Two sibling sites, 176 and 129 million impressions: the smaller one gets twice the clicks

I operate four recipe sites arranged in two pairs. Each pair is the same project in two languages: one site in English and its sibling in Spanish, same subject, same people behind it, same age. That makes these pairs something you almost never get to measure cleanly, because the one large variable that changes between siblings is the language. Twelve months of numbers say something I did not expect, and it repeats in both pairs.

Quick answer

• The Spanish half converts about three times better. CTR of 3.2% against 1.2% in the large pair, and 1.8% against 0.5% in the small one.
• Fewer impressions, more clicks: the Spanish site in the large pair has 27% fewer impressions than its English sibling and 93% more clicks.
• The small pair controls the variable by accident: nearly identical impressions, a worse average position in Spanish, and still 3.5 times the clicks.
• Why it happens, I do not know. Four candidate explanations, and these numbers cannot separate them.

That is the whole argument in one picture: the top panel is essentially flat and the bottom one multiplies by three and a half. Same exposure, different outcome.

Is Spanish SEO worth it compared to English?

These numbers answer whether a Spanish half pays off once it exists. The prior question, which pages of a site earn a Spanish version at all and why translating everything is the common mistake, is the subject of Spanish SEO.

In two pairs of sibling sites I operate, the Spanish half turns impressions into clicks about three times better than the English half at comparable positions. For a US audience that is entirely Hispanic, Spanish wins outright. For a mixed audience, measure CTR per language before splitting budget. For a global niche, English still delivers far more impressions.

What kind of comparison this is

The dominant mental model is that English is the big market and Spanish is the small one, so a bilingual project splits its effort in that proportion. Under that idea the deciding metric is search volume, and English wins every time.

These numbers do not contradict the volume part. They contradict the outcome. The English site in the large pair generates 47 million more impressions than its sibling and two million fewer clicks. If effort had been allocated by impression volume, more of it would have gone to the half that performs worse.

What makes this uncommon is not the size of the numbers, it is the comparison. Almost everything published about multilingual SEO compares different sites with different owners, and then language is one variable among twenty. Here they are siblings.

What the US search results say about this

I searched this before writing, and the first page answers a nearby question rather than this one. It is mostly market sizing: more than 62 million people in the United States identify as Hispanic, a large share of US internet users move between both languages, and most online shoppers prefer to buy in their first language. All true and all demographic.

The performance numbers that do circulate come from paid search, not organic. The one most often repeated is a Google study on car-part queries showing a cost per click of $0.79 in Spanish against $2.65 in English, roughly 3.5 times cheaper, alongside a 5.7% uplift in clicks for Spanish results. I could not trace that 5.7% back to a primary source I can link, so it does not get used as evidence here.

It also would not answer the question even if it were solid. Paid search is an auction: cheaper Spanish clicks mean fewer advertisers bidding, which is a statement about competition, not about how people behave with organic results. What is missing on that first page is an organic comparison between two halves of the same project, which is what these four properties happen to be.

The small pair is the one that counts

Big numbers impress and small numbers prove. In the large pair the Spanish site has fewer impressions, so you could object that it compares two different populations and that CTR rises because the query mix is more specific.

The small pair removes that objection on its own:

• Nearly identical impressions: about 930,000 in English and 910,000 in Spanish, a 2% difference.
• The average position is worse in Spanish: 13.1 against 12.2. The better-performing half sits lower on the page.
• Clicks are 3.5 times higher: roughly 16,400 against 4,700.

Same exposure, a worse spot on the page, and three and a half times the outcome. That combination is not explained by query mix, because mix would move impressions too, and the impressions are the same. Reading these four columns without drawing more than they support is covered in how to set up Search Console.

When this does not apply

When the business sells in English. Better CTR over an audience that does not buy is not an advantage, and the deciding metric becomes conversion, not the click.

When the niche is technical and its vocabulary was born in English. Spanish speakers often search those terms in English, and the two halves stop being comparable.

When the Spanish site is a machine translation of the English one. These pairs are not, and that difference probably explains part of the result. What that costs is the subject of translating a website for SEO.

When there is only one pair. With two agreeing I am willing to publish it; with one I would have left it as an anecdote.

The four candidate explanations

This is where the article parts company with most of what gets published. I do not know why this happens, and these numbers cannot decide it. The four I consider:

1. The English results page is more crowded. Generative summaries, ads, carousels and rich results occupy the screen before the first blue link, and in English they arrived earlier and across more queries. Position 5.8 in English can sit considerably lower on an actual screen than 5.4 in Spanish.

2. Competition for the click is different. English pits you against large media brands people recognize. In Spanish the neighborhood is thinner and an unfamiliar result gets a fairer hearing.

3. The query mix differs. The Spanish half may receive more specific queries, where the searcher knows what they want. Working against this explanation is the small pair, which has the same impressions.

4. Average position is an average. This is the heaviest objection and it goes last on purpose. An average moves on its own when the mix changes, so "comparable positions" is a weaker claim than it looks. Closing it would take a fixed-cohort comparison of queries, and I have not run one.

All four are compatible. The likeliest answer is several at once, and anyone reading this with two halves of their own can check whether it happens to them too.

Mistakes that repeat

• Allocating effort by search volume. It is the decision these numbers contradict, and it usually gets made before anyone measures anything.
• Translating keywords instead of researching them. The equivalent query in the other language is rarely the literal translation, which is the point of free keyword research.
• Comparing sites with different owners. With two unrelated projects, language is one variable among many and the conclusion does not hold.
• Quoting paid-search economics as if they were organic. A cheaper click in an auction says something about advertisers, not about searchers.
• Watching impressions only. It is the column that rises the most and says the least about whether anyone arrived.
• Assuming the Hispanic market is the small one. In these two cases it is the one producing more clicks, and what is particular about it is in SEO for Latino businesses in the US.

Data and transparency

Figures come from Search Console, pulled through the API on August 10, 2026, with a read-only credential, over a twelve-month window.

All four sites are client work, so they are not named and the figures are rounded. They are described by type, which is recipe publishing, and by their relationship to each other, which is the only part the argument needs. The accurate description of my role is that I operate them, not that I own them.

Google documents a logging error covering most of this window. Its Data Anomalies page states that a bug kept impressions from being recorded accurately between May 13, 2025 and April 27, 2026, affecting impressions, CTR and average position, and not clicks. The sibling comparison survives it, because both halves of each pair lived through the same error and the bias applies equally, and clicks are the untouched column: 2.1 against 4.1 million, and 4,700 against 16,400. What should not be taken at face value are the absolute impressions and the exact CTR. Checked at the source on August 13, 2026.

Position here is an average, not a median. The twelve-month property summary only reports it that way. A weighted median would be the better comparison and I do not have one for this window, which is what supports the fourth objection above.

There is no proof of cause. This article measures a difference and lists hypotheses. Anyone citing these numbers as proof that Spanish converts better is claiming more than was measured here.

What this changes

The conclusion I take is not that Spanish converts better. It is that I had been splitting effort by impression volume for years without ever comparing the two halves, with all four properties in the same account and the comparison two queries away.

That is what makes this case uncomfortable to write. It did not need an expensive tool or a new methodology. It needed putting four rows in the same table, which is a five-minute operation nobody performs because each property gets looked at on its own.

And the last objection is the one worth keeping. With two pairs agreeing, the difference is measured and not explained, and the temptation to fill that gap with a convincing cause is exactly the mechanism that produces the unsourced numbers this blog spends its time taking apart. An honest measurement with a hole in it beats a tidy explanation I cannot defend.

Frequently asked questions