- It opens a curiosity gap — names the subject, withholds the turn — in roughly eight words.
- It sets an expectation the video clearly beats, not one it strains to reach.
- I judged my best title beside three real competitor titles, not on its own.
- I tested one variable and kept the version that won clicks and held watch-time.
The one line to keep
A title is a label you taste before the video. Promise a notch under what you deliver, and the same video lands as a win.
Framework drawn from Dan Ariely's Predictably Irrational — relativity (we judge by comparison, not absolute value), anchoring (the first number sets the frame), and above all expectations (information given before an experience changes the experience itself). Title data — a number lifting click-through by roughly 20–30%, top titles near eight words and 45–55 characters, the curiosity gap as the most effective honest technique, and platform-native A/B title testing — is drawn from 2026 video-title benchmark reporting. Tools named (ChatGPT, Claude) are examples, not endorsements. A popular-science, how-to reading; intellectual property belongs to the original author. © vlog.bluecatbot.com 2026.