Across 95,086 Google queries over seven and a half months, questions shaped like “what is the difference between X and Y” earned clicks at 0.85%, while lookups shaped like “X in Spanish” earned 0.12%. That is a seven-fold gap between two kinds of demand that a keyword tool would price almost identically.
We teach Spanish, so the queries here are Spanish queries. The shape of the finding is not about Spanish. It is about what happens to a search when the answer fits in a box, and it decides which pages are still worth writing.
Key takeaways
- Comparison queries converted at 0.85%. Translation queries converted at 0.12%, a ratio of 6.9 to 1.
- Comparisons are 2.1% of our impressions and 11.8% of our clicks.
- Translation lookups are the largest block of impressions we have, 2,214,930 of them, and they returned 2,701 clicks.
- The same split shows up where nobody is searching: on our YouTube channel, the four most-commented videos of all time are all comparison videos, carrying 28.8% of every viewer comment we have ever received.
- In the 90 days to September 18, 2026, the gap was wider still: 0.85% against 0.10%.
The short version
Write the comparison, skip the translation. When a query can be answered by a single word, the search engine answers it and keeps the visit. When a query requires a judgment between two options that both look correct, the searcher still comes to read a person’s explanation. Query volume is not the number to optimize toward. Query shape is.
Why this question does not have an answer yet
The published research measures how many clicks disappeared. It does not say which questions keep theirs. SparkToro reports that 68.01% of US Google searches ended without a click in early 2026, and Ahrefs reports a 58% lower click-through rate for top-ranking pages when an AI Overview is present, from 300,000 keywords in December 2025. Both are averages across every kind of query at once.
For anyone deciding what to write next week, the average is the wrong number. What you need to know is which shape of question still sends a reader, and that requires one site’s own queries sorted by shape and measured against each other. That is what this study is: 95,086 queries from a single property, split five ways, with the click rate of each group next to the others.
The five query shapes, measured
We sorted every named query into one of five shapes using fixed patterns, then measured each group. The sort happens in code, the patterns are published below, and a query is tested against comparison first so that “difference between por and para” lands as a comparison rather than as a question.
| Query shape | Example | Queries | Impressions | Clicks | CTR |
|---|---|---|---|---|---|
| Comparison | tu vs su | 2,153 | 150,566 | 1,276 | 0.85% |
| Meaning | nada meaning | 10,468 | 1,310,033 | 2,005 | 0.15% |
| Question | what does orale mean | 16,932 | 1,021,268 | 1,495 | 0.15% |
| Other | no mames | 49,342 | 2,328,116 | 3,282 | 0.14% |
| Translation | morning in Spanish | 16,187 | 2,214,930 | 2,701 | 0.12% |
The volume trap
Translation queries are our biggest impression source and our worst click source. They brought 2,214,930 impressions and returned 2,701 clicks. Comparisons brought 150,566 impressions, about a fifteenth as many, and returned 1,276 clicks, about half as many. Per impression, a comparison is worth seven of them.
This is why keyword volume alone is a poor guide to what to write. A research tool will tell you that “morning in spanish” has enormous demand, and it does. What it will not tell you is that the demand is already satisfied above the results, so ranking for it buys visibility that never converts into a reader.
The same pattern shows up where nobody is searching
Our YouTube comments agree with our Search Console, and the two datasets have nothing to do with each other. We pulled every comment on the channel on September 20, 2026: 437 viewer comments across 157 public videos. The four videos that collected the most comments in nine years are all comparisons.
- “What’s the Difference Between Tu and Su in Spanish?” 42 comments
- “Mexican Slang: Prieto and Guero in Spanish” 34 comments
- “What’s the Difference Between No Problemo and No Problema?” 29 comments
- “What’s the Difference Between Morning and Tomorrow in Spanish?” 21 comments
Those four carry 28.8% of every viewer comment the channel has ever received. People argue about the pair. Nobody argues about a translation.
Our classroom says it too
In 113 one-to-one class sessions we coded from January 2025 to September 2026, the recurring sticking points are nearly all pairs. Preterite against imperfect appeared in the sessions of eight different students. Ser against estar, tu against usted, saber against conocer, por against para: the questions adults bring to a paid teacher are the same shape as the questions that still earn a click on Google. The full taxonomy is in the class corpus study.
What we do with this
Three rules came out of it, and we now apply them before any new post gets written. Ask whether the answer fits in one line, and if it does, write it as a section inside a larger piece rather than as its own page. Prefer the pair over the word, because a pair needs a person. Publish the reasoning, not just the verdict, because the reasoning is the part an answer box cannot copy without sending someone to read it.
How we counted this
Source: Google Search Console API, domain property sc-domain:speakbetterspanish.com, pulled September 20, 2026. YouTube Data API for the comment corpus, pulled the same day.
Window: February 4 to September 18, 2026, the full history available for this property. 227 days, 95,086 named queries, 7,025,331 impressions.
Classification: each query is tested in a fixed order against comparison (“vs”, “versus”, “difference”, “which”), then question (opens with what, how, why, when, where, which, who, is, are, do, does, did, can, should, could, will), then meaning (“meaning”, “means”, “definition”, “translate”), then translation (“in spanish”, “en espanol”, “spanish word for”), and anything left is other. Brand queries are excluded from the table. A query lands in exactly one shape.
Limits: one site in one niche. The classifier is crude by design so that it can be rerun and checked. Any pattern-based sort puts some queries in the wrong bucket, and the ratio moves by a few tenths of a point depending on the order the tests run. It does not move enough to change the finding: the gap between comparison and translation is sevenfold at the year scale and close to ninefold in the most recent quarter.
Data: query shape table (CSV) and comment buckets (CSV). Free to reuse with attribution.
Questions people ask about this data
Why do comparison queries get more clicks?
Our data shows the gap, not the cause. The most plausible reading is that a one-word answer can be printed on the results page while a comparison needs context, examples and a judgment about which option a real speaker would use. A searcher who wants to know whether to say tu or su has to read something. A searcher who wants the Spanish for “morning” does not.
Does this mean translation content is worthless?
No. It means translation pages should not be built as standalone posts chasing volume. We keep ours, because they rank and they feed internal links, and we stopped adding new ones. The effort moved to pairs and to questions where a native speaker’s judgment is the whole answer.
Is 0.85% a good click-through rate?
It is poor by any historical benchmark and it is seven times better than the alternative on the same site in the same period. Both things are true. The useful comparison is not against an industry average, it is against the other shapes of query that the same site could target with the same amount of work.
How would I run this on my own site?
Pull your Search Console queries with the API rather than the interface, so you get tens of thousands of rows instead of the top thousand. Sort them by shape with simple patterns, then compare clicks per impression by group rather than by page. The whole analysis is a short script. Ours is described above and the output tables are published.
References
- Fishkin, R. In 2026, Less than One Third of Google Searches Still Send a Click. SparkToro, using Similarweb US clickstream data, January to April 2026.
- Law, R. and Guan, X. AI Overviews Reduce Clicks by 58%. Ahrefs, 300,000 keywords, December 2025.
- Speak Better Spanish. Click-through rate by query shape, 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, with the query set, the YouTube corpus and the class corpus all pulled the same day.
Recent Comments