Spanish SEO services: the two things that go wrong before you hire anyone

A page I audited ranked in position 1 for its target query and received zero clicks. The translation was accurate. The problem was that the title said "chicken stew" while the people searching typed "sudado de pollo recipe", and nobody recognized their own dish in the result. That page is the entire argument of this article: a correct translation and a working Spanish page are different things, and the gap between them is where most bilingual SEO budgets disappear.

What is Spanish SEO?

By treating Spanish as a separate market rather than a translated copy. A new site should pick one language and do it properly; a mid-authority site adds the second with its own keyword research; a consolidated one runs both with hreflang declared, which only works when every pair points back at the other.

Quick answer

• Translation is not the strategy. Accurate words nobody searches for rank for nothing.
• Spanish is not one market. The terms differ between the US, Mexico, Colombia and Spain.
• Spoken language does not predict search language. The same person switches by topic.
• Separate pages per language, connected with reciprocal hreflang. A half-declared pair is worse than none.

What kind of problem this is

The dominant assumption is that this is a language problem with a language solution: hire a translator, publish, done. That assumption is comfortable and it produces the failure described in the opening.

The reason it fails is that search is not a translation of speech. A person choosing words to type is not choosing the words they would say. They are choosing the shortest string they believe will produce the result they want, and that string is often a loanword, a mix, or a term that would sound wrong out loud.

So this is a research problem before it is a writing problem, and the research has to happen in each language separately. Translating the keyword list is the same mistake one level up.

Three things the data shows and intuition does not

Spanish is several markets, not one

An autocomplete harvest across Colombia, Spain and Mexico produced 14,715 unique suggestions from 11,220 seed queries, of which 8,875 survived cleaning. The suggestions differ by country for the same underlying need, and a term that is standard in Bogotá can be unfamiliar in Mexico City.

For a business serving Spanish speakers inside the United States, this matters more, not less: that audience is not a single origin. Harvesting with es-es data and publishing it in Miami is a specific way to be precisely wrong.

The same person switches language by topic

A second-generation owner may speak Spanish at home and search in English out of habit. A recent arrival may search in Spanish for personal services and in English for anything involving paperwork or law.

That is why the language question cannot be answered from demographics. It is answered from Search Console, which lists the queries a site already appears for in both languages, and it is answered per topic rather than per audience.

Accuracy and findability are independent

This is the lesson of the opening case, and it generalizes. A professional translator optimizes for accuracy, which is their job. Search rewards recognition, which is a different property.

The check costs nothing: search the translated term yourself and see whether the results look like your business. If the page one results are about something else, the term is not the one your market uses. The broader diagnostic for this pattern is in the Spanish guide on ranking without clicks.

How to do it in the right order

1. Export your existing queries by language and country. Search Console already knows which ones you appear for.
2. Harvest autocomplete separately per market, using seeds written natively in each language rather than translated ones.
3. Check what Google shows for each target query. If the top results are in English for a Spanish query, that audience accepts English answers.
4. Write per language and per intent, not per translated page. Some topics need both versions, many need only one.
5. Connect equivalents with reciprocal hreflang, and only when they genuinely are equivalents.

That last point has a rule worth stating plainly. If page A declares page B and B does not declare A, Google discards the entire signal rather than half of it. Both directions or neither. The mechanics of the rest of the research process are in the Spanish guide on free keyword research.

When NOT to publish in both languages

When one language clearly dominates your customer base and you know it from data, not assumption. One excellent page beats two mediocre ones.

When you cannot maintain both. A Spanish version that goes stale while the English one updates becomes a liability, and it is visible to anyone who reads both.

When the budget covers one properly. Splitting it evenly between two half-done versions is the most common way to end up with neither working.

What to expect

That third row is worth dwelling on. If a translated site already exists, the highest-return work is usually not rewriting it: it is checking which titles contain the words the audience actually types, and fixing those. In the audit that opens this article, that single change was the most profitable work in the whole project.

The two things that go wrong, and only one is about language

Bilingual projects fail in two places, and the first one is not the one anyone budgets for.

The pairing. Declaring two versions of a page requires each one to point back at the other, and a single error in a cluster makes the whole cluster get ignored. Half a signal is not half a benefit, it is none. That is why a site with one language half-declared is in a worse position than a site that declared nothing, and why the pairing is a build-time check rather than a launch task.

And the maintenance. Every price change, every new service and every correction now has two homes. The half that goes stale is almost always the second language, because the person writing lives in the first one. A page that contradicts the current offer does more damage than no page.

Neither of those is a translation problem, and both are the reason most bilingual sites end up with one working half.

What the second language actually requires

Its own keyword research, because the words are not the translations. The method is in free keyword research, applied twice rather than once.

Its own pages, not machine-translated copies, with the reasoning and the policy limits in translating a website for SEO.

And a decision about the local side, because a bilingual audience still searches by proximity: local SEO for small business and Google Business Profile.

The market-level picture, with which sectors this changes most, is in SEO for Latino businesses in the USA. What the whole job costs is in SEO cost for a small business, and how to judge a provider proposing it, in how to choose an SEO agency.

The first page is mostly one company

Checked on 13 August 2026, every result for this query is an agency landing page, which is expected for a commercial term. What is less expected is the concentration: one firm holds half of the first page across four different domains, including a separate site built around the same brand name. The page reads like a market and is closer to one estate.

The second thing they share is what they sell on. Cultural nuance, native speakers, Spanish keyword research, resonating with the community. All of that is real work and none of it is the part that decides whether the Spanish version of a US site earns anything.

That part is structural and testable, and it is missing from every one of those pages: whether the two languages live on separate URLs, whether they declare each other, and whether the Spanish version is a full page or a translated shell hanging off the English one. A site that gets those wrong can hire the best writer in the market and still have both versions competing for the same query.

It is not hard to see why the pages are written that way. Cultural fit is easy to promise and impossible to fail out loud, and a hreflang error shows up in a report with a date on it.

Mistakes that repeat

• Translating the keyword list. It is the original mistake one level up.
• Treating Spanish as one market. The terms differ by country and the differences matter.
• Inferring search language from spoken language. The same person switches by topic.
• One page serving both languages. It splinters the signal and reads badly to everyone.
• Declaring hreflang in one direction. Google discards the whole signal, not half.
• Rewriting pages when the titles were the problem. Far more expensive, far less effective.

Data and transparency

That every top result for this query is an agency landing page, that one firm accounts for about half of them across four domains, and that none of the pages reviewed mentions URL structure or language annotation, is an own observation from 13 August 2026 and is recorded as a pattern in the results, not as a source.

That hreflang requires reciprocal return tags and that a single error in a cluster causes the whole cluster to be ignored is documented behaviour, checked on 12 August 2026. No figure is given here for how many implementations are faulty: the ones circulating are attributed only to unnamed studies.

The autocomplete harvest figures (11,220 seed queries, 14,715 unique suggestions, 8,875 useful after cleaning, across Colombia, Spain and Mexico) come from my own research conducted in August 2026. The 312-page recipe site with a position-1 page receiving zero clicks is client work: it is not named, and its figures are rounded. The reciprocity requirement for hreflang is Google's documented behavior. No figures on the size of the Spanish-speaking search market appear here because I could not verify any to a primary source. Across the sites I operate and audit, Search Console records more than 300 million impressions a year. Verified as of August 2026.

What this changes

There is a version of this conversation that never gets past identity: understand the culture, respect the language, build trust with the community. All true, and none of it tells you what to publish.

The operational version is narrower. Open Search Console, look at which language your existing queries arrive in, and notice that it splits by topic rather than by person. That single observation replaces most of the strategy debate, and it takes twenty minutes with data you already own. The businesses that get bilingual search wrong are rarely the ones that misunderstood their customers. They are the ones that never checked what those customers type.

Frequently asked questions