YouTube and your newsletter: an audience that's yours

After this playbook you’ll know what YouTube really measures, how to test titles and thumbnails without guessing, and how to turn viewers into a newsletter that lands in the inbox. Labels: ✅ primary (the platform or company itself) and ⚠ secondary or self-reported.

How it really works

• YouTube ✅: signals are “clicks, watchtime, survey responses, sharing, likes, and dislikes”. There is “no set formula”.
• YouTube ✅ (valued watchtime): viewers rate videos in surveys with 1-5 stars; only 4-5 stars counts as valued watch time. A model predicts that score for everyone who doesn’t fill in the survey. So a click alone says little.
• YouTube Test & Compare ✅: tests up to 3 titles, 3 thumbnails or combinations, and picks the winner on watch time share, not on click-through rate. Not for Shorts.
• YouTube Analytics ✅: “Intro” is the percentage of viewers still watching after the first 30 seconds. YouTube recommends experimenting with it.
• Gmail ✅: every sender needs SPF or DKIM and a spam rate below 0.3% (Google advises below 0.1%). Anyone sending 5,000+ emails a day to Gmail needs SPF and DKIM, DMARC and one-click unsubscribe. Since November 2025 Gmail rejects non-compliant mail more strictly, including permanently.
• Yahoo ✅: similar requirements, plus: process unsubscribes within 2 days (like Gmail) and “Don’t purchase mailing lists”.

Who does it perfectly

Veritasium (YouTube)

• What: in “Clickbait is Unreasonably Effective” (2021), Derek Muller tested a clearer title and thumbnail and distinguished between honest and misleading clickbait.
• Result: the video performed clearly better after new packaging ⚠ (via press; exact figures not verified).
• Steal this: exaggerate the emotion or the stakes, never the facts.

The Rundown AI (newsletter, AI niche)

• What: daily 5-minute AI newsletter, plus interviews with AI executives, courses and spin-offs per topic.
• Result: says itself 1 million subscribers in February 2025 and “3,500,000+ readers” on the homepage now ⚠ (self-reported).
• Steal this: one fixed format with a fixed reading time, the same every day.

There’s An AI For That (newsletter, AI niche)

• What: an AI tool directory with a newsletter, scaled up from 3 times a week to daily.
• Result: 1.7 million subscribers in less than 2 years, with a $0 ad budget ⚠ (self-reported via a beehiiv case study).
• Steal this: build something useful (an overview, a tool) and let the newsletter be the reason to come back.

Morning Brew (newsletter)

• What: every subscriber gets a personal share link, with rewards at milestones.
• Result: referrals are cited as around 30-35% of growth ⚠ (figures differ per source and year).
• Steal this: a low first threshold (3 sign-ups) with a digital reward.

Do this

1. Title and thumbnail first, then film

Without a click your video doesn’t exist for the system; without a satisfied viewer the click yields nothing ✅.

1. Write 10 ideas as titles. Feel free to let AI make 20 variants, choose yourself.
2. Sketch a thumbnail for the best 3: one subject, one emotion or contrast, max. 3-4 words.
3. Check: does what the thumbnail promises appear in the first 30 seconds? No: adjust it or don’t make it.

TitleMain keyword (optional)

Score

–

Free · runs entirely in your browser · nothing is stored or sent

2. Let Test & Compare settle the debate

Test & Compare rewards watch time, not clicks

Test & Compare picks the winner on watch time share ✅. A misleading variant that draws lots of clicks but makes people drop off therefore often loses. In desktop Studio: video → title or thumbnail → A/B Testing. Make variants that really differ (different concept, not a different colour) and start with older videos, as YouTube itself recommends. Keep a log of what won.

3. Deliver in the first 30 seconds

1. Seconds 0-5: show what the thumbnail promised. No logo intro, no “welcome back”.
2. Seconds 5-30: why this matters and what the viewer will have at the end.
3. After 48 hours: Analytics → Retention → Intro and Dips. Compare with your own average, not with other people’s screenshots.

A leaked MrBeast production document names the same three core metrics (click-through rate, average view duration and % viewed) and says the first minute must prove the thumbnail’s promise ⚠ (authenticity not officially confirmed).

4. Let your Dutch video also run in English

Automatic dubbing NL → EN

YouTube supports automatic dubbing from Dutch to English and back ✅. For eligible creators it’s on by default. Set the video language correctly, otherwise YouTube translates from the wrong language, and check the dub once: jargon often goes wrong and dubs can’t be edited ✅. According to YouTube, creators with multi-language audio get more than a quarter of their watch time from the non-primary language ⚠ (via press).

5. Build your list from video 1

1. Make one concrete free thing per topic (checklist, prompt pack, template).
2. Landing page with one field and double opt-in (proof of consent).
3. Mention it in the video at the moment the problem comes up, plus in the description and a pinned comment. Measure sign-ups per video with a separate link.
4. Welcome sequence: email 1 delivers the magnet and asks “where are you stuck?”. According to estimates cited by beehiiv, more than 80% of independent newsletters use nothing more than one welcome email ⚠, so you already stand out here.

6. Grow via other newsletters

Recommending earns recommendations

According to beehiiv, newsletters taking part in the free recommendation network grow 2.75 times faster, and those who recommend others get recommended 32 times more often themselves ⚠ (platform self-report; correlation, not proof of cause). Put 5 newsletters with an overlapping audience in your recommendations and ask for reciprocity. Then: a referral programme with milestone rewards (Morning Brew model).

Grey: possible, with risk

Tactic	Risk	Safer
Clickbait packaging	“A thumbnail that misleads viewers” = removal and possibly a strike ✅. Disappointed viewers give low survey scores ✅	Everything on the thumbnail is in the video, preferably in the first 30 seconds
Faceless AI at scale	Spam policy now literally prohibits “Using automated tools or AI to churn out high volumes of similar content with minimal changes” ✅. Enforcement sometimes hits good channels too ⚠	Faceless is allowed. Your own angle, your own demos, fewer but better videos
Paid recommendations (paying per subscriber)	Subscribers who clicked once on a “recommended” screen drop off more often, which can cause spam complaints and deliverability damage	Small budget, compare click-through rate per source after 30 days, clean out inactive subscribers after 60-90 days
Giveaways	Prize hunters with a prize unrelated to your topic	Niche prize (a call, your template), separate tick boxes for entry and newsletter

Black: how it works and why you don’t do it

Practice	What it is	Why it goes wrong
Sub4sub	Subscribing in exchange for a subscription back	Literally prohibited in YouTube’s fake engagement policy ✅
Bought views, likes, subscribers	Services or bots that inflate numbers	Removal up to channel termination, even if a hired agency did it ✅
Detection evasion	Re-uploading someone else’s video mirrored, sped up or cropped	Explicitly prohibited in the spam policy ✅
Bought or scraped email lists	Addresses without consent	Yahoo: “Don’t purchase mailing lists” ✅. Above 0.3% spam complaints you lose your right to support at Gmail and your delivery drops sharply ✅. Also in breach of the opt-in rules for email marketing in the EU ⚠
Pre-ticked consent	The newsletter box is already ticked	Not valid consent; Yahoo explicitly names it ✅

Security

• Channel hijacking often starts with a fake sponsor email. An attachment steals your session cookie; the channel becomes a crypto scam or gets resold. Don’t open files from sponsor emails and consider Google Advanced Protection.
• Newsletters get hijacked the same way. A Substack journalist clicked a fake “suspicious activity” email and the newsletter became a Trezor imitation. Turn on 2FA and never log in via a link in an email.
• After a platform leak comes the phishing. Substack leaked email addresses and phone numbers (reported Feb 2026): ignore emails and texts about “account problems”.

Myths

• “Upload every day, or your channel sinks.” No official basis. At VidCon 2026 YouTube said uploading less often isn’t penalised ⚠ (second-hand report).
• “An AI label costs reach.” YouTube: “Disclosing AI content alone won’t limit a video’s audience or impact its eligibility to earn money” ✅.
• “Videos must be exactly 8 (or 10, or 20) minutes.” YouTube’s creator liaison: “As long as it needs to be, but no longer” ⚠ (widely quoted, transcript not read ourselves).
• “My 50% open rate is good news.” Apple MPP counts opens that aren’t there ⚠; beehiiv calls open rates “less reliable” because of this ✅.

Measuring

• Impressions click-through rate alongside average view duration: never one of the two on its own.
• Intro % and Dips in your retention report.
• Traffic sources (Analytics → Reach): search, browse or suggested? Make more of what works.
• Sign-ups per video via a separate link.
• Newsletter: clicks, replies, unsubscribes and spam rate (Google Postmaster Tools). Open rate only as a trend.

Your first 30 days

For a creator or freelancer in the AI niche:

1. Day 1: choose a newsletter platform and your own (sub)domain as sender, never an @gmail.com address.
2. Day 2: set up SPF, DKIM and DMARC (p=none is fine) via your platform’s instructions. Sign up for Google Postmaster Tools.
3. Day 3: make one lead magnet and a welcome sequence of 3 emails, with double opt-in.
4. Week 1: write 10 titles for “how to do X with AI” videos (search traffic is easier than browse). Pick 3 with an honest thumbnail.
5. Week 2: record video 1. Thumbnail promise in seconds 0-5, lead magnet mentioned at the moment the problem comes up.
6. Week 2-3: turn on automatic dubbing and check the English track once.
7. Week 3: start a Test & Compare on your title or thumbnail. Put 5 newsletters in your recommendations.
8. Day 30: look at Intro %, traffic sources, sign-ups per video and your emails’ click-through rate. Change one thing, not ten.

Checklist

0/4

• I come up with title and thumbnail before I film
• What the thumbnail promises is in the first 30 seconds
• SPF, DKIM and DMARC are set up on my own domain
• I steer on clicks and replies, not on open rate

Your progress is only stored in this browser.

Test yourself

1. 1What does YouTube Test & Compare use to pick the winning title or thumbnail?

   The highest click-through rateThe most watch timeThe most likes

2. 2What counts as 'valued watchtime' according to YouTube?

   All watch time above 30 secondsWatch time that viewers give 4 or 5 stars in surveys (or that a model predicts as such)Watch time from subscribers

3. 3Why is open rate a poor measure for your newsletter?

   Because Apple Mail Privacy Protection loads tracking pixels even if nobody opens the emailBecause Gmail prohibits open ratesBecause open rates are always too low