A Translated Front Door
A website can be accurately translated and yet provide a worse experience. The words would have arrived, but the journey wouldn't have.
Patients who prefer Spanish report waiting longer, finding it more difficult to book appointments, and having more trouble understanding clinic staff, even at clinics with Spanish-language websites. Some aspect of the translation must be failing, while another aspect that should pass through is instead being lost.
The Spanish-language page states "agenda en línea" (schedule online). The words are correct and accurate. The translation is right. When the patient clicks through to the scheduling tool, they find that it has not been localized. The labels on the form are in English, and the error messages are as well. The confirmation email is sent in English. Likewise, the automated reminder call, when the patient goes all the way through to a booked appointment, is in English.
It is not a case of a translation failure; the translation has carried out precisely what it was meant to do. This is a flaw in system design based on the mistaken idea that translating the entry point of a process is equivalent to translating the process itself. The patient who needed Spanish to begin the journey still needs it to complete it. What the organization has constructed is a front door that has been translated and a house that accepts only English.
The difference between linguistic accuracy and functional equivalence is something the traditional multilingual-web checklist fails to account for: translated pages are counted, but completed tasks are counted differently, and most organizations do not count the latter.
What the Evidence Shows About the Gap
A 2025 systematic review of the experiences of patients who prefer Spanish found that language-concordant care—that is, having a provider who speaks the patient's language—was generally associated with higher-quality communication. However, the effects varied and were inconsistent across studies. The presence of a Spanish-speaking provider in the room does not automatically eliminate the gap in the patient experience.
The interpreter's findings were more specific: they showed that patient experience improved only when the interpreter's quality was high; low-quality interpretations were linked to poorer outcomes than no interpretation at all. It is just as important whether the mechanism is present as whether it works; simply ticking the box indicating that an interpreter was provided does not indicate that the patient was actually able to go through with the decision they had to make.
The finding regarding the disparity is the one that deserves focus: patients who preferred Spanish reported worse experiences, particularly with appointments, wait times, and understanding from clinic staff. This is not a failure in communication at the level of vocabulary; it is a failure of the system at the level of access; that is, it relates to the stages in the care process where language accuracy was present but functional equivalence was not.
The translation considers whether the words have arrived, whereas parity examines whether the patient can still complete the journey. The two questions are not the same, and most multilingual web audits address only the first.
The Semantic Parity Test
Semantic parity is not the same as translation; it holds that language access should be judged not by whether words exist in another language, but by whether the same decision, confidence, and action can be made in that language.
The distinction produces four levels that most multilingual experiences partially move through and stop at:
Translation
The words, when expressed in another language, are where most organizations conclude the task is complete.
Comprehension
Comprehension involves a patient who speaks the target language being able to grasp the meaning of the words. Translation may fail to achieve proper understanding even if it is technically accurate because it is contextually incorrect—for example, when a clinical term is translated literally into a language where a different expression would be understood, or when a tone comes across as formal and distant in a patient relationship that requires warmth.
Actionability
Actionability means that a patient can act based on what they have understood; in this respect, the translated front door and the English-only house become apparent. The instruction "Schedule online" is understandable in Spanish, but if the scheduling tool it links to is not, the instruction is not actionable.
Parity
The idea is that no matter which language the patient uses when going through the system, their experience, including the confidence they have when making decisions, the extent of the information they receive, and how easy the process is, should be the same. Currently, almost no health center website meets this standard, and almost none are being checked to see whether they do.
The four stages are translation, comprehension, actionability, and parity. While most multilingual health center websites have achieved the first stage, a few have reached the second, and almost none have been created with the fourth in mind.
What a Parity Audit Looks Like
The usual web audit that deals with multiple languages counts the number of pages. For example, how many of the site's pages are offered in Spanish? What percentage of the content has been translated? These are merely indicators of access; they show the availability of different languages, not the quality of the experience this provides.
A parity audit involves testing various tasks. Select the five things that a new patient would most need to do on your website:
Find out about the services
Locate a branch
Understand what to bring to an appointment
Find out if they are eligible for the services
Book an appointment.
Carry out each of these tasks in English and then do them in Spanish. Record all instances in which the Spanish version results in a different outcome, such as a broken form, a confirmation page that reverts to English, a phone number that connects to a staff member who does not speak Spanish, or an error message the patient cannot understand. The difference between the two pathways is known as the parity gap.
The parity issue isn't something that can be addressed solely by improving web design; it is a problem of system design. If a website manages the Spanish-language scheduling process properly but then passes the task to a phone system that can't handle it, there is a parity failure at the handoff. In the case of a patient-facing portal offered in Spanish yet providing clinical summaries in English only, there is a parity failure at the document level. These kinds of failures cannot be detected by looking at the number of pages; they can only be seen when the task is actually carried out and where it breaks is measured.
Why This Matters Beyond Compliance
The provision of language access in federally qualified health centers is a matter of compliance; both Title VI of the Civil Rights Act and the ACA mandate meaningful access for patients who have limited English proficiency. Health centers have a duty to meet this standard, and most have developed systems which they believe meet it.
The semantic parity argument is not primarily about compliance; it is rather a trust argument. When a patient encounters a translated front door and an English-only scheduling tool, they can infer the organization's attitude towards them. This understanding does not come from the words, since the words were accurate, but from the experience of feeling that the Spanish page was created to meet a requirement rather than to serve the patient. This distinction is perceived before it is put into words, and it affects whether or not the patient returns.
Health centers that provide services in communities where a large part of the patient population prefers Spanish, Haitian Creole, Vietnamese, or Somali are working to earn the trust of patients who have good reason to be cautious when approaching healthcare organizations. The parity gap is not simply a neutral technical malfunction; it signals which group's experience the organization was designed to meet and which group it was built for.
Filling the gap is not the result of a single initiative; it is an ongoing audit process that involves testing the tasks, identifying gaps, correcting handoffs, and then testing again. The organizations that do this on purpose are creating something that the number of translated pages cannot measure: a patient experience that is, in fact, equivalent, not merely linguistically adequate.
The fact that a page has been translated proves that someone carried out the work. Completing the task in the patient's own language is evidence that the work was actually done for the patient. Yet only one of these two points closes the gap.
Some ideas are worth discussing in the context of your organization.
