We compared every comment ever left on our YouTube channel, 437 of them, against 95,086 Google queries that reached our site, and the two audiences want different things. Search brings people looking up a word. Comments come from people arguing about whether a word is rude.
Both datasets belong to the same small Spanish school and were pulled on the same day, which is what makes the comparison worth anything. Nobody usually gets to see the same learners behave in two places at once.
Key takeaways
- 437 viewer comments across 157 public videos, against 456 replies from us. The channel has 280,152 lifetime views and 2,980 subscribers.
- Praise is the most common comment type at 33.9%. Questions are second at 14.2%.
- 6.4% of comments argue about regional usage and 4.6% correct us or another commenter, usually as native or heritage speakers.
- Four videos, all of them comparisons, hold 28.8% of every comment the channel has received.
- The grammar that fills paid classes barely shows up in search. Queries about the subjunctive brought us 92 impressions in seven months. Queries about Mexican slang brought 709,187.
The short version
Public demand and private difficulty are not the same thing. What learners type into a search engine is overwhelmingly vocabulary: one word, one meaning, one answer. What they stop to write a comment about is judgment: which word a real Mexican would use, whether a phrase is offensive, whether their teacher was right. What they actually pay to fix, which is the subject of our class corpus study, is a third list again.
Two datasets that usually never meet
Search data and audience data are normally collected by different people for different reasons, so nobody puts them side by side. Keyword tools report what gets typed. Social analytics report what gets engagement. Neither tells a teacher what a learner is actually stuck on, and the gap matters because search demand is now a weaker signal than it used to be: SparkToro found that 68.01% of US Google searches ended without a click in early 2026, so the questions people type are increasingly answered before anyone reaches a teacher.
This study joins three sources that belong to the same small school and were pulled on the same day: every comment the YouTube channel has received, every query Search Console reports for the website, and the coded record of 113 paid one-to-one classes. Same audience, three different behaviors, three different lists.
What is in the comment corpus
We pulled every comment thread on the channel through the YouTube Data API on September 20, 2026. That returned 893 records across 157 public videos: 456 of them are our own replies, leaving 437 viewer comments, of which 376 are top-level comments and 61 are replies. Fifty-nine arrived in the last twelve months. This is nine years of a small channel, and it is small: the whole corpus is the size of a single thread on a large channel.
| Comment type | Comments | Share |
|---|---|---|
| Praise for the explanation or the teacher | 148 | 33.9% |
| A question | 62 | 14.2% |
| Regional usage (“here in Tamaulipas we say”) | 28 | 6.4% |
| Comparison or stated confusion | 25 | 5.7% |
| Pronunciation and accent | 21 | 4.8% |
| A correction, often from a native speaker | 20 | 4.6% |
| Personal context (a spouse, patients, Duolingo, moving) | 18 | 4.1% |
| A request for a specific video | 5 | 1.1% |
A comment can fall into more than one group, so the column does not sum to 437. The buckets are keyword patterns, published in the method box, not human judgment calls.
The comments cluster on four videos
Four videos hold 126 of the 437 comments, and all four ask what the difference is between two things. Tu and su collected 42 comments. Prieto and guero, two words about skin tone that Mexicans use casually and learners find alarming, collected 34. No problemo against no problema collected 29, and morning against tomorrow collected 21.
Nothing else on the channel comes close. A video that explains a single word gets a thank you. A video that tells you which of two words to use gets an argument, and the argument is the engagement.
Native speakers show up to police the answer
Twenty of the 437 comments are corrections, and they concentrate on context rather than grammar. The recurring subject is whether a word is acceptable in public, which is exactly the question a textbook cannot answer and a dictionary will not touch. One commenter from Mexico wrote, in Spanish, that anyone learning with us is doing it right and that the pronunciation is genuinely good. Another asked, plainly, whether gringo is an ethnic slur.
For a teaching channel this is a feature. The corrections arrive in public, on the page, where the next learner reads them.
Where search and the classroom disagree
The topics that dominate our search traffic are vocabulary, and the topics that dominate paid classes are not searched for at all. Over the same seven and a half months, queries about Mexican slang words brought 709,187 impressions to the site. Queries containing “subjunctive” brought 92.
| Topic | Search queries | Impressions | Clicks | CTR | YouTube comments |
|---|---|---|---|---|---|
| Mexican slang words | 1,580 | 709,187 | 2,480 | 0.35% | 15 |
| Time words (mañana, ahorita) | 1,332 | 306,559 | 73 | 0.02% | 13 |
| The me gusta family | 1,655 | 298,050 | 294 | 0.10% | 13 |
| Tu, su and usted as a pair | 120 | 31,830 | 272 | 0.86% | 25 |
| Por against para | 126 | 1,033 | 0 | 0.00% | 0 |
| Ser against estar | 23 | 453 | 0 | 0.00% | 9 |
| The subjunctive | 25 | 92 | 0 | 0.00% | 1 |
What this means if you teach or write for learners
Search volume measures curiosity, not need. A learner who cannot hold a conversation still searches for single words, because a single word is the only question small enough to type. The gap between what gets searched and what gets taught is not evidence that learners are wrong about their own needs. It is evidence that search is a poor instrument for finding out what they are.
For anyone building content in this space, the practical version is short. Use search demand to decide which door people walk through, and use classroom evidence to decide what is in the room.
How we counted this
Sources: YouTube Data API, read-only, for the Speak Better Spanish channel, pulled September 20, 2026. Google Search Console API for speakbetterspanish.com, February 4 to September 18, 2026. Class figures come from 113 coded one-to-one class summaries, January 2025 to September 2026.
Comment corpus: the full uploads list is 878 items, of which 283 are public. 157 public videos carry at least one comment. We pulled every comment thread with replies: 893 records, 456 posted by the channel and 437 by viewers.
Classification: each comment is tested against fixed keyword patterns for praise, questions, regional usage, comparison or confusion, pronunciation, corrections, personal context and content requests. A comment can match several. This is a tally, not a reading, and it will misfile some comments in both directions.
Search topics: each topic is a pattern applied to the query set, for example the slang row matches gringo, guero, prieto, michi, no mames, no manches, chido and wey. Impressions and clicks are Google’s counts for our property only, so they measure demand that reached us, not total demand.
Limits: 437 comments over nine years is a small corpus from a channel with 2,980 subscribers, and YouTube comments skew toward people with strong opinions. Nothing here is a survey and none of it is representative of Spanish learners in general.
Data: comment buckets (CSV). Free to reuse with attribution.
Questions people ask about this data
What do Spanish learners ask about most?
In public, on video, they ask which of two similar words to use, and they argue about whether a word is polite. In search, they look up single words and short phrases: how to say a word in Spanish, what one word means. The two behaviors come from the same people and produce different lists.
Why is there so little search demand for the subjunctive?
Our data shows the absence, not the reason. One likely explanation is that learners who need the subjunctive are already working with a teacher or a textbook, and learners who do not know they need it have no words to search with. Twenty-five queries containing the word reached us in seven and a half months, for 92 impressions total.
Are YouTube comments a useful research source?
They are useful for one thing: finding out what people disagree with. Our corrections and context arguments concentrate on exactly the points where textbooks and real Mexican usage part ways, which is harder to find in any other dataset. They are useless for measuring how common a problem is, because commenting is rare and unrepresentative.
Can I reuse these numbers?
Yes, with attribution to Speak Better Spanish and a link to this page. The comment bucket table is published as a CSV above, and questions go to hannah@speakbetterspanish.com.
References
- Fishkin, R. In 2026, Less than One Third of Google Searches Still Send a Click. SparkToro, January to April 2026.
- Speak Better Spanish. YouTube comment types, September 20, 2026.
Latest updates
- September 20, 2026, later the same day: Added the background section, the outside references and the data visuals. Figures unchanged.
- September 20, 2026: First publication, from a full comment pull and a same-day Search Console pull.
Recent Comments