Nobody can tell you how many people search a term on YouTube, and every tool that shows a number is guessing

Almost everyone who searches for YouTube keyword research wants one specific thing: a monthly number next to a phrase. How many people search how to start a faceless channel on YouTube every month. That number is not published. Not inside YouTube Studio, not in the public API, not to any third party tool, not to agencies, not to anyone.

Every tool that shows you 12,400 searches a month for a YouTube term is showing you a model. The usual ingredients are Google Search volume from the Keyword Planner, click behaviour on Google results that lead to YouTube, autocomplete ordering, and extrapolation from a panel. Some of those models are decent at rank order. None of them is a measurement, and the vendors that are honest say so in their own documentation.

This matters commercially, not academically. Chasing a precise number costs weeks that produce zero videos, and the decision you actually have to make, which is what to record this week, never needed the exact figure. It needs three things: rank order between candidates, evidence that demand is not already served, and proof that the audience exists at all. All three are available for free, and the rest of this post is where.

There is one first party source of YouTube search behaviour and it lives inside your own account: YouTube Studio, Analytics, the Trends tab, which older guides still call the Research tab. It shows four cards. Top searches, based on your audience and your saves. Breakout videos, which are videos performing unusually well for creators similar to you. Recent videos, tied to what your viewers watched lately. And content gaps for Shorts, which are search terms where viewers cannot find enough quality results.

Now the limits, because they are what nobody writes down. Audience interest is expressed on a qualitative scale that runs from very low to very high. There is no number anywhere in the panel. The window is the last 28 days, built on videos watched more than 1,000 times a week in that period. And the YouTube help page itself states that some of the insights are limited to certain countries, languages and devices, without listing which ones. Since July 2026 Studio has been rolling out an updated experience, so the tab may not look the way a 2024 tutorial shows it.

Read honestly, that panel is a direction finder and not a measuring tape. It is still the most valuable free input you have, because it is the only one that comes from the platform where the video will actually live. If you run channels in Portuguese or Spanish and the panel looks thin, that is not your account being broken, it is the coverage limitation talking.

This is also where the gap between knowing and doing opens up. The panel tells you a topic is underserved this month, and then you still have nine hours of work per video ahead of you. FalconVid closes that gap from the other end: from Pro upwards a Senior AI Analyst with a fixed name, face and voice reads your account every two days and writes, in your language, the specific move to make next inside the product.

Google Trends has a YouTube Search filter that is separate from Web Search, and most creators never switch it on. Switching it on changes the answer, because a term can be enormous on Google and dead on YouTube. Nobody opens YouTube to search for cheap flights to lisbon, and a lot of people open YouTube to search for how to tie a tie who would never type it into Google.

The output is a relative index from 0 to 100. The 100 is the peak of that particular series, in that window, in that country. It is not a volume and it never becomes one. Two terms compared side by side tell you which is bigger than the other in relative terms and nothing about how big either is. Treating the index as a number of searches is the single most common mistake in this whole subject.

Where it earns its keep is shape over time. A term sitting at 12 in May and 78 in December is a fourth quarter topic, and that is a calendar decision rather than a keyword decision. A term with a flat line across three years is evergreen, which is the kind that keeps paying eighteen months after upload. We measured that difference in views and revenue in evergreen topics against news topics after twelve months, and the two curves are not close.

Google Ads Keyword Planner does publish monthly search volumes, and those volumes are for Google Search. Using them as YouTube volumes is using the right instrument on the wrong patient. They are still useful for one job: mapping the family of a topic, so you can see that a subject has ten adjacent questions rather than one. Take the shape, leave the number.

YouTube autocomplete is the opposite trade. Zero volume attached, but every suggestion is a phrase real people typed into the box, ordered by YouTube's own signals. It costs nothing and takes two minutes: type your topic, then add each letter from a to z after it, and copy every suggestion that is a real question rather than a brand name. That list is the best source of title phrasing that exists, because it is the exact wording of the demand.

The working combination is boring and effective. Keyword Planner for the family of the topic, autocomplete for the phrasing of the title, the Trends tab for whether YouTube viewers specifically want it, and Google Trends for whether the interest is rising, flat or seasonal. Four free inputs, none of which is a volume, and together they beat any single paid number.

The first is the content gap. That is the only moment a platform tells you directly that demand exceeds supply, and it is worth more than any estimated volume, because volume without a gap just means you are joining a queue. YouTube surfaces it explicitly for Shorts, and the same logic applies manually to long form: search the term and look for the moment where the top results are either off topic or four years old.

The second is the breakout video. A video outperforming what its own channel normally does is a demand signal with a date stamped on it, and it is much sharper than a yearly average. Reading them properly, including what to copy and what to ignore, is a skill in itself, and we broke it down in what to take from a competitor and what to leave.

The third is revealed demand, which is simply the view counts already sitting on the results page. If a term has six videos above 200,000 views published in the last eighteen months, the audience is proven. If the best result in three years is 12,000 views, the audience is not there, no matter what any tool estimates. An estimated volume is a prediction. A view count is a receipt.

This is exactly the job the FalconVid Spy does, and the Chrome extension does it on the YouTube page itself while you browse: it reads the channels already monetising a topic, the rhythm they publish at and the outliers inside their catalogue. The house phrase for it is blunt on purpose, you do not start from zero, you start from what already monetises. If you are choosing a subject rather than a single video, the longer method is in how to find a niche that travels before you record anything.

Even a perfect keyword only buys the search slice of your views. On an established channel the majority of watch time arrives from browse and from suggested, which are recommendation surfaces and not search. We measured how the three behave differently, and what each one rewards, in where YouTube views actually come from.

The practical consequence changes what you research for. A topic chosen only for search demand has a ceiling built into it. The topic that wins is the one a person can search for today and that the system can also place next to a video somebody just finished watching. That is why breakout videos matter more than volume: they prove the recommendation side already accepted the subject.

Then there is the arithmetic that ends the argument. From 1 February 2027 a new channel needs 1,000 subscribers plus 8,000 qualified public watch hours in 365 days, double the old 4,000. Eight thousand hours is 480,000 minutes. A twelve minute video watched at 40 percent delivers 4.8 minutes per view, so the door is 100,000 views. At a thousand views per video that is a hundred videos, and at nine and a half to thirteen and a half hours of work each that is 950 to 1,350 hours by hand.

Nobody researches their way through that. This is where the parallel side of FalconVid stops being marketing: five AI specialists, researcher, scriptwriter, narrator, editor and sound design, work on the same video at the same time, a finished video lands in up to 30 minutes, and the plan decides how many run at once, 2 on Starter and 50 on Scale. The same hundred videos cost 100,800 credits on economy mode, about US$ 316.

None of the above needs a subscription, and all of it fits into one sitting per week. The order matters, because each step narrows the list before the expensive step runs.

Run it once and you will notice something uncomfortable: the research finishes long before the production does. That asymmetry is the whole reason this niche is full of people with a spreadsheet of forty validated topics and eleven published videos.

• Studio, Analytics, Trends tab. Write down every content gap and every breakout video you see. Five minutes.
• YouTube search box: your topic, then your topic plus each letter from a to z. Copy the ten suggestions that are real questions. Five minutes.
• Google Trends with the YouTube Search filter, your country, twelve months. Mark which of the ten are rising and which are seasonal. Five minutes.
• Search your three finalists on YouTube and read the view counts and the dates of the top twelve results. Nothing above 50,000 views in eighteen months means the demand is not there. Five minutes.
• Publish, and let the video answer. One published video returns more usable data in seven days than a month of research does.

Research has a hard ceiling and publishing does not, which means the bottleneck was never the keyword. It is everything that happens after you decide: the research pass, the script, the narration, the roughly ninety shots a twelve minute video needs at a cut every six to ten seconds, the thumbnail, the YouTube title, description and tags, and getting it out on a date.

That whole list is what FalconVid runs. You approve the calendar, not each individual step, and the videos come out on it. Inside the video, research plus script is a fixed 307 credits and it is identical in all three quality modes, which is the honest way to see how small the writing is next to the visual generation. The same twelve minute video costs 1,008 credits on economy, 8,676 on balanced and 26,760 on premium, and the mode is chosen per video, on every plan.

Around it sits the rest: the Spy and the Chrome extension for the research above, the channel DNA that keeps identity consistent across videos, a calendar that extends itself so the channel never runs dry, publishing to YouTube, Instagram, TikTok, Rumble and Facebook, 63 narration languages, duplicating a finished project into another language paying only the difference, and the Studio if you want to shorten an intro or swap a shot instead of regenerating. From Pro upwards the Senior AI Analyst writes every two days with the next move.

Keyword research tells you what the audience wants this month. It never tells you how to get a hundred videos out before February 2027. That second half is the one that is actually for sale, and you can see the whole machine on the FalconVid home page or read what the analyst does on the AI channel analytics assistant page.