Never run out of YouTube video ideas: the five data sources that decide the next video for you

Every channel starts with a list. Ten, fifteen, sometimes twenty ideas written in the first week of excitement, and they carry the channel for two or three months. Then the list ends. The creator opens the editor with nothing in mind, publishes a weaker video to keep the streak alive, then skips a week, then another. The channel's upload history shows the pattern with brutal clarity: a burst of videos, a thinning, silence. It is the most common death on YouTube and it has nothing to do with talent.

The cause is where the ideas came from. The first list was inspiration: what the creator found interesting, what they had seen work elsewhere, what felt like the channel. Inspiration is a reservoir, and reservoirs empty. Demand is a river. The audience keeps searching, keeps commenting, keeps watching some videos three times longer than others, and every one of those behaviours is a video idea that already has a viewer waiting for it. The channels that publish for years are not more creative; they read the river instead of the reservoir.

A video idea worth making has to answer three questions at once: is someone already looking for this, has this channel earned the right to answer it, and can it be made this week. Inspiration only answers the first by accident. The five sources below answer all three, and the last section shows what happens when the reading, the scoring and the calendar are done for you.

• The first list lasts 12 to 15 videos, then the blank page arrives
• Inspiration is a reservoir and empties; demand is a river and does not
• An idea worth making: someone is looking for it, the channel can answer it, it can be made this week

The richest source of ideas is the channel itself, and almost nobody reads it. Sort your videos by views and look at the median, not the average: the median is what a normal video on this channel does, and any video at three times the median is the audience telling you what they want more of. A channel with a median of 800 views and one video at 4,000 has been handed its next five ideas: the sequel, the opposite case, the beginner version, the advanced version and the mistakes version of that same topic. The guide to finding a channel's outlier video explains the multiplier reading in detail, and it applies to your own channel before anyone else's.

The sequel rule is the single most reliable idea generator on the platform, because an outlier is not a lucky video, it is a proof of demand: the algorithm found an audience for that topic and will look for that audience again when you publish the sequel. Series built on an outlier also solve the second question, whether the channel has earned the right to answer, because the audience already came to this channel for that topic once.

The trap is reading the wrong metric. Views are the outcome; the cause is in average view duration and click-through rate. A video with average views but 60% retention is an idea the audience loved and the packaging failed, so the sequel gets a new thumbnail and the same topic. A video with high views and 25% retention is a thumbnail that oversold, and the sequel needs to deliver what the first one promised. Two videos, two different next moves, and both are visible in fifteen minutes of your own analytics.

• Median, never average: 3 times the median is a signal, 10 times is an anchor
• Every outlier carries five sequels: opposite case, beginner, advanced, mistakes, part two
• Read retention and CTR to know whether the topic or the packaging won

YouTube Studio has a Research tab inside Analytics, labelled Trends in newer versions of Studio, that most creators have never opened. It shows the searches your viewers ran across YouTube, and it flags content gaps: searches with real volume where viewers did not find a satisfying result. A content gap in your niche is an idea that has already passed the first question, someone is looking for it, and has a bonus: the competition for it is weak by definition. Ten minutes in that tab, once a week, is worth more than an afternoon of brainstorming.

The second search source is the autocomplete box. Type your niche's core phrase and a letter, and YouTube completes it with what people actually type, in the order they type it. "How to invest" followed by a, b, c, d gives forty real queries in five minutes, each one a title. Autocomplete does not show volume, and no third party tool has the real number either, but it shows direction and phrasing, and phrasing is what the title needs to match.

The third is the search terms report for your own videos: the exact queries that brought viewers to the channel through YouTube search. When a query brings traffic to a video that only half answers it, the full answer is the next video, and it will rank faster than anything else you make because the channel already ranks for the query. These three reports together produce more ideas per week than a channel can publish, which is the point: the problem stops being finding ideas and becomes choosing between them.

• Research tab: viewer searches and flagged content gaps
• Autocomplete: the phrasing people actually type, letter by letter
• Search terms report: queries you already rank for and only half answer

The channel next door is also running experiments, and its results are public. Every competitor in your niche has a median and a handful of videos far above it, and those outliers are demand signals that cost you nothing to collect. The method of YouTube competitor research is to list ten channels of a size similar to yours or one step ahead, find each one's anchor videos, and extract the topic, not the video. Ten channels, ten anchors, three topics that repeat across them: those three are the strongest ideas in the niche this quarter.

The distinction between modelling and copying is what keeps this source honest and the channel safe. Copying reproduces the title, the structure and the thumbnail; the algorithm already has that video and has no reason to recommend a second one. Modelling takes the proven demand and answers it with this channel's angle, voice and examples, which is a different video the audience has not seen. The competitor proved the topic works; the channel proves it can do it better or differently.

The manual cost of this source is what kills it: an hour per competitor to sort, read and record, ten hours a month for ten channels, and the data is stale in three weeks because new outliers appear every week. This is exactly what Spy does inside FalconVid: it finds the outlier videos in the niche, ranks them by multiplier over the channel's median, and turns the topic into a proposal on your calendar, so the collection runs continuously instead of once in the first month.

• Ten channels, ten anchors, three repeating topics
• Model the demand, never copy the video
• By hand: ten hours a month and stale in three weeks; Spy runs it continuously

A comment that asks a question is a video idea with a guaranteed first viewer. Read the last 200 comments on the channel and sort them into three piles: questions the video did not answer, disagreements with something the video said, and requests for a topic. Each pile is a different kind of next video. The unanswered question becomes the follow-up; the disagreement becomes the debate video, which holds retention because the audience already has an opinion; the request becomes the video with a commenter's name in the first line, which is the cheapest loyalty a channel can buy.

The Community tab turns this from passive reading into active polling. A post with two or three candidate topics and a vote tells you which one to make next before a minute of production is spent, and it warms the audience for the video's first hour, which is the hour that decides its reach. The guide to community posts on a faceless channel covers what to post and how often; for the idea problem, the poll is the tool.

Comments are also the source that scales worst by hand. A channel publishing three videos a week in two languages receives hundreds of comments a month, and the questions are buried among thanks and spam. On FalconVid the comments get automatic replies in the commenter's language, and the questions that recur are exactly what the Senior AI Analyst reads when it looks for the next move, so the pile sorts itself.

• Three piles: unanswered question, disagreement, request
• Community poll: the audience picks the next video before production starts
• Recurring questions are the Analyst's raw material

Some niches never have an idea problem, because the world supplies one every morning: finance, technology, sport, geopolitics, gaming, anything where something happened today. The source for those niches is the feed, RSS feeds from the outlets of the niche, filtered for the topics the channel covers, and the video is the explanation of what happened for the audience that does not read the outlet. The trade-off is the shelf life: a news video earns most of its views in 48 hours, while an evergreen guide earns for years, and the comparison of RSS feeds versus manual topics walks through when each one pays.

The feed also works for evergreen niches in a quieter way. A study published this week, a rule that changed, a product that launched, is the hook that makes an old topic new again: "how to save for retirement" is a reservoir video; "what the new contribution limit changes about how to save for retirement" is a river video with a date on it and a search spike behind it. Reading three feeds for ten minutes a week produces the hooks; the evergreen backbone stays the same.

On FalconVid the feed is a channel setting rather than a chore: connect the RSS sources, set the filters, and the matching stories arrive on the calendar as proposals, with the research done before the script. For a news channel that is the entire idea pipeline; for an evergreen channel it is the weekly injection of timeliness that keeps the library from looking frozen.

• News niches: the feed is the calendar, 48 hour shelf life
• Evergreen niches: the feed supplies the hook, the backbone stays
• RSS as a channel setting: matching stories arrive as calendar proposals

The five sources produce more ideas than any channel can make, which creates the second problem: choosing. The fix is an idea bank with a score. Every idea gets three marks from 1 to 5: demand, from the source that produced it, an outlier at 10 times the median is a 5, an autocomplete phrase with no volume signal is a 2; fit, whether this channel has already proven it can hold that audience; and effort, whether it can be made this week with the research the channel already has. Sort by the sum, and the top of the list is the calendar for the next month.

The bank should hold 30 ideas ahead of the calendar at all times. Thirty is a month of daily publishing or three months at twice a week, and it is the buffer that makes the blank page impossible: the week the creator is sick, travelling or simply tired, the calendar still has a video. Refilling it is the 20 minute weekly ritual: five minutes in your own analytics for outliers, five in the Research tab and the search terms, five on competitor anchors, five on comments, and the feed read in the gaps. Twenty minutes, once a week, and the bank never drops below 30.

This is the ritual the growing channels run and the dying ones skip, and it is the single habit that separates them more than editing quality, thumbnail style or upload frequency. On FalconVid the ritual is compressed further, because the sources are read continuously and the calendar arrives already proposed: the 20 minutes become the approval, not the research.

• Score every idea: demand, fit, effort, each from 1 to 5
• Keep 30 ideas ahead of the calendar at all times
• 20 minutes a week: 5 on outliers, 5 on search, 5 on competitors, 5 on comments

Start from the reading. The AI YouTube analyst inside FalconVid reads the channel's analytics the way this guide describes, outliers against the median, retention against click-through, the search terms and the recurring comment questions, and from Pro upward the Senior AI Analyst, a fixed person for your account with a name, a face and a voice, writes to you every two days with the next move: which topic to sequel, which video's thumbnail to swap, which comment thread became a video. It is the 20 minute ritual done by someone who never skips the week.

Then the proposal. Spy finds the outliers across the niche and the calendar arrives with topics already scored and titled, from research on what is being searched and what the growing channels published this week. You approve, reorder or replace, and the AI specialists produce in parallel: research before the script, so the video answers the question with facts rather than memory, scriptwriter, narrator in 63 languages with ultra realistic premium voices or your cloned voice, scene generation in the channel's style, sound design, thumbnail, video SEO with title, description, tags and chapters, and 9:16 Shorts cut from the long video. A 12 minute video is ready in up to 30 minutes, and several run at once: 2 simultaneous generations on Starter, 5 on Pro, 10 on Business, 25 on Agency and 50 on Scale.

The channel DNA is what makes sequels feel like sequels: the same narrator, the same visual identity and the same tone on video 40 as on video 4, without a decision per video. The Studio lets you watch the first version and shorten the intro, swap a scene or the music, or open the timeline, instead of regenerating. And publishing goes to YouTube, Instagram, TikTok, Rumble and Facebook at the scheduled time, with comment replies in the commenter's language, so the five sources keep feeding the bank while the videos ship.

• The Analyst reads outliers, retention, search terms and comments every two days
• Spy plus research: the calendar arrives scored and titled, you approve
• DNA keeps series consistent; Studio fixes without regenerating

By hand, the honest budget for reading the five sources is two to three hours a week, on top of the 8 to 12 hours a manual video already costs, and that is why the reading is the first thing to be skipped when the week gets full. The channel that skips it publishes from the reservoir until the reservoir is empty, and then stops. The cost is not the hours; it is the channel.

On the line the cost is credits and the hours are the approval. On today's ruler a 12 minute video is 1,731 credits in economy, 4,597 in balanced and 16,131 in premium. Starter includes 15,000 credits, 8 videos of 12 minutes in economy, 1 channel and 2 simultaneous generations: the honest first month is two balanced videos for the sequels of the outliers plus three in economy for the search gaps, 14,387 of the 15,000. Pro includes 30,000 credits, 17 videos in economy or 6 in balanced, 5 channels and 5 at once, plus the dedicated server and the Senior AI Analyst, which is the plan where the reading is done for you every two days.

Business includes 95,000 credits, 54 videos in economy, 10 channels and 10 at once, the plan for a portfolio where each channel runs its own bank. Agency includes 190,000 credits, 109 videos, 25 channels and 25 at once, and Scale 320,000 credits, 184 videos, 50 channels and 50 at once. The free plan makes the first four videos of up to 3 minutes every 30 days with no card, enough to watch a sequel of your own outlier get made, and the annual plan charges 10 months instead of 12.

• By hand: 2 to 3 hours a week of reading, the first thing skipped
• Today's ruler: 1,731 economy, 4,597 balanced, 16,131 premium for 12 minutes
• Starter: 8 in economy, or 2 balanced plus 3 economy, 14,387 of 15,000
• Pro: the Senior AI Analyst reads the five sources every two days
• Free plan: 4 videos of up to 3 minutes every 30 days, no card