In plain words

Machine translation is when a computer automatically converts text from one language to another without a human translator. Studies show that machine-translated medical instructions contain errors in some languages that could cause harm, which is why federal law gives you the right to a free, qualified human interpreter for consent forms, insurance appeals, and other high-stakes health decisions.

AI tools, including the chatbot on this site, can be genuinely helpful when you are trying to reach mental health care in the United States. They can explain what a letter says, summarize a confusing form, draft a list of questions, or help you practice a phone call before you make it. But an AI is a helper, not an authority. It is not a doctor, a therapist, an interpreter, or a lawyer, and it can be confidently wrong.

The safest pattern is to prepare with AI, then do the high-stakes step with a trusted person. Use Ask Chatbot to get ready: understand the general topic of a document, draft your questions, write your phone script. Then bring what you have prepared to a community health worker, a clinic interpreter, or a trusted navigator for anything that matters. This guide explains where AI tools are useful, where to be careful, and what to do with a person instead.

At a glance
  • Prepare with AI, then do the high-stakes step with a trusted person.
  • AI is not a doctor, interpreter, or lawyer. You have the right to a free, qualified interpreter for anything you sign.
  • Do not share your immigration status, ID numbers, or full documents. Most consumer chatbots are not covered by HIPAA.

Your privacy on LINC

LINC does not ask you to sign in or give your name, and it does not keep your conversations. When you send a message, it goes to an AI provider (Anthropic) to write the reply, and Anthropic does not use it to train its models. LINC does not collect your identity and does not share information with immigration authorities. Even so, to stay safe, do not type your immigration status, ID numbers, or full documents, and cover names and ID numbers before you share a photo of a document.

If immigration is part of your situation, this matters. Typing your visa status, case number, or identification details into any AI tool creates a digital record outside your control. Even though LINC does not store your messages, as a precaution describe your situation in general terms rather than entering specific identifiers. Use the chatbot to understand your rights and prepare your questions; bring those steps to a community health worker or legal advocate who is bound by confidentiality.

What the chatbot can help you do

These tasks are low-risk because you can catch and correct any mistake before you act on it.

The chatbot can help you
  • Draft two or three clear questions to bring to your appointment.
  • Write a phone script or email, then practice it or share it with someone you trust.
  • Get a plain-language summary of a confusing sentence or paragraph.
  • Find a verified LINC guide, an official resource, or local interpreter services.
  • Practice saying how you feel in simple sentences.
  • Turn a big worry into a short list of small steps.
  • Learn what a word or program means, like "copay," "sliding scale fee," or "referral."

Think of the chatbot as a preparation tool. It helps you get ready. A community health worker, a clinic interpreter, or a trusted provider does the high-stakes work with you.

What the chatbot must not do

There are some things an AI tool should never do for you, no matter how confident it sounds.

  • Diagnose an illness or recommend a specific medication. Only a licensed clinician who knows your full health history can do that safely.
  • Replace a professional interpreter for anything you sign or agree to. Consent forms, insurance appeals, grievance letters, and treatment plans all require a qualified human interpreter.
  • Provide mental health treatment. The chatbot can help you find care; it cannot provide therapy or counseling.
  • Be your first response in a crisis. If you are in danger or thinking about harming yourself, do not type to an AI. See the crisis section below.
  • Fill out or submit an official form for you. A chatbot can explain what a form is asking, but a person you trust should help you complete and send anything official, such as an insurance appeal or a benefits application.
  • Give you legal or immigration advice. Questions about your visa, your case, or your rights should go to a qualified legal advocate, not an AI tool.
  • Confirm a phone number, address, or eligibility rule as final. These change often, and AI can give an old or made-up answer. Always check the official source before you rely on it.

AI translation: useful, but not reliable for high-stakes wording

AI translation can give you a general idea of what a document says, but it makes more errors in some languages than others, and medical and legal wording is where those errors matter most and cause the most harm. Research that checked machine translation of emergency department discharge instructions into Spanish and Chinese found that some translations were inaccurate, and a share of the errors could have harmed a patient (Khoong et al., 2019).

How well AI translates depends a lot on the language. AI learns to translate from huge amounts of text, so it does better with languages that have a lot of text online, like Spanish, and worse with languages that have less. These are sometimes called lower-resource languages. In one study of Google Translate for emergency department discharge instructions across seven languages, the overall meaning was kept in about 82% of the translations, but accuracy ranged from about 94% in Spanish down to about 55% in Armenian; the authors concluded the tool “should not be relied on for patient instructions” (Taira et al., 2021). Newer AI tools have the same gap. A 2024 study found that Google Translate and ChatGPT translated pediatric discharge instructions about as well as professional translators for Spanish and Portuguese, but fell short for Haitian Creole, where they made more errors that could cause clinical harm (Brewster et al., 2024). A fluent-sounding answer can still be wrong, and it is more likely to be wrong in a lower-resource language.

You have the right to a free, qualified interpreter. Under Section 1557 of the Affordable Care Act (the health civil-rights law) and Title VI of the Civil Rights Act of 1964, federally funded health programs, including most hospitals, clinics, and health insurance plans, must provide a qualified interpreter at no cost to you (HHS Office for Civil Rights). You do not have to pay, and you do not have to use a family member or friend. Using an untrained person carries real risks: a study in two pediatric emergency departments found that ad hoc interpreters, such as untrained bilingual staff, made a higher share of errors with potential clinical consequences than trained professional interpreters (Flores et al., 2012). Call the clinic or health plan before your appointment and ask for a language line interpreter in your language.

For details on how to request an interpreter, see Your right to an interpreter.

Safe and unsafe ways to ask an AI tool

How you write your question changes how safe it is. A safe question describes your situation in general words. An unsafe question hands over your private details. Here are examples you can copy and change to fit your own situation.

SafeDescribe the general topic"Explain, in simple words, what this kind of letter usually means and what I should do next."
AvoidHand over private detailsPasting a letter with your name, date of birth, case number, or address, or "my visa is [type] and my case number is [number]."

Safe prompts. These describe the general topic and ask for plain-language help:

  • “I got a letter from a mental health clinic about an appointment. Can you explain, in simple words, what this kind of letter usually means and what I should do next?”
  • “Help me write three short questions to ask a counselor about starting therapy.”
  • “Write a simple phone script I can use to call a clinic and ask if they have a therapist who speaks my language.”
  • “Explain what a ‘sliding scale fee’ means in plain language.”

Unsafe prompts. Do not do these, because they share information that identifies you:

  • Pasting the full text of a letter that shows your name, date of birth, case number, or address.
  • “My name is [full name], my visa is [type], and my case number is [number]. What are my chances?”
  • “Here is my whole medical record. Tell me what is wrong with me.”
  • Uploading a photo of a document before you cover your name and ID numbers.

Notice the pattern. Safe prompts ask about the general kind of document or the general topic. Unsafe prompts give away information that points to you. You can almost always get the help you need with the general version.

Privacy: what not to share with AI

Before you type anything into an AI tool, consider whether you would be comfortable if a stranger could read it later.

Never type any of these into an AI tool, including the chatbot on this site:

  • Your immigration or visa status, your A-number, or your case number
  • Your Social Security number or any other ID number
  • Your full name together with other details that identify you
  • Your date of birth combined with your full name
  • Your home address or a phone number tied to your name
  • Bank, credit card, or payment numbers
  • Passwords or security codes
  • The full text or a full photo of an official document before you cover the private parts

The points below explain the safe way to get help while keeping this information to yourself.

  • Do not enter your immigration status, visa type, case number, or Social Security number into any AI tool, including the chatbot on this site.
  • Do not paste a full document into a chat window. If you want to understand a letter, describe what it is about instead of copying the full text.
  • If you want a document explained, sharing a photo can be safer than typing out all your personal details, but only with the sensitive parts covered first. Before you take the photo, place a small piece of paper over your name, date of birth, ID numbers, and case numbers. This way the chatbot can explain what the document is about without you entering that information as text.
  • Most consumer AI chatbots are not covered by HIPAA (a US health privacy law). HIPAA protects your medical records at a hospital, clinic, or health plan, but it does not automatically cover general-purpose AI apps and chatbots. The Federal Trade Commission has explained that many health and wellness apps can collect, share, or sell your information in ways HIPAA does not restrict, and it offers a tool that walks through which laws apply to a given health app. The FTC’s updated Health Breach Notification Rule (2024) makes clear that health app makers must notify users after a breach, but that rule addresses what happens after a breach, not day-to-day data sharing.
  • Some apps store the text you type and may use it to improve future models. Read the privacy policy of any AI tool before you type sensitive information into it.

For more on your medical privacy rights, see Your privacy and medical records.

Be careful with AI chatbots and AI therapy apps

Some apps and chatbots are sold as a way to get therapy or mental health support from AI. Be careful with these. They can feel warm and like they understand you, but they are not a trained therapist, and they are not a substitute for one.

In November 2025, the American Psychological Association warned people not to use AI chatbots and wellness apps in place of care from a qualified mental health professional. The APA said these tools do not yet have the science or the rules to keep users safe, and that their ability to help someone in a crisis is “limited and unpredictable” (American Psychological Association, 2025a).

The APA also names specific risks. Unlike a trained therapist, a chatbot tends to agree with you and keep affirming you, even when you say something that is harmful or untrue about yourself. Some chatbots have claimed to be licensed therapists when they are not. In the worst cases, a chatbot can keep a person from reaching real human help, or even encourage self-harm (American Psychological Association, 2025b).

This does not mean AI is useless for your mental health. You can still use the chatbot on this site to find care, understand a letter, or get ready for an appointment. The rule is simple: use AI to reach a person who can help you, and let a licensed professional do the actual counseling.

Never use AI in a crisis

If you are in danger or thinking about harming yourself, do not talk to an AI. Go directly to a person.

If you are in crisis

Do not type to an AI. Call or text 988, the Suicide and Crisis Lifeline: free, confidential, 24 hours a day, with interpreters in more than 240 languages. Call 911 for immediate physical danger.

  • Call or text 988 (the Suicide and Crisis Lifeline) (SAMHSA). It is free, confidential, and available 24 hours a day, 7 days a week. For Spanish, press 2 when you call, or text the word “Ayuda” to 988. For other languages, stay on the line and ask for an interpreter; 988 connects callers with interpreters in more than 240 languages through Language Line Solutions, typically within about 20 seconds (988 Lifeline). If you are deaf or hard of hearing, you can reach 988 in American Sign Language through its videophone option, or use TTY by dialing 711 then 988. Chat is also available in English and Spanish at 988lifeline.org.
  • Call 911 for immediate physical danger.

The chatbot on this site is built to recognize crisis messages and send you directly to 988 and to emergency resources. But do not wait for that. If you are in crisis, go to 988 first.

For local and national crisis resources, see If you are in crisis.

How to check whether an AI answer is right

AI tools can sound certain and still be wrong. This is especially true for specific facts, phone numbers, addresses, and eligibility rules that change over time. A good habit can protect you.

After the chatbot gives you an answer, check it against one official source before you act. Ask yourself: Does the answer point to a specific .gov website? If you go to that website, does it actually say what the chatbot said? For treatment and provider information, the official SAMHSA locator at FindTreatment.gov is a reliable place to confirm. For general health information, MedlinePlus from the National Library of Medicine provides free guidance on how to judge whether health information is trustworthy.

Watch for answers that sound very confident but do not name a specific source you can check. That is a sign to ask a community health worker or a clinic navigator to confirm before you make a decision based on it. For a fuller look at why AI produces confident wrong answers and a simple habit for checking them, see AI gets things wrong.

Try it now. The next time the chatbot gives you an answer that matters, reply with one line: "Where does this come from, and what should I check to be sure?" Then say the answer back in your own words and ask, "Did I get that right?" These two short follow-ups are the whole checking habit.

MedlinePlus "Evaluating Health Information" page from the U.S. National Library of Medicine, with steps for judging whether health information is trustworthy (medlineplus.gov, captured June 30, 2026).

Prepare with AI, then act with a trusted person

The safest pattern for using AI to navigate mental health care is this: prepare with AI, then do the high-stakes step with a real person.

Use Ask Chatbot to get ready: draft your questions, understand the general topic of a document, write your phone script. Then bring what you have prepared to a community health worker, a clinic interpreter, or a trusted navigator. Let that person review it with you, make the call alongside you, or sit with you when you sign a form.

This approach gives you the benefit of AI as a language tool and a practice partner while keeping a qualified human being in the loop for every decision that matters. No AI tool should be the last check before a high-stakes health decision.

Sources

LINC is a research prototype, not a medical or legal service. Translation can be wrong; you have the right to a free, qualified interpreter. For anything high-stakes, do it with a community health worker or someone you trust. In a crisis, call or text 988, or 911 for immediate danger.